Cursor
Desktop app for Mac, Windows and Linux. Agents run on your computer, on your own server or in the vendor's cloud.
Product websitecursor.com Compare with others Missing or wrong data?
- Desktop app
- Mac, Windows and Linux
- Browser
- Yes
- Phone
- IPhone and phone browser
- Command line
- Yes
- Agents
- Cursor Agent
- Your subscription
- Not listed
- Agents run
- On your computer, on your own server and in the vendor's cloud
- Laptop closed
- Keeps going
- Isolation
- Worktree and VM
- Parallel agents
- Yes
- Scheduled runs
- Yes
- Coordinator agent
- Yes
- Built-in issues
- Yes
- Pull requests
- Yes
- Review screen
- Yes
- Team
- Shared sessions, roles and SSO
- Self-hosting
- No
- Licence
- Proprietary
- Price from
- Free
- How it charges
- Free plan, subscription, per seat, usage-based and bring your own key
- Paid plans from
- $10 or less/month
Receipts
- Checked
- · oldest answer 2026-10-09
- Findings
- 50 published · 44 cross-checked by a second agent · read them all
- Sources
- 101 public pages on 3 sites: cursor.com, github.com, help.openai.com
- Screenshot
- docs.cursor.com, retrieved · first party docs · shows appearance, not capability
- Data
- /products/cursor.json · .md
- Method
- Same rules for all 103 researched ADEs. How we check · Report wrong data
What it does
Where you use it
Cursor is reachable from a desktop app (Mac, Windows, Linux), a CLI, a web client at cursor.com/agents, a native iPhone/iPad app (iOS/iPadOS 26+) and chat/issue integrations. There is no native Android app (documented as planned); on Android the docs direct users to install the cursor.com/agents web client as a PWA from Chrome.
Source and scope for Where you use itAgents it runs
Cursor runs its own coding-agent loop with selectable models. OpenAI also distributes a separate Codex IDE extension installable inside the editor; that extension does not establish Cursor-native agent/session integration or a Cloud Agents harness option.
Source and scope for Agents it runsAccounts and subscriptions
A Cursor account uses the new X Subscription or a retained legacy Cursor plan. Own model API keys are a separate documented route; third-party agent subscription reuse remains unestablished.
Source and scope for Accounts and subscriptionsWhere agents run
Work can run on the local machine, on remote SSH hosts, on Cursor-managed VMs, or on self-managed workers; managed cloud runs continue independently of the laptop.
Source and scope for Where agents runHow work is kept apart
Local agents use worktrees and an OS sandbox; cloud agents run in per-agent Firecracker microVMs with configurable network egress and redacted secrets.
Source and scope for How work is kept apartParallel work
Many agents can run in parallel across local worktrees, remote SSH machines, self-managed machines and the cloud; no numeric cloud concurrency limit is stated.
Source and scope for Parallel workScheduled runs
Automations run on schedules, source-control events, Slack, Linear, Sentry, PagerDuty and webhooks; Projects can subscribe to PRs, CI and schedules.
Source and scope for Scheduled runsGit and pull requests
Cloud agents push branches, make signed commits, open draft PRs and respond to PR comments; local work uses worktrees.
Source and scope for Git and pull requestsLicence and source
The Terms grant a limited right to access and use the Service and reserve all rights not granted; the public cursor/cursor repository holds no product source.
Source and scope for Licence and sourcePrice
New subscriptions cost $8 Plus, $30 Premium, $100 Super or $200 Ultra monthly; Teams seats $30/$100/$200. Legacy rates continue for existing accounts. Third-party models are billed separately.
Source and scope for PriceTyped filters
Every value links to its receipt below. Read plan and surface restrictions before using a filter to decide.
| Filter | Value |
|---|---|
| Desktop OS: macOS | Yes cross-checked |
| Desktop OS: Windows | Yes cross-checked |
| Desktop OS: Linux | Yes cross-checked |
| Web client | Yes cross-checked |
| Phone access: Native iOS | Yes cross-checked |
| Phone access: Native Android | No cross-checked |
| Phone access: Phone browser | Yes cross-checked |
| CLI | Yes cross-checked |
| Coding agents: Claude Code | Unknown |
| Coding agents: Codex | Unknown |
| Coding agents: Cursor Agent | Yes |
| Coding agents: Gemini CLI | Unknown |
| Coding agents: OpenCode | Unknown |
| Coding agents: Pi | Unknown |
| Coding agents: Grok | Unknown |
| Coding agents: Aider | Unknown |
| Coding agents: Antigravity | Unknown |
| Use my subscription | Unknown |
| Where agents run: On your computer | Yes cross-checked |
| Where agents run: On your own server | Yes cross-checked |
| Where agents run: In the vendor's cloud | Yes cross-checked |
| Workspace isolation: Git worktree | Yes cross-checked |
| Workspace isolation: Container | Unknown |
| Workspace isolation: VM or microVM | Yes cross-checked |
| Parallel runs | Yes cross-checked |
| Coordinator | Yes cross-checked |
| Native issue tracking | Yes |
| Unattended triggers | Yes cross-checked |
| Shared prompting | Yes cross-checked |
| Team roles | Yes cross-checked |
| SSO | Yes cross-checked |
| Pull request workflow | Yes cross-checked |
| Review interface | Yes cross-checked |
| Product license: Apache-2.0 | No. Proprietary cross-checked |
| Product license: MIT | No. Proprietary cross-checked |
| Product license: GPL-3.0 | No. Proprietary cross-checked |
| Product license: GPL-3.0-or-later | No. Proprietary cross-checked |
| Product license: AGPL-3.0 | No. Proprietary cross-checked |
| Product license: AGPL-3.0-or-later | No. Proprietary cross-checked |
| Product license: Elastic License 2.0 | No. Proprietary cross-checked |
| Product license: Custom terms | No. Proprietary cross-checked |
| Product license: Proprietary | Yes. Proprietary cross-checked |
| Self-hosting | No |
| Telemetry opt-out | Unknown |
| Free access | Yes cross-checked |
| Starting price: Free | Yes. Free cross-checked |
| Starting price: Up to $20/month | No. Free cross-checked |
| Starting price: $21–40/month | No. Free cross-checked |
| Starting price: $41–100/month | No. Free cross-checked |
| Starting price: Over $100/month | No. Free cross-checked |
| Starting price: Contact for pricing | No. Free cross-checked |
| Pricing model: Free, no paid plan | No |
| Pricing model: Free plan | Yes cross-checked |
| Pricing model: Subscription | Yes cross-checked |
| Pricing model: Per seat | Yes cross-checked |
| Pricing model: Usage-based | Yes cross-checked |
| Pricing model: Per machine or compute | No |
| Pricing model: Bring your own key | Yes cross-checked |
| Pricing model: One-time purchase | No |
| Paid plans from: No paid plan | No. $10 or less/month |
| Paid plans from: $10 or less/month | Yes. $10 or less/month |
| Paid plans from: $11–20/month | No. $10 or less/month |
| Paid plans from: $21–40/month | No. $10 or less/month |
| Paid plans from: $41–100/month | No. $10 or less/month |
| Paid plans from: Over $100/month | No. $10 or less/month |
| Paid plans from: Contact sales | No. $10 or less/month |
| Lifecycle status: Active | Unknown |
| Lifecycle status: Sunsetting | Unknown |
| Keeps going with laptop closed | Yes cross-checked |
| Open-source license | No. Proprietary cross-checked |
| Any desktop app | Yes cross-checked |
| Paid plan up to $20/month | Yes. $10 or less/month |
| Paid plan up to $40/month | Yes. $10 or less/month |
| Any phone client | Yes cross-checked |
Filter receipts
Desktop OS: macOS: Yes cross-checked
Desktop OS: macOS
Typed value: Yes.
Cursor is reachable from a desktop app (Mac, Windows, Linux), a CLI, a web client at cursor.com/agents, a native iPhone/iPad app (iOS/iPadOS 26+) and chat/issue integrations. There is no native Android app (documented as planned); on Android the docs direct users to install the cursor.com/agents web client as a PWA from Chrome.
- Desktop os: Desktop app for Mac, Windows and Linux; no OS versions listed.
- No native Android app is documented; "Android is planned" with no timing. The Android PWA route is documented only on the Cloud Agents page, not the mobile app page, and was not installed or run.
- iOS app requires iOS/iPadOS 26 or later; English only.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Desktop app for Mac, Windows and Linux; Cursor account required.
Desktop OS: Windows: Yes cross-checked
Desktop OS: Windows
Typed value: Yes.
Cursor is reachable from a desktop app (Mac, Windows, Linux), a CLI, a web client at cursor.com/agents, a native iPhone/iPad app (iOS/iPadOS 26+) and chat/issue integrations. There is no native Android app (documented as planned); on Android the docs direct users to install the cursor.com/agents web client as a PWA from Chrome.
- Desktop os: Desktop app for Mac, Windows and Linux; no OS versions listed.
- No native Android app is documented; "Android is planned" with no timing. The Android PWA route is documented only on the Cloud Agents page, not the mobile app page, and was not installed or run.
- iOS app requires iOS/iPadOS 26 or later; English only.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Desktop app for Mac, Windows and Linux; Cursor account required.
Desktop OS: Linux: Yes cross-checked
Desktop OS: Linux
Typed value: Yes.
Cursor is reachable from a desktop app (Mac, Windows, Linux), a CLI, a web client at cursor.com/agents, a native iPhone/iPad app (iOS/iPadOS 26+) and chat/issue integrations. There is no native Android app (documented as planned); on Android the docs direct users to install the cursor.com/agents web client as a PWA from Chrome.
- Desktop os: Desktop app for Mac, Windows and Linux; no OS versions listed.
- No native Android app is documented; "Android is planned" with no timing. The Android PWA route is documented only on the Cloud Agents page, not the mobile app page, and was not installed or run.
- iOS app requires iOS/iPadOS 26 or later; English only.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Desktop app for Mac, Windows and Linux; Cursor account required.
Web client: Yes cross-checked
Web client
Typed value: Yes.
Cursor is reachable from a desktop app (Mac, Windows, Linux), a CLI, a web client at cursor.com/agents, a native iPhone/iPad app (iOS/iPadOS 26+) and chat/issue integrations. There is no native Android app (documented as planned); on Android the docs direct users to install the cursor.com/agents web client as a PWA from Chrome.
- Mobile apps: Native app shipped for iPhone (iOS 26.0+) and iPad (iPadOS 26.0+), English only. No native Android app: the mobile page says "Android is planned"; the Cloud Agents page separately directs Android users to install cursor.com/agents as a PWA from Chrome (Install App).
- Browser: cursor.com/agents web client.
- Mobile browser: Web client usable on any device, including Android as a PWA.
- No native Android app is documented; "Android is planned" with no timing. The Android PWA route is documented only on the Cloud Agents page, not the mobile app page, and was not installed or run.
- iOS app requires iOS/iPadOS 26 or later; English only.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Clients (iOS, web, desktop, Slack, GitHub/Bitbucket comments, Linear, API); paid plan required; managed VMs; environments; no stated concurrency limit; team visibility read-only unless admin enables follow-ups; billed at model API pricing; spend limit set at first use; screenshots, videos and logs as artifacts; draft/merge-ready PRs; Android: open cursor.com/agents in Chrome and tap Install App for a PWA; links to the native iOS app page.
Phone access: Native iOS: Yes cross-checked
Phone access: Native iOS
Typed value: Yes.
Cursor is reachable from a desktop app (Mac, Windows, Linux), a CLI, a web client at cursor.com/agents, a native iPhone/iPad app (iOS/iPadOS 26+) and chat/issue integrations. There is no native Android app (documented as planned); on Android the docs direct users to install the cursor.com/agents web client as a PWA from Chrome.
- Mobile apps: Native app shipped for iPhone (iOS 26.0+) and iPad (iPadOS 26.0+), English only. No native Android app: the mobile page says "Android is planned"; the Cloud Agents page separately directs Android users to install cursor.com/agents as a PWA from Chrome (Install App).
- Browser: cursor.com/agents web client.
- Mobile browser: Web client usable on any device, including Android as a PWA.
- Task actions: Mobile app: start agents on a Cloud machine, Team Pool or My Machine, send follow-ups, review diffs and checks, ask agents to fix failing checks, merge PRs.
- Notifications: Mobile push notification when an agent finishes a turn; Live Activities for up to eight agents; Slack status reactions and completion messages.
- Cross device continuity: Agents started on desktop, web or mobile appear across clients; Move to Cloud and /remote-control (Cursor 3.9.8+) continue local sessions from the phone. Remote Control must be turned on under Settings > Agents, and on Teams and Enterprise a team admin enables it.
- No native Android app is documented; "Android is planned" with no timing. The Android PWA route is documented only on the Cloud Agents page, not the mobile app page, and was not installed or run.
- iOS app requires iOS/iPadOS 26 or later; English only.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Clients (iOS, web, desktop, Slack, GitHub/Bitbucket comments, Linear, API); paid plan required; managed VMs; environments; no stated concurrency limit; team visibility read-only unless admin enables follow-ups; billed at model API pricing; spend limit set at first use; screenshots, videos and logs as artifacts; draft/merge-ready PRs; Android: open cursor.com/agents in Chrome and tap Install App for a PWA; links to the native iOS app page.
-
Finding: iPhone/iPad app (iOS/iPadOS 26+): start agents, follow-ups, PR review and merge actions, push notifications, Live Activities, Remote Control of local sessions; "Android is planned" with no timing; available on Start, Pro, Pro+, Ultra, Teams, Enterprise.
Phone access: Native Android: No cross-checked
Phone access: Native Android
Typed value: No.
Cursor is reachable from a desktop app (Mac, Windows, Linux), a CLI, a web client at cursor.com/agents, a native iPhone/iPad app (iOS/iPadOS 26+) and chat/issue integrations. There is no native Android app (documented as planned); on Android the docs direct users to install the cursor.com/agents web client as a PWA from Chrome.
- Mobile apps: Native app shipped for iPhone (iOS 26.0+) and iPad (iPadOS 26.0+), English only. No native Android app: the mobile page says "Android is planned"; the Cloud Agents page separately directs Android users to install cursor.com/agents as a PWA from Chrome (Install App).
- Task actions: Mobile app: start agents on a Cloud machine, Team Pool or My Machine, send follow-ups, review diffs and checks, ask agents to fix failing checks, merge PRs.
- Notifications: Mobile push notification when an agent finishes a turn; Live Activities for up to eight agents; Slack status reactions and completion messages.
- Cross device continuity: Agents started on desktop, web or mobile appear across clients; Move to Cloud and /remote-control (Cursor 3.9.8+) continue local sessions from the phone. Remote Control must be turned on under Settings > Agents, and on Teams and Enterprise a team admin enables it.
- No native Android app is documented; "Android is planned" with no timing. The Android PWA route is documented only on the Cloud Agents page, not the mobile app page, and was not installed or run.
- iOS app requires iOS/iPadOS 26 or later; English only.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: iPhone/iPad app (iOS/iPadOS 26+): start agents, follow-ups, PR review and merge actions, push notifications, Live Activities, Remote Control of local sessions; "Android is planned" with no timing; available on Start, Pro, Pro+, Ultra, Teams, Enterprise.
Phone access: Phone browser: Yes cross-checked
Phone access: Phone browser
Typed value: Yes.
Cursor is reachable from a desktop app (Mac, Windows, Linux), a CLI, a web client at cursor.com/agents, a native iPhone/iPad app (iOS/iPadOS 26+) and chat/issue integrations. There is no native Android app (documented as planned); on Android the docs direct users to install the cursor.com/agents web client as a PWA from Chrome.
- Mobile apps: Native app shipped for iPhone (iOS 26.0+) and iPad (iPadOS 26.0+), English only. No native Android app: the mobile page says "Android is planned"; the Cloud Agents page separately directs Android users to install cursor.com/agents as a PWA from Chrome (Install App).
- Browser: cursor.com/agents web client.
- Mobile browser: Web client usable on any device, including Android as a PWA.
- No native Android app is documented; "Android is planned" with no timing. The Android PWA route is documented only on the Cloud Agents page, not the mobile app page, and was not installed or run.
- iOS app requires iOS/iPadOS 26 or later; English only.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Clients (iOS, web, desktop, Slack, GitHub/Bitbucket comments, Linear, API); paid plan required; managed VMs; environments; no stated concurrency limit; team visibility read-only unless admin enables follow-ups; billed at model API pricing; spend limit set at first use; screenshots, videos and logs as artifacts; draft/merge-ready PRs; Android: open cursor.com/agents in Chrome and tap Install App for a PWA; links to the native iOS app page.
CLI: Yes cross-checked
CLI
Typed value: Yes.
Cursor is reachable from a desktop app (Mac, Windows, Linux), a CLI, a web client at cursor.com/agents, a native iPhone/iPad app (iOS/iPadOS 26+) and chat/issue integrations. There is no native Android app (documented as planned); on Android the docs direct users to install the cursor.com/agents web client as a PWA from Chrome.
- Cli: Cursor CLI (agent) on macOS, Linux, WSL and Windows PowerShell.
- No native Android app is documented; "Android is planned" with no timing. The Android PWA route is documented only on the Cloud Agents page, not the mobile app page, and was not installed or run.
- iOS app requires iOS/iPadOS 26 or later; English only.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: CLI on macOS, Linux, WSL and Windows; agent, plan and ask modes; print mode; "&" prefix hands a conversation to a Cloud Agent; session resume; sandbox controls.
Coding agents: Cursor Agent: Yes
Coding agents: Cursor Agent
Typed value: Yes.
Cursor runs its own coding-agent loop with selectable models. OpenAI also distributes a separate Codex IDE extension installable inside the editor; that extension does not establish Cursor-native agent/session integration or a Cloud Agents harness option.
- Harnesses: Cursor supplies its own coding-agent loop, orchestrating instructions, tools and a chosen model. Other coding-agent harnesses were not established.
- External coding harnesses remain unknown without Cursor-native session/launcher integration. The optional Codex extension uses its own UI, authentication and agent loop.
-
Finding: Cursor Agent orchestrates model-specific instructions, file/search/terminal tools and a selected model. This establishes the Cursor harness itself, not external agent loops.
Where agents run: On your computer: Yes cross-checked
Where agents run: On your computer
Typed value: Yes.
Work can run on the local machine, on remote SSH hosts, on Cursor-managed VMs, or on self-managed workers; managed cloud runs continue independently of the laptop.
- Local: Local agents in the editor and CLI, optionally in worktrees.
- Byo remote: Remote SSH environments in the Agents Window and My Machines workers.
- Hybrid: Cloud-hosted agent loop with tool execution on own machines, and handoff between local and cloud.
- Default VM resources are limited; larger VMs need Enterprise support and custom resource configuration is "coming soon".
- Team Pools require Enterprise; My Machines plan requirement not stated.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Workspace for parallel agents across local, cloud and remote SSH; worktree isolation; handoff cloud to local and back; /in-cloud and /autopilot cloud subagents.
-
Finding: Isolated Git checkouts in Agents Window; /worktree and /best-of-n in IDE; .cursor/worktrees.json setup scripts; default maximum 25 worktrees per machine with automatic cleanup.
Where agents run: On your own server: Yes cross-checked
Where agents run: On your own server
Typed value: Yes.
Work can run on the local machine, on remote SSH hosts, on Cursor-managed VMs, or on self-managed workers; managed cloud runs continue independently of the laptop.
- Local: Local agents in the editor and CLI, optionally in worktrees.
- Byo remote: Remote SSH environments in the Agents Window and My Machines workers.
- Self hosted: Self-Hosted Machines run tools on user infrastructure; the agent loop and inference stay in Cursor's cloud.
- Hybrid: Cloud-hosted agent loop with tool execution on own machines, and handoff between local and cloud.
- Provisioning: Cursor provisions managed VMs; self-hosted pools use a controller that starts machines on demand, with Kubernetes and partner-host templates.
- Payer: Managed Cloud Agents are billed at model API pricing with infrastructure included; self-hosted users pay their own machines.
- Default VM resources are limited; larger VMs need Enterprise support and custom resource configuration is "coming soon".
- Team Pools require Enterprise; My Machines plan requirement not stated.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Tool execution on own machine while agent loop and inference stay in Cursor cloud; Team Pools need Enterprise; up to 200 workers per user and 1,000 per team; outbound HTTPS only; partner hosts such as AWS Lambda, Modal, Vercel; user pays for own infrastructure.
-
Finding: Workspace for parallel agents across local, cloud and remote SSH; worktree isolation; handoff cloud to local and back; /in-cloud and /autopilot cloud subagents.
Where agents run: In the vendor's cloud: Yes cross-checked
Where agents run: In the vendor's cloud
Typed value: Yes.
Work can run on the local machine, on remote SSH hosts, on Cursor-managed VMs, or on self-managed workers; managed cloud runs continue independently of the laptop.
- Managed cloud: Managed isolated VMs (Firecracker microVMs).
- Provisioning: Cursor provisions managed VMs; self-hosted pools use a controller that starts machines on demand, with Kubernetes and partner-host templates.
- Payer: Managed Cloud Agents are billed at model API pricing with infrastructure included; self-hosted users pay their own machines.
- Expiry: Workspace recycled after idle (follow-ups reset the timer); VM snapshots kept 90 days of inactivity.
- Client disconnect: Local machine need not stay connected to a managed cloud run.
- Other device continuation: Runs started on one client continue from web or mobile.
- Default VM resources are limited; larger VMs need Enterprise support and custom resource configuration is "coming soon".
- Team Pools require Enterprise; My Machines plan requirement not stated.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Clients (iOS, web, desktop, Slack, GitHub/Bitbucket comments, Linear, API); paid plan required; managed VMs; environments; no stated concurrency limit; team visibility read-only unless admin enables follow-ups; billed at model API pricing; spend limit set at first use; screenshots, videos and logs as artifacts; draft/merge-ready PRs; Android: open cursor.com/agents in Chrome and tap Install App for a PWA; links to the native iOS app page.
-
Finding: Firecracker microVM per agent in a separate AWS account; workspace recycled after idle; snapshots kept rolling 90 days inactive; conversation state kept indefinitely unless deleted via API; Privacy Mode (not Legacy); commands auto-run; draft PRs; HSM-signed commits; SOC 2 Type 2 statement; region not stated.
Workspace isolation: Git worktree: Yes cross-checked
Workspace isolation: Git worktree
Typed value: Yes.
Local agents use worktrees and an OS sandbox; cloud agents run in per-agent Firecracker microVMs with configurable network egress and redacted secrets.
- Worktree: Worktrees in the Agents Window and /worktree in the IDE.
- Isolation was not tested.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Isolated Git checkouts in Agents Window; /worktree and /best-of-n in IDE; .cursor/worktrees.json setup scripts; default maximum 25 worktrees per machine with automatic cleanup.
Workspace isolation: VM or microVM: Yes cross-checked
Workspace isolation: VM or microVM
Typed value: Yes.
Local agents use worktrees and an OS sandbox; cloud agents run in per-agent Firecracker microVMs with configurable network egress and redacted secrets.
- Vm: Firecracker-based microVM per agent in a separate AWS account.
- Secrets: Runtime Secrets are injected as environment variables and stripped from transcripts, tool output and commits.
- Code separation: "One agent cannot see another agent's code, environment, or state."
- Security boundary: VM boundary; Cloud Agent VMs sit in a separate AWS account from Cursor's other production infrastructure.
- Isolation was not tested.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Firecracker microVM per agent in a separate AWS account; workspace recycled after idle; snapshots kept rolling 90 days inactive; conversation state kept indefinitely unless deleted via API; Privacy Mode (not Legacy); commands auto-run; draft PRs; HSM-signed commits; SOC 2 Type 2 statement; region not stated.
Parallel runs: Yes cross-checked
Parallel runs
Typed value: Yes.
Many agents can run in parallel across local worktrees, remote SSH machines, self-managed machines and the cloud; no numeric cloud concurrency limit is stated.
- Concurrent sessions: Agents Window runs many parallel agents across local, cloud and remote SSH.
- Status visibility: Agents Window and the mobile app list agents; Live Activities track up to eight agents at once.
- Docs state "run as many agents as you want in parallel" without a number; spend limits and machine capacity still apply.
- Observed concurrency was not tested.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Workspace for parallel agents across local, cloud and remote SSH; worktree isolation; handoff cloud to local and back; /in-cloud and /autopilot cloud subagents.
Coordinator: Yes cross-checked
Coordinator
Typed value: Yes.
Subagents (built-in and custom) and Project coordinators delegate work with isolated context and return final summaries; they can be resumed by ID.
- Roles: Built-in Explore, Bash and Browser subagents plus custom subagents; Projects have a coordinator that does not write code.
- Model harness effort: Each subagent can set its own model; model options such as effort are written in brackets in the model field.
- Messages: Parent prompts the subagent; users can talk to Project agents directly.
- Results return: Subagent returns a final message; intermediate output stays in its context.
- Orchestration owner: Parent agent or Project coordinator.
- Activation: Automatic by description, explicit by name, /in-cloud and /autopilot for cloud subagents.
- Live progress: Background subagents write their output to ~/.cursor/subagents/. On mobile, tapping a subagent card reads its child transcript (mobile-current).
- Followup: Passing the agent ID resumes a subagent with context preserved.
- Invocation: Foreground, background or parallel invocation.
- Cross-harness delegation was not found; subagent definitions from .claude/.codex are read but run in Cursor.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: iPhone/iPad app (iOS/iPadOS 26+): start agents, follow-ups, PR review and merge actions, push notifications, Live Activities, Remote Control of local sessions; "Android is planned" with no timing; available on Start, Pro, Pro+, Ultra, Teams, Enterprise.
-
Finding: Coordinator agent plans and delegates to cloud agents, shared context files sync across machines, subscriptions to Slack/PR/CI/schedules, runs in the cloud; paid plans only, not Privacy Mode (Legacy). Dependencies between tasks, plan view and cost visibility not described.
-
Finding: Live TypeScript SDK page lists no Python SDK page; it points to an SDK Bridge for other languages. Stored ACC-05 sdk note reads 'a Python SDK page is listed but not read'.
-
Finding: Built-in Explore/Bash/Browser and custom subagents (also read from .claude/.codex agent folders); per-subagent model; foreground, background and parallel; final summary returned; resumable by ID; cloud subagents via /in-cloud.
Native issue tracking: Yes
Native issue tracking
Typed value: Yes.
Projects has a persistent shared task list and board, distinct from member coordinator chats and worker agents. Projects requires Cloud Agents access; shared Projects requires a team account and Cursor 3.26.13 or later.
- Project epic views: Project members share a task list and board.
- Task session separation: Members have separate coordinator/worker chats but work from the same Project tasks.
- Metadata: The page establishes a shared list and board without enumerating all issue metadata or lifecycle operations.
- First-party documentation review only; no installed-product, account or multi-user scenario was executed. Unanswered atomic questions remain open.
-
Finding: Projects maintains work over months and exposes a shared task list and board distinct from each member own coordinator/workers. Team members work from the same tasks. Sharing needs a team account and Cursor 3.26.13+; Projects requires a Cloud Agents plan (Premium+, Teams, Enterprise or legacy paid plans).
Unattended triggers: Yes cross-checked
Unattended triggers
Typed value: Yes.
Automations run on schedules, source-control events, Slack, Linear, Sentry, PagerDuty and webhooks; Projects can subscribe to PRs, CI and schedules.
- Schedules: Preset or cron schedules.
- Events: GitHub, GitLab, Bitbucket Cloud, Slack, Linear, Sentry, PagerDuty and webhook triggers.
- Recurring tasks: Automations can review recent PR commits for bugs and summarize changes to the codebase on a schedule (automations page, retrieved 2026-10-09).
- Client independence: Automations and Projects run in the cloud; closing the laptop does not stop a Project.
- Scheduled triggers may run late but do not start before the indicated time; scheduler location, overlap policy and run history are not stated.
- Pull requests from forks fail for source-control triggers, except the pull-request-merged trigger.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Schedule/cron, GitHub/GitLab/Bitbucket events, Slack, webhook, Linear, Sentry, PagerDuty triggers; PR, Slack and MCP tools; persistent Memories; billed as cloud agent usage; scheduled runs may be delayed.
Shared prompting: Yes cross-checked
Shared prompting
Typed value: Yes.
Agents are visible to members of the team they were started under and are read-only for viewers; follow-ups are owner-only unless an admin enables them for service accounts or everyone. Each viewer connects their own source-control account and needs access to the repository; team membership alone does not grant access.
- Visibility: Agents are visible to members of the Cursor team they were started under. Cursor also checks each viewer's repository access: the viewer connects their own source control account and needs access to the repository. Team membership alone does not grant access (cloud page).
- Coprompting: Read-only by default; team follow-ups can be disabled, limited to service-account agents, or open to all members.
- Permissions: Admin-controlled team follow-up setting (disabled, service accounts only, or all).
- Presence, ordered co-prompting and live comments were not found; shared transcripts have no comments.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Clients (iOS, web, desktop, Slack, GitHub/Bitbucket comments, Linear, API); paid plan required; managed VMs; environments; no stated concurrency limit; team visibility read-only unless admin enables follow-ups; billed at model API pricing; spend limit set at first use; screenshots, videos and logs as artifacts; draft/merge-ready PRs; Android: open cursor.com/agents in Chrome and tap Install App for a PWA; links to the native iOS app page.
-
Finding: Team follow-ups: disabled (owner only), service accounts only, or all; admin-only toggles for summaries and long running agents; network mode configurable at user, team and environment level.
Team roles: Yes cross-checked
Team roles
Typed value: Yes.
Teams have Member, Admin and Unpaid Admin roles with SAML SSO; Enterprise adds SCIM, organization groups and MDM restrictions.
- Organization roles: Member, Admin, Unpaid Admin on teams; organization groups on Enterprise.
- Membership: Email invitations, invite links, SSO auto-join and domain matching; admins can remove members, and removal does not delete the member’s Cursor account.
- Sso: SAML 2.0 SSO on Teams (Okta, Azure AD, Google Workspace and OneLogin listed).
- Provisioning: SCIM 2.0 on Enterprise with SSO.
- Revocation: SCIM 2.0 manages membership. Audit events document credentials_revoked with sessions revoked, retained or revoke_failed; sessions can be retained when membership in another team in the organization remains. Runtime enforcement was not tested.
- Plan scope: SCIM, MDM policies, pooled usage and audit logs are Enterprise; SAML SSO is on Teams.
- Project- and task-level roles were not found.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Member, Admin and Unpaid Admin roles; invite, link, SSO and domain join; SAML 2.0 SSO on Team plans; seats $40 Standard and $120 Premium per user per month; yearly saves 20%; removal does not delete the member account. SCIM is covered on the IAM page, not here.
-
Identity and access management
Finding: SAML 2.0, SCIM 2.0 (Enterprise with SSO), organization groups, MDM team ID restrictions.
SSO: Yes cross-checked
SSO
Typed value: Yes.
Teams have Member, Admin and Unpaid Admin roles with SAML SSO; Enterprise adds SCIM, organization groups and MDM restrictions.
- Organization roles: Member, Admin, Unpaid Admin on teams; organization groups on Enterprise.
- Membership: Email invitations, invite links, SSO auto-join and domain matching; admins can remove members, and removal does not delete the member’s Cursor account.
- Sso: SAML 2.0 SSO on Teams (Okta, Azure AD, Google Workspace and OneLogin listed).
- Provisioning: SCIM 2.0 on Enterprise with SSO.
- Revocation: SCIM 2.0 manages membership. Audit events document credentials_revoked with sessions revoked, retained or revoke_failed; sessions can be retained when membership in another team in the organization remains. Runtime enforcement was not tested.
- Plan scope: SCIM, MDM policies, pooled usage and audit logs are Enterprise; SAML SSO is on Teams.
- Project- and task-level roles were not found.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Member, Admin and Unpaid Admin roles; invite, link, SSO and domain join; SAML 2.0 SSO on Team plans; seats $40 Standard and $120 Premium per user per month; yearly saves 20%; removal does not delete the member account. SCIM is covered on the IAM page, not here.
-
Identity and access management
Finding: SAML 2.0, SCIM 2.0 (Enterprise with SSO), organization groups, MDM team ID restrictions.
Pull request workflow: Yes cross-checked
Pull request workflow
Typed value: Yes.
Cloud agents push branches, make signed commits, open draft PRs and respond to PR comments; local work uses worktrees.
- Branches: New cursor/... branch per agent by default.
- Commits: Agent commits are signed with an HSM-backed key and show Verified.
- Pull requests: Draft PRs, autoCreatePR and merge-ready PRs.
- Comments: @cursor comments on GitHub/Bitbucket PRs start or continue agents.
- Rebase and conflict handling were not documented beyond mobile "update branch".
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Clients (iOS, web, desktop, Slack, GitHub/Bitbucket comments, Linear, API); paid plan required; managed VMs; environments; no stated concurrency limit; team visibility read-only unless admin enables follow-ups; billed at model API pricing; spend limit set at first use; screenshots, videos and logs as artifacts; draft/merge-ready PRs; Android: open cursor.com/agents in Chrome and tap Install App for a PWA; links to the native iOS app page.
-
Finding: Firecracker microVM per agent in a separate AWS account; workspace recycled after idle; snapshots kept rolling 90 days inactive; conversation state kept indefinitely unless deleted via API; Privacy Mode (not Legacy); commands auto-run; draft PRs; HSM-signed commits; SOC 2 Type 2 statement; region not stated.
-
Finding: v1 API in public beta ("APIs may change before general availability"); webhooks not yet in v1; Basic/Bearer API-key auth; agents, runs, stream, cancel, artifacts, environments, secrets, usage endpoints; up to 20 repos; only one active run per agent.
Review interface: Yes cross-checked
Review interface
Typed value: Yes.
Cloud runs produce diffs, PRs and artifacts (screenshots, videos, log references) with remote desktop takeover and port previews.
- Diffs: The mobile app shows full diffs; PRs carry the agent’s artifacts.
- Screenshots: Screenshots.
- Recordings: Videos.
- Test outputs: Logs and log references.
- Task links: Artifacts attach to the PR. The API lists artifacts and returns a temporary 15-minute presigned S3 URL for each download (api-endpoints, retrieved 2026-10-09).
- Prompt leakage: Artifacts in PR descriptions use long, unguessable URLs that are viewable without authentication.
- Artifacts are kept with the agent transcript: indefinite by default and deletable; the Delete Agent API removes the transcript and artifacts (cloud-security retention table, retrieved 2026-10-09).
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: iPhone/iPad app (iOS/iPadOS 26+): start agents, follow-ups, PR review and merge actions, push notifications, Live Activities, Remote Control of local sessions; "Android is planned" with no timing; available on Start, Pro, Pro+, Ultra, Teams, Enterprise.
-
Finding: Computer use, artifacts (screenshots, videos, log references) attached to PRs, remote desktop takeover, team MCP (HTTP/stdio), subscriptions up to 180 days, CI auto-fix for GitHub Actions on Teams only (max 10 follow-ups).
Product license: Apache-2.0: No. Proprietary cross-checked
Product license: Apache-2.0
Typed value: No. Proprietary.
The Terms grant a limited right to access and use the Service and reserve all rights not granted; the public cursor/cursor repository holds no product source.
- Client license: The Terms grant a limited right to access and use the Service and prohibit reverse engineering and derivative works; no separate client licence or EULA was found.
- Server license: The hosted Service is governed by the Terms; section 5.1 states that there are no implied licences.
- Terms were read at a high level (sections 1.1, 1.5, 5.1); an exact client licence file or EULA was not located.
- Marketplace plugins must be open source, but that does not apply to the client.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Limited right to access and use the Service (1.1); reverse engineering and derivative works prohibited (1.5); no implied licences and all rights reserved (5.1); Beta Services may be discontinued at any time (1.6); Anysphere may modify, suspend or discontinue the Services (9, 10). Last updated Sep 3, 2026.
Product license: MIT: No. Proprietary cross-checked
Product license: MIT
Typed value: No. Proprietary.
The Terms grant a limited right to access and use the Service and reserve all rights not granted; the public cursor/cursor repository holds no product source.
- Client license: The Terms grant a limited right to access and use the Service and prohibit reverse engineering and derivative works; no separate client licence or EULA was found.
- Server license: The hosted Service is governed by the Terms; section 5.1 states that there are no implied licences.
- Terms were read at a high level (sections 1.1, 1.5, 5.1); an exact client licence file or EULA was not located.
- Marketplace plugins must be open source, but that does not apply to the client.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Limited right to access and use the Service (1.1); reverse engineering and derivative works prohibited (1.5); no implied licences and all rights reserved (5.1); Beta Services may be discontinued at any time (1.6); Anysphere may modify, suspend or discontinue the Services (9, 10). Last updated Sep 3, 2026.
Product license: GPL-3.0: No. Proprietary cross-checked
Product license: GPL-3.0
Typed value: No. Proprietary.
The Terms grant a limited right to access and use the Service and reserve all rights not granted; the public cursor/cursor repository holds no product source.
- Client license: The Terms grant a limited right to access and use the Service and prohibit reverse engineering and derivative works; no separate client licence or EULA was found.
- Server license: The hosted Service is governed by the Terms; section 5.1 states that there are no implied licences.
- Terms were read at a high level (sections 1.1, 1.5, 5.1); an exact client licence file or EULA was not located.
- Marketplace plugins must be open source, but that does not apply to the client.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Limited right to access and use the Service (1.1); reverse engineering and derivative works prohibited (1.5); no implied licences and all rights reserved (5.1); Beta Services may be discontinued at any time (1.6); Anysphere may modify, suspend or discontinue the Services (9, 10). Last updated Sep 3, 2026.
Product license: GPL-3.0-or-later: No. Proprietary cross-checked
Product license: GPL-3.0-or-later
Typed value: No. Proprietary.
The Terms grant a limited right to access and use the Service and reserve all rights not granted; the public cursor/cursor repository holds no product source.
- Client license: The Terms grant a limited right to access and use the Service and prohibit reverse engineering and derivative works; no separate client licence or EULA was found.
- Server license: The hosted Service is governed by the Terms; section 5.1 states that there are no implied licences.
- Terms were read at a high level (sections 1.1, 1.5, 5.1); an exact client licence file or EULA was not located.
- Marketplace plugins must be open source, but that does not apply to the client.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Limited right to access and use the Service (1.1); reverse engineering and derivative works prohibited (1.5); no implied licences and all rights reserved (5.1); Beta Services may be discontinued at any time (1.6); Anysphere may modify, suspend or discontinue the Services (9, 10). Last updated Sep 3, 2026.
Product license: AGPL-3.0: No. Proprietary cross-checked
Product license: AGPL-3.0
Typed value: No. Proprietary.
The Terms grant a limited right to access and use the Service and reserve all rights not granted; the public cursor/cursor repository holds no product source.
- Client license: The Terms grant a limited right to access and use the Service and prohibit reverse engineering and derivative works; no separate client licence or EULA was found.
- Server license: The hosted Service is governed by the Terms; section 5.1 states that there are no implied licences.
- Terms were read at a high level (sections 1.1, 1.5, 5.1); an exact client licence file or EULA was not located.
- Marketplace plugins must be open source, but that does not apply to the client.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Limited right to access and use the Service (1.1); reverse engineering and derivative works prohibited (1.5); no implied licences and all rights reserved (5.1); Beta Services may be discontinued at any time (1.6); Anysphere may modify, suspend or discontinue the Services (9, 10). Last updated Sep 3, 2026.
Product license: AGPL-3.0-or-later: No. Proprietary cross-checked
Product license: AGPL-3.0-or-later
Typed value: No. Proprietary.
The Terms grant a limited right to access and use the Service and reserve all rights not granted; the public cursor/cursor repository holds no product source.
- Client license: The Terms grant a limited right to access and use the Service and prohibit reverse engineering and derivative works; no separate client licence or EULA was found.
- Server license: The hosted Service is governed by the Terms; section 5.1 states that there are no implied licences.
- Terms were read at a high level (sections 1.1, 1.5, 5.1); an exact client licence file or EULA was not located.
- Marketplace plugins must be open source, but that does not apply to the client.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Limited right to access and use the Service (1.1); reverse engineering and derivative works prohibited (1.5); no implied licences and all rights reserved (5.1); Beta Services may be discontinued at any time (1.6); Anysphere may modify, suspend or discontinue the Services (9, 10). Last updated Sep 3, 2026.
Product license: Elastic License 2.0: No. Proprietary cross-checked
Product license: Elastic License 2.0
Typed value: No. Proprietary.
The Terms grant a limited right to access and use the Service and reserve all rights not granted; the public cursor/cursor repository holds no product source.
- Client license: The Terms grant a limited right to access and use the Service and prohibit reverse engineering and derivative works; no separate client licence or EULA was found.
- Server license: The hosted Service is governed by the Terms; section 5.1 states that there are no implied licences.
- Terms were read at a high level (sections 1.1, 1.5, 5.1); an exact client licence file or EULA was not located.
- Marketplace plugins must be open source, but that does not apply to the client.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Limited right to access and use the Service (1.1); reverse engineering and derivative works prohibited (1.5); no implied licences and all rights reserved (5.1); Beta Services may be discontinued at any time (1.6); Anysphere may modify, suspend or discontinue the Services (9, 10). Last updated Sep 3, 2026.
Product license: Custom terms: No. Proprietary cross-checked
Product license: Custom terms
Typed value: No. Proprietary.
The Terms grant a limited right to access and use the Service and reserve all rights not granted; the public cursor/cursor repository holds no product source.
- Client license: The Terms grant a limited right to access and use the Service and prohibit reverse engineering and derivative works; no separate client licence or EULA was found.
- Server license: The hosted Service is governed by the Terms; section 5.1 states that there are no implied licences.
- Terms were read at a high level (sections 1.1, 1.5, 5.1); an exact client licence file or EULA was not located.
- Marketplace plugins must be open source, but that does not apply to the client.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Limited right to access and use the Service (1.1); reverse engineering and derivative works prohibited (1.5); no implied licences and all rights reserved (5.1); Beta Services may be discontinued at any time (1.6); Anysphere may modify, suspend or discontinue the Services (9, 10). Last updated Sep 3, 2026.
Product license: Proprietary: Yes. Proprietary cross-checked
Product license: Proprietary
Typed value: Yes. Proprietary.
The Terms grant a limited right to access and use the Service and reserve all rights not granted; the public cursor/cursor repository holds no product source.
- Client license: The Terms grant a limited right to access and use the Service and prohibit reverse engineering and derivative works; no separate client licence or EULA was found.
- Server license: The hosted Service is governed by the Terms; section 5.1 states that there are no implied licences.
- Terms were read at a high level (sections 1.1, 1.5, 5.1); an exact client licence file or EULA was not located.
- Marketplace plugins must be open source, but that does not apply to the client.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Limited right to access and use the Service (1.1); reverse engineering and derivative works prohibited (1.5); no implied licences and all rights reserved (5.1); Beta Services may be discontinued at any time (1.6); Anysphere may modify, suspend or discontinue the Services (9, 10). Last updated Sep 3, 2026.
Self-hosting: No
Self-hosting
Typed value: No.
Cursor is a vendor-hosted service; only tool execution can be self-hosted, and cancellation reverts the account to Hobby.
- Self hosting: Self-Hosted Machines keep the checkout on own hardware but Cursor still runs the agent loop and inference.
- Vendor dependencies: Workers need outbound HTTPS to api2.cursor.sh and artifact storage; file contents, terminal output and diffs are sent to Cursor during runs.
- Export formats and history portability beyond API/CSV were not established.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Tool execution on own machine while agent loop and inference stay in Cursor cloud; Team Pools need Enterprise; up to 200 workers per user and 1,000 per team; outbound HTTPS only; partner hosts such as AWS Lambda, Modal, Vercel; user pays for own infrastructure.
Free access: Yes cross-checked
Free access
Typed value: Yes.
Free has limited core use. Plus $8 lacks Cloud Agents; Premium $30 and higher add Cloud access. Existing legacy paid accounts keep their access.
- Ongoing free: Free $0 has limited Agent/Chat/Tab use with Cursor models.
- Paid only: Cloud Agents, automations, Origin, Bugbot and third-party model access require Premium+ or eligible legacy plans.
- Free scenarios: Limited local core Agent/Chat/Tab usage; free tier excludes Cloud Agents.
- Usage caps: Free limited usage; paid weekly pools described as tier multiples without fixed task count.
- No workload, billing or authenticated client test was performed; unestablished details remain unknown.
-
Finding: New-customer X Subscription: Free $0, Plus $8, Premium $30, Super $100, Ultra $200 monthly; Teams tiers $30/$100/$200 per user. Third-party model use excluded and separately billed. Legacy Cursor plans continue only for existing customers. Taxes extra.
-
Finding: New X Subscription yearly prices: Plus $80, Premium $300, Super $1,000, Ultra $2,000. Weekly Cursor/Grok usage pool; Plus excludes Cloud and third-party models. Legacy plans remain opt-in migration, not available to new customers.
Starting price: Free: Yes. Free cross-checked
Starting price: Free
Typed value: Yes. Free.
Free has limited core use. Plus $8 lacks Cloud Agents; Premium $30 and higher add Cloud access. Existing legacy paid accounts keep their access.
- Ongoing free: Free $0 has limited Agent/Chat/Tab use with Cursor models.
- Paid only: Cloud Agents, automations, Origin, Bugbot and third-party model access require Premium+ or eligible legacy plans.
- Free scenarios: Limited local core Agent/Chat/Tab usage; free tier excludes Cloud Agents.
- Usage caps: Free limited usage; paid weekly pools described as tier multiples without fixed task count.
- No workload, billing or authenticated client test was performed; unestablished details remain unknown.
-
Finding: New-customer X Subscription: Free $0, Plus $8, Premium $30, Super $100, Ultra $200 monthly; Teams tiers $30/$100/$200 per user. Third-party model use excluded and separately billed. Legacy Cursor plans continue only for existing customers. Taxes extra.
-
Finding: New X Subscription yearly prices: Plus $80, Premium $300, Super $1,000, Ultra $2,000. Weekly Cursor/Grok usage pool; Plus excludes Cloud and third-party models. Legacy plans remain opt-in migration, not available to new customers.
Starting price: Up to $20/month: No. Free cross-checked
Starting price: Up to $20/month
Typed value: No. Free.
Free has limited core use. Plus $8 lacks Cloud Agents; Premium $30 and higher add Cloud access. Existing legacy paid accounts keep their access.
- Ongoing free: Free $0 has limited Agent/Chat/Tab use with Cursor models.
- Paid only: Cloud Agents, automations, Origin, Bugbot and third-party model access require Premium+ or eligible legacy plans.
- Free scenarios: Limited local core Agent/Chat/Tab usage; free tier excludes Cloud Agents.
- Usage caps: Free limited usage; paid weekly pools described as tier multiples without fixed task count.
- No workload, billing or authenticated client test was performed; unestablished details remain unknown.
-
Finding: New-customer X Subscription: Free $0, Plus $8, Premium $30, Super $100, Ultra $200 monthly; Teams tiers $30/$100/$200 per user. Third-party model use excluded and separately billed. Legacy Cursor plans continue only for existing customers. Taxes extra.
-
Finding: New X Subscription yearly prices: Plus $80, Premium $300, Super $1,000, Ultra $2,000. Weekly Cursor/Grok usage pool; Plus excludes Cloud and third-party models. Legacy plans remain opt-in migration, not available to new customers.
Starting price: $21–40/month: No. Free cross-checked
Starting price: $21–40/month
Typed value: No. Free.
Free has limited core use. Plus $8 lacks Cloud Agents; Premium $30 and higher add Cloud access. Existing legacy paid accounts keep their access.
- Ongoing free: Free $0 has limited Agent/Chat/Tab use with Cursor models.
- Paid only: Cloud Agents, automations, Origin, Bugbot and third-party model access require Premium+ or eligible legacy plans.
- Free scenarios: Limited local core Agent/Chat/Tab usage; free tier excludes Cloud Agents.
- Usage caps: Free limited usage; paid weekly pools described as tier multiples without fixed task count.
- No workload, billing or authenticated client test was performed; unestablished details remain unknown.
-
Finding: New-customer X Subscription: Free $0, Plus $8, Premium $30, Super $100, Ultra $200 monthly; Teams tiers $30/$100/$200 per user. Third-party model use excluded and separately billed. Legacy Cursor plans continue only for existing customers. Taxes extra.
-
Finding: New X Subscription yearly prices: Plus $80, Premium $300, Super $1,000, Ultra $2,000. Weekly Cursor/Grok usage pool; Plus excludes Cloud and third-party models. Legacy plans remain opt-in migration, not available to new customers.
Starting price: $41–100/month: No. Free cross-checked
Starting price: $41–100/month
Typed value: No. Free.
Free has limited core use. Plus $8 lacks Cloud Agents; Premium $30 and higher add Cloud access. Existing legacy paid accounts keep their access.
- Ongoing free: Free $0 has limited Agent/Chat/Tab use with Cursor models.
- Paid only: Cloud Agents, automations, Origin, Bugbot and third-party model access require Premium+ or eligible legacy plans.
- Free scenarios: Limited local core Agent/Chat/Tab usage; free tier excludes Cloud Agents.
- Usage caps: Free limited usage; paid weekly pools described as tier multiples without fixed task count.
- No workload, billing or authenticated client test was performed; unestablished details remain unknown.
-
Finding: New-customer X Subscription: Free $0, Plus $8, Premium $30, Super $100, Ultra $200 monthly; Teams tiers $30/$100/$200 per user. Third-party model use excluded and separately billed. Legacy Cursor plans continue only for existing customers. Taxes extra.
-
Finding: New X Subscription yearly prices: Plus $80, Premium $300, Super $1,000, Ultra $2,000. Weekly Cursor/Grok usage pool; Plus excludes Cloud and third-party models. Legacy plans remain opt-in migration, not available to new customers.
Starting price: Over $100/month: No. Free cross-checked
Starting price: Over $100/month
Typed value: No. Free.
Free has limited core use. Plus $8 lacks Cloud Agents; Premium $30 and higher add Cloud access. Existing legacy paid accounts keep their access.
- Ongoing free: Free $0 has limited Agent/Chat/Tab use with Cursor models.
- Paid only: Cloud Agents, automations, Origin, Bugbot and third-party model access require Premium+ or eligible legacy plans.
- Free scenarios: Limited local core Agent/Chat/Tab usage; free tier excludes Cloud Agents.
- Usage caps: Free limited usage; paid weekly pools described as tier multiples without fixed task count.
- No workload, billing or authenticated client test was performed; unestablished details remain unknown.
-
Finding: New-customer X Subscription: Free $0, Plus $8, Premium $30, Super $100, Ultra $200 monthly; Teams tiers $30/$100/$200 per user. Third-party model use excluded and separately billed. Legacy Cursor plans continue only for existing customers. Taxes extra.
-
Finding: New X Subscription yearly prices: Plus $80, Premium $300, Super $1,000, Ultra $2,000. Weekly Cursor/Grok usage pool; Plus excludes Cloud and third-party models. Legacy plans remain opt-in migration, not available to new customers.
Starting price: Contact for pricing: No. Free cross-checked
Starting price: Contact for pricing
Typed value: No. Free.
Free has limited core use. Plus $8 lacks Cloud Agents; Premium $30 and higher add Cloud access. Existing legacy paid accounts keep their access.
- Ongoing free: Free $0 has limited Agent/Chat/Tab use with Cursor models.
- Paid only: Cloud Agents, automations, Origin, Bugbot and third-party model access require Premium+ or eligible legacy plans.
- Free scenarios: Limited local core Agent/Chat/Tab usage; free tier excludes Cloud Agents.
- Usage caps: Free limited usage; paid weekly pools described as tier multiples without fixed task count.
- No workload, billing or authenticated client test was performed; unestablished details remain unknown.
-
Finding: New-customer X Subscription: Free $0, Plus $8, Premium $30, Super $100, Ultra $200 monthly; Teams tiers $30/$100/$200 per user. Third-party model use excluded and separately billed. Legacy Cursor plans continue only for existing customers. Taxes extra.
-
Finding: New X Subscription yearly prices: Plus $80, Premium $300, Super $1,000, Ultra $2,000. Weekly Cursor/Grok usage pool; Plus excludes Cloud and third-party models. Legacy plans remain opt-in migration, not available to new customers.
Pricing model: Free, no paid plan: No
Pricing model: Free, no paid plan
Typed value: No.
New subscriptions cost $8 Plus, $30 Premium, $100 Super or $200 Ultra monthly; Teams seats $30/$100/$200. Legacy rates continue for existing accounts. Third-party models are billed separately.
- Currency: Dollar-denominated public prices; taxes excluded.
- Taxes: Prices exclude applicable taxes.
- No workload, billing or authenticated client test was performed; unestablished details remain unknown.
-
Finding: New-customer X Subscription: Free $0, Plus $8, Premium $30, Super $100, Ultra $200 monthly; Teams tiers $30/$100/$200 per user. Third-party model use excluded and separately billed. Legacy Cursor plans continue only for existing customers. Taxes extra.
Pricing model: Free plan: Yes cross-checked
Pricing model: Free plan
Typed value: Yes.
New subscriptions cost $8 Plus, $30 Premium, $100 Super or $200 Ultra monthly; Teams seats $30/$100/$200. Legacy rates continue for existing accounts. Third-party models are billed separately.
- Currency: Dollar-denominated public prices; taxes excluded.
- Taxes: Prices exclude applicable taxes.
- No workload, billing or authenticated client test was performed; unestablished details remain unknown.
-
Finding: New-customer X Subscription: Free $0, Plus $8, Premium $30, Super $100, Ultra $200 monthly; Teams tiers $30/$100/$200 per user. Third-party model use excluded and separately billed. Legacy Cursor plans continue only for existing customers. Taxes extra.
Pricing model: Subscription: Yes cross-checked
Pricing model: Subscription
Typed value: Yes.
New subscriptions cost $8 Plus, $30 Premium, $100 Super or $200 Ultra monthly; Teams seats $30/$100/$200. Legacy rates continue for existing accounts. Third-party models are billed separately.
- Currency: Dollar-denominated public prices; taxes excluded.
- Taxes: Prices exclude applicable taxes.
- No workload, billing or authenticated client test was performed; unestablished details remain unknown.
-
Finding: New-customer X Subscription: Free $0, Plus $8, Premium $30, Super $100, Ultra $200 monthly; Teams tiers $30/$100/$200 per user. Third-party model use excluded and separately billed. Legacy Cursor plans continue only for existing customers. Taxes extra.
Pricing model: Per seat: Yes cross-checked
Pricing model: Per seat
Typed value: Yes.
New subscriptions cost $8 Plus, $30 Premium, $100 Super or $200 Ultra monthly; Teams seats $30/$100/$200. Legacy rates continue for existing accounts. Third-party models are billed separately.
- Currency: Dollar-denominated public prices; taxes excluded.
- Taxes: Prices exclude applicable taxes.
- No workload, billing or authenticated client test was performed; unestablished details remain unknown.
-
Finding: New-customer X Subscription: Free $0, Plus $8, Premium $30, Super $100, Ultra $200 monthly; Teams tiers $30/$100/$200 per user. Third-party model use excluded and separately billed. Legacy Cursor plans continue only for existing customers. Taxes extra.
Pricing model: Usage-based: Yes cross-checked
Pricing model: Usage-based
Typed value: Yes.
New subscriptions cost $8 Plus, $30 Premium, $100 Super or $200 Ultra monthly; Teams seats $30/$100/$200. Legacy rates continue for existing accounts. Third-party models are billed separately.
- Interval: Monthly or annual: Plus $80/year, Premium $300/year, Super $1,000/year, Ultra $2,000/year.
- Seats: Teams mixes Premium $30, Super $100 or Ultra $200 seats per user monthly.
- Included usage: Weekly shared Cursor/Grok pools. Plus excludes Cloud and third-party models; Premium and above add those routes.
- Overages: New plans buy prepaid credit for third-party models and top-offs; legacy plans retain on-demand usage.
- Model cost: Third-party models use API list price plus eligible Teams/Enterprise Cursor Token Rate, with published exemptions.
- No workload, billing or authenticated client test was performed; unestablished details remain unknown.
-
Finding: New X Subscription uses weekly Cursor model pools and prepaid credits for third-party models. Legacy monthly dual pools/on-demand remain for existing accounts. Eligible Teams/Enterprise third-party requests add $0.25/million Cursor Token Rate with stated exemptions.
-
Finding: New X Subscription yearly prices: Plus $80, Premium $300, Super $1,000, Ultra $2,000. Weekly Cursor/Grok usage pool; Plus excludes Cloud and third-party models. Legacy plans remain opt-in migration, not available to new customers.
Pricing model: Per machine or compute: No
Pricing model: Per machine or compute
Typed value: No.
New subscriptions cost $8 Plus, $30 Premium, $100 Super or $200 Ultra monthly; Teams seats $30/$100/$200. Legacy rates continue for existing accounts. Third-party models are billed separately.
- Currency: Dollar-denominated public prices; taxes excluded.
- Taxes: Prices exclude applicable taxes.
- No workload, billing or authenticated client test was performed; unestablished details remain unknown.
-
Finding: New-customer X Subscription: Free $0, Plus $8, Premium $30, Super $100, Ultra $200 monthly; Teams tiers $30/$100/$200 per user. Third-party model use excluded and separately billed. Legacy Cursor plans continue only for existing customers. Taxes extra.
Pricing model: Bring your own key: Yes cross-checked
Pricing model: Bring your own key
Typed value: Yes.
A Cursor account uses the new X Subscription or a retained legacy Cursor plan. Own model API keys are a separate documented route; third-party agent subscription reuse remains unestablished.
- Api accounts: Own keys for OpenAI, Anthropic, Google, Azure or Bedrock bill providers directly; eligible Teams/Enterprise requests still add Cursor Token Rate. Standard non-reasoning OpenAI chat restrictions and Tab exceptions remain scoped to the API-key help.
- Provider restrictions: Some providers restrict models by location, and a model may be missing from the list in those regions. Claude Fable models require data-retention approval for Enterprise, Teams and Privacy Mode individual customers; zero data retention does not apply to own keys.
- Whether Agent or Cloud Agents accept custom API keys is not stated in the help page.
- Use of other vendors' agent subscriptions (for example a Claude or ChatGPT plan) was not found in the pages read.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: OpenAI, Anthropic, Google, Azure OpenAI and AWS Bedrock keys; chat models only; Tab unaffected; whether Agent/Cloud Agents accept keys is not stated.
Pricing model: One-time purchase: No
Pricing model: One-time purchase
Typed value: No.
New subscriptions cost $8 Plus, $30 Premium, $100 Super or $200 Ultra monthly; Teams seats $30/$100/$200. Legacy rates continue for existing accounts. Third-party models are billed separately.
- Currency: Dollar-denominated public prices; taxes excluded.
- Taxes: Prices exclude applicable taxes.
- No workload, billing or authenticated client test was performed; unestablished details remain unknown.
-
Finding: New-customer X Subscription: Free $0, Plus $8, Premium $30, Super $100, Ultra $200 monthly; Teams tiers $30/$100/$200 per user. Third-party model use excluded and separately billed. Legacy Cursor plans continue only for existing customers. Taxes extra.
Paid plans from: No paid plan: No. $10 or less/month
Paid plans from: No paid plan
Typed value: No. $10 or less/month.
New subscriptions cost $8 Plus, $30 Premium, $100 Super or $200 Ultra monthly; Teams seats $30/$100/$200. Legacy rates continue for existing accounts. Third-party models are billed separately.
- Currency: Dollar-denominated public prices; taxes excluded.
- Taxes: Prices exclude applicable taxes.
- No workload, billing or authenticated client test was performed; unestablished details remain unknown.
-
Finding: New-customer X Subscription: Free $0, Plus $8, Premium $30, Super $100, Ultra $200 monthly; Teams tiers $30/$100/$200 per user. Third-party model use excluded and separately billed. Legacy Cursor plans continue only for existing customers. Taxes extra.
Paid plans from: $10 or less/month: Yes. $10 or less/month
Paid plans from: $10 or less/month
Typed value: Yes. $10 or less/month.
New subscriptions cost $8 Plus, $30 Premium, $100 Super or $200 Ultra monthly; Teams seats $30/$100/$200. Legacy rates continue for existing accounts. Third-party models are billed separately.
- Currency: Dollar-denominated public prices; taxes excluded.
- Taxes: Prices exclude applicable taxes.
- No workload, billing or authenticated client test was performed; unestablished details remain unknown.
-
Finding: New-customer X Subscription: Free $0, Plus $8, Premium $30, Super $100, Ultra $200 monthly; Teams tiers $30/$100/$200 per user. Third-party model use excluded and separately billed. Legacy Cursor plans continue only for existing customers. Taxes extra.
Paid plans from: $11–20/month: No. $10 or less/month
Paid plans from: $11–20/month
Typed value: No. $10 or less/month.
New subscriptions cost $8 Plus, $30 Premium, $100 Super or $200 Ultra monthly; Teams seats $30/$100/$200. Legacy rates continue for existing accounts. Third-party models are billed separately.
- Currency: Dollar-denominated public prices; taxes excluded.
- Taxes: Prices exclude applicable taxes.
- No workload, billing or authenticated client test was performed; unestablished details remain unknown.
-
Finding: New-customer X Subscription: Free $0, Plus $8, Premium $30, Super $100, Ultra $200 monthly; Teams tiers $30/$100/$200 per user. Third-party model use excluded and separately billed. Legacy Cursor plans continue only for existing customers. Taxes extra.
Paid plans from: $21–40/month: No. $10 or less/month
Paid plans from: $21–40/month
Typed value: No. $10 or less/month.
New subscriptions cost $8 Plus, $30 Premium, $100 Super or $200 Ultra monthly; Teams seats $30/$100/$200. Legacy rates continue for existing accounts. Third-party models are billed separately.
- Currency: Dollar-denominated public prices; taxes excluded.
- Taxes: Prices exclude applicable taxes.
- No workload, billing or authenticated client test was performed; unestablished details remain unknown.
-
Finding: New-customer X Subscription: Free $0, Plus $8, Premium $30, Super $100, Ultra $200 monthly; Teams tiers $30/$100/$200 per user. Third-party model use excluded and separately billed. Legacy Cursor plans continue only for existing customers. Taxes extra.
Paid plans from: $41–100/month: No. $10 or less/month
Paid plans from: $41–100/month
Typed value: No. $10 or less/month.
New subscriptions cost $8 Plus, $30 Premium, $100 Super or $200 Ultra monthly; Teams seats $30/$100/$200. Legacy rates continue for existing accounts. Third-party models are billed separately.
- Currency: Dollar-denominated public prices; taxes excluded.
- Taxes: Prices exclude applicable taxes.
- No workload, billing or authenticated client test was performed; unestablished details remain unknown.
-
Finding: New-customer X Subscription: Free $0, Plus $8, Premium $30, Super $100, Ultra $200 monthly; Teams tiers $30/$100/$200 per user. Third-party model use excluded and separately billed. Legacy Cursor plans continue only for existing customers. Taxes extra.
Paid plans from: Over $100/month: No. $10 or less/month
Paid plans from: Over $100/month
Typed value: No. $10 or less/month.
New subscriptions cost $8 Plus, $30 Premium, $100 Super or $200 Ultra monthly; Teams seats $30/$100/$200. Legacy rates continue for existing accounts. Third-party models are billed separately.
- Currency: Dollar-denominated public prices; taxes excluded.
- Taxes: Prices exclude applicable taxes.
- No workload, billing or authenticated client test was performed; unestablished details remain unknown.
-
Finding: New-customer X Subscription: Free $0, Plus $8, Premium $30, Super $100, Ultra $200 monthly; Teams tiers $30/$100/$200 per user. Third-party model use excluded and separately billed. Legacy Cursor plans continue only for existing customers. Taxes extra.
Paid plans from: Contact sales: No. $10 or less/month
Paid plans from: Contact sales
Typed value: No. $10 or less/month.
New subscriptions cost $8 Plus, $30 Premium, $100 Super or $200 Ultra monthly; Teams seats $30/$100/$200. Legacy rates continue for existing accounts. Third-party models are billed separately.
- Currency: Dollar-denominated public prices; taxes excluded.
- Taxes: Prices exclude applicable taxes.
- No workload, billing or authenticated client test was performed; unestablished details remain unknown.
-
Finding: New-customer X Subscription: Free $0, Plus $8, Premium $30, Super $100, Ultra $200 monthly; Teams tiers $30/$100/$200 per user. Third-party model use excluded and separately billed. Legacy Cursor plans continue only for existing customers. Taxes extra.
Keeps going with laptop closed: Yes cross-checked
Keeps going with laptop closed
Typed value: Yes.
Work can run on the local machine, on remote SSH hosts, on Cursor-managed VMs, or on self-managed workers; managed cloud runs continue independently of the laptop.
- Managed cloud: Managed isolated VMs (Firecracker microVMs).
- Provisioning: Cursor provisions managed VMs; self-hosted pools use a controller that starts machines on demand, with Kubernetes and partner-host templates.
- Payer: Managed Cloud Agents are billed at model API pricing with infrastructure included; self-hosted users pay their own machines.
- Client disconnect: Local machine need not stay connected to a managed cloud run.
- Other device continuation: Runs started on one client continue from web or mobile.
- Default VM resources are limited; larger VMs need Enterprise support and custom resource configuration is "coming soon".
- Team Pools require Enterprise; My Machines plan requirement not stated.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Clients (iOS, web, desktop, Slack, GitHub/Bitbucket comments, Linear, API); paid plan required; managed VMs; environments; no stated concurrency limit; team visibility read-only unless admin enables follow-ups; billed at model API pricing; spend limit set at first use; screenshots, videos and logs as artifacts; draft/merge-ready PRs; Android: open cursor.com/agents in Chrome and tap Install App for a PWA; links to the native iOS app page.
-
Finding: Coordinator agent plans and delegates to cloud agents, shared context files sync across machines, subscriptions to Slack/PR/CI/schedules, runs in the cloud; paid plans only, not Privacy Mode (Legacy). Dependencies between tasks, plan view and cost visibility not described.
Open-source license: No. Proprietary cross-checked
Open-source license
Typed value: No. Proprietary.
The Terms grant a limited right to access and use the Service and reserve all rights not granted; the public cursor/cursor repository holds no product source.
- Client license: The Terms grant a limited right to access and use the Service and prohibit reverse engineering and derivative works; no separate client licence or EULA was found.
- Server license: The hosted Service is governed by the Terms; section 5.1 states that there are no implied licences.
- Terms were read at a high level (sections 1.1, 1.5, 5.1); an exact client licence file or EULA was not located.
- Marketplace plugins must be open source, but that does not apply to the client.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Limited right to access and use the Service (1.1); reverse engineering and derivative works prohibited (1.5); no implied licences and all rights reserved (5.1); Beta Services may be discontinued at any time (1.6); Anysphere may modify, suspend or discontinue the Services (9, 10). Last updated Sep 3, 2026.
Any desktop app: Yes cross-checked
Any desktop app
Typed value: Yes.
Cursor is reachable from a desktop app (Mac, Windows, Linux), a CLI, a web client at cursor.com/agents, a native iPhone/iPad app (iOS/iPadOS 26+) and chat/issue integrations. There is no native Android app (documented as planned); on Android the docs direct users to install the cursor.com/agents web client as a PWA from Chrome.
- Desktop os: Desktop app for Mac, Windows and Linux; no OS versions listed.
- No native Android app is documented; "Android is planned" with no timing. The Android PWA route is documented only on the Cloud Agents page, not the mobile app page, and was not installed or run.
- iOS app requires iOS/iPadOS 26 or later; English only.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Desktop app for Mac, Windows and Linux; Cursor account required.
Paid plan up to $20/month: Yes. $10 or less/month
Paid plan up to $20/month
Typed value: Yes. $10 or less/month.
New subscriptions cost $8 Plus, $30 Premium, $100 Super or $200 Ultra monthly; Teams seats $30/$100/$200. Legacy rates continue for existing accounts. Third-party models are billed separately.
- Currency: Dollar-denominated public prices; taxes excluded.
- Taxes: Prices exclude applicable taxes.
- No workload, billing or authenticated client test was performed; unestablished details remain unknown.
-
Finding: New-customer X Subscription: Free $0, Plus $8, Premium $30, Super $100, Ultra $200 monthly; Teams tiers $30/$100/$200 per user. Third-party model use excluded and separately billed. Legacy Cursor plans continue only for existing customers. Taxes extra.
Paid plan up to $40/month: Yes. $10 or less/month
Paid plan up to $40/month
Typed value: Yes. $10 or less/month.
New subscriptions cost $8 Plus, $30 Premium, $100 Super or $200 Ultra monthly; Teams seats $30/$100/$200. Legacy rates continue for existing accounts. Third-party models are billed separately.
- Currency: Dollar-denominated public prices; taxes excluded.
- Taxes: Prices exclude applicable taxes.
- No workload, billing or authenticated client test was performed; unestablished details remain unknown.
-
Finding: New-customer X Subscription: Free $0, Plus $8, Premium $30, Super $100, Ultra $200 monthly; Teams tiers $30/$100/$200 per user. Third-party model use excluded and separately billed. Legacy Cursor plans continue only for existing customers. Taxes extra.
Any phone client: Yes cross-checked
Any phone client
Typed value: Yes.
Cursor is reachable from a desktop app (Mac, Windows, Linux), a CLI, a web client at cursor.com/agents, a native iPhone/iPad app (iOS/iPadOS 26+) and chat/issue integrations. There is no native Android app (documented as planned); on Android the docs direct users to install the cursor.com/agents web client as a PWA from Chrome.
- Mobile apps: Native app shipped for iPhone (iOS 26.0+) and iPad (iPadOS 26.0+), English only. No native Android app: the mobile page says "Android is planned"; the Cloud Agents page separately directs Android users to install cursor.com/agents as a PWA from Chrome (Install App).
- Browser: cursor.com/agents web client.
- Mobile browser: Web client usable on any device, including Android as a PWA.
- Task actions: Mobile app: start agents on a Cloud machine, Team Pool or My Machine, send follow-ups, review diffs and checks, ask agents to fix failing checks, merge PRs.
- Notifications: Mobile push notification when an agent finishes a turn; Live Activities for up to eight agents; Slack status reactions and completion messages.
- Cross device continuity: Agents started on desktop, web or mobile appear across clients; Move to Cloud and /remote-control (Cursor 3.9.8+) continue local sessions from the phone. Remote Control must be turned on under Settings > Agents, and on Teams and Enterprise a team admin enables it.
- No native Android app is documented; "Android is planned" with no timing. The Android PWA route is documented only on the Cloud Agents page, not the mobile app page, and was not installed or run.
- iOS app requires iOS/iPadOS 26 or later; English only.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Clients (iOS, web, desktop, Slack, GitHub/Bitbucket comments, Linear, API); paid plan required; managed VMs; environments; no stated concurrency limit; team visibility read-only unless admin enables follow-ups; billed at model API pricing; spend limit set at first use; screenshots, videos and logs as artifacts; draft/merge-ready PRs; Android: open cursor.com/agents in Chrome and tap Install App for a PWA; links to the native iOS app page.
-
Finding: iPhone/iPad app (iOS/iPadOS 26+): start agents, follow-ups, PR review and merge actions, push notifications, Live Activities, Remote Control of local sessions; "Android is planned" with no timing; available on Start, Pro, Pro+, Ultra, Teams, Enterprise.
All public findings
Findings may summarize a source. They are not presented as verbatim statements.
cursor:ACC-01: Cursor is reachable from a desktop app (Mac, Windows, Linux), a CLI, a web client at cursor.com/agents, a native iPhone/iPad app (iOS/iPadOS 26+) and chat/issue integrations. There is no native Android app (documented as planned); on Android the docs direct users to install the cursor.com/agents web client as a PWA from Chrome. cross-checked
cursor:ACC-01
Cursor is reachable from a desktop app (Mac, Windows, Linux), a CLI, a web client at cursor.com/agents, a native iPhone/iPad app (iOS/iPadOS 26+) and chat/issue integrations. There is no native Android app (documented as planned); on Android the docs direct users to install the cursor.com/agents web client as a PWA from Chrome.
- No native Android app is documented; "Android is planned" with no timing. The Android PWA route is documented only on the Cloud Agents page, not the mobile app page, and was not installed or run.
- iOS app requires iOS/iPadOS 26 or later; English only.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Clients (iOS, web, desktop, Slack, GitHub/Bitbucket comments, Linear, API); paid plan required; managed VMs; environments; no stated concurrency limit; team visibility read-only unless admin enables follow-ups; billed at model API pricing; spend limit set at first use; screenshots, videos and logs as artifacts; draft/merge-ready PRs; Android: open cursor.com/agents in Chrome and tap Install App for a PWA; links to the native iOS app page.
-
Finding: iPhone/iPad app (iOS/iPadOS 26+): start agents, follow-ups, PR review and merge actions, push notifications, Live Activities, Remote Control of local sessions; "Android is planned" with no timing; available on Start, Pro, Pro+, Ultra, Teams, Enterprise.
-
Finding: CLI on macOS, Linux, WSL and Windows; agent, plan and ask modes; print mode; "&" prefix hands a conversation to a Cloud Agent; session resume; sandbox controls.
-
Finding: @Cursor starts agents and follow-ups in threads; status reactions; plan posted in thread; routing rules; team follow-ups setting controls who can follow up.
-
Finding: Desktop app for Mac, Windows and Linux; Cursor account required.
-
Finding: agent acp exposes Cursor via Agent Client Protocol to JetBrains, Neovim, Zed and custom clients; modes, permissions, MCP; maturity not stated.
cursor:ACC-02: Cursor runs its own coding-agent loop with selectable models. OpenAI also distributes a separate Codex IDE extension installable inside the editor; that extension does not establish Cursor-native agent/session integration or a Cloud Agents harness option.
cursor:ACC-02
Cursor runs its own coding-agent loop with selectable models. OpenAI also distributes a separate Codex IDE extension installable inside the editor; that extension does not establish Cursor-native agent/session integration or a Cloud Agents harness option.
- External coding harnesses remain unknown without Cursor-native session/launcher integration. The optional Codex extension uses its own UI, authentication and agent loop.
-
Finding: TypeScript SDK page covers local and cloud runs, CURSOR_API_KEY or apiKey authentication, Agent.getUsage() for billed usage and dollar cost, and an SDK Bridge for other languages. Team Admin API keys are not yet supported. The docs navigation also lists a Python page, which was not read.
-
Finding: New X Subscription uses weekly Cursor model pools and prepaid credits for third-party models. Legacy monthly dual pools/on-demand remain for existing accounts. Eligible Teams/Enterprise third-party requests add $0.25/million Cursor Token Rate with stated exemptions.
-
Finding: SKILL.md skills in .cursor, .agents, .claude and .codex directories; team distribution through marketplaces.
-
Finding: Built-in Explore/Bash/Browser and custom subagents (also read from .claude/.codex agent folders); per-subagent model; foreground, background and parallel; final summary returned; resumable by ID; cloud subagents via /in-cloud.
-
Finding: Cursor Agent orchestrates model-specific instructions, file/search/terminal tools and a selected model. This establishes the Cursor harness itself, not external agent loops.
-
Finding: The hooks page says Cursor supports loading hooks from third-party tools like Claude Code (details on a separate Third Party Hooks page), gives a separate default timeout for Claude Code hooks and exposes CLAUDE_PROJECT_DIR as a Claude-compatibility alias.
-
Codex IDE extension inside Cursor
Finding: OpenAI describes installing its separate Codex IDE extension inside Cursor and signing in with ChatGPT or an API key. It explicitly distinguishes this extension from Cursor own model picker/agent. Extension installation alone does not establish integration through Cursor native session/agent layer.
cursor:ACC-03: A Cursor account uses the new X Subscription or a retained legacy Cursor plan. Own model API keys are a separate documented route; third-party agent subscription reuse remains unestablished. cross-checked
cursor:ACC-03
A Cursor account uses the new X Subscription or a retained legacy Cursor plan. Own model API keys are a separate documented route; third-party agent subscription reuse remains unestablished.
- Whether Agent or Cloud Agents accept custom API keys is not stated in the help page.
- Use of other vendors' agent subscriptions (for example a Claude or ChatGPT plan) was not found in the pages read.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Coordinator agent plans and delegates to cloud agents, shared context files sync across machines, subscriptions to Slack/PR/CI/schedules, runs in the cloud; paid plans only, not Privacy Mode (Legacy). Dependencies between tasks, plan view and cost visibility not described.
-
Finding: New X Subscription uses weekly Cursor model pools and prepaid credits for third-party models. Legacy monthly dual pools/on-demand remain for existing accounts. Eligible Teams/Enterprise third-party requests add $0.25/million Cursor Token Rate with stated exemptions.
-
Finding: OpenAI, Anthropic, Google, Azure OpenAI and AWS Bedrock keys; chat models only; Tab unaffected; whether Agent/Cloud Agents accept keys is not stated.
-
Finding: Usage docs separate weekly pools/prepaid credits on X Subscription from legacy Cursor monthly dual pools/on-demand. Dashboard tracks usage and reset; no fixed workload allowance established.
-
Finding: Enterprise US-only data residency; some providers restrict models by location; Start plan India only.
-
Finding: Cloud Agents are the only feature requiring Cursor to store code; repo copies deleted after run; Enterprise US residency covers Cloud Agents with 10% uplift; CMEK for Cloud Agent data; telemetry and indexing not covered.
-
Finding: Desktop app for Mac, Windows and Linux; Cursor account required.
-
Finding: New X Subscription yearly prices: Plus $80, Premium $300, Super $1,000, Ultra $2,000. Weekly Cursor/Grok usage pool; Plus excludes Cloud and third-party models. Legacy plans remain opt-in migration, not available to new customers.
cursor:ACC-04: Cloud Agents need a paid plan and connected source control (or a no-repo start); setup is agent-led or Dockerfile-based; self-managed machines attach through the CLI worker. cross-checked
cursor:ACC-04
Cloud Agents need a paid plan and connected source control (or a no-repo start); setup is agent-led or Dockerfile-based; self-managed machines attach through the CLI worker.
- Local git without a remote was not established.
- Start-from-scratch cloud agents need a paid plan and Origin (cloud setup page) and are documented only in the cloud page and changelog.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Clients (iOS, web, desktop, Slack, GitHub/Bitbucket comments, Linear, API); paid plan required; managed VMs; environments; no stated concurrency limit; team visibility read-only unless admin enables follow-ups; billed at model API pricing; spend limit set at first use; screenshots, videos and logs as artifacts; draft/merge-ready PRs; Android: open cursor.com/agents in Chrome and tap Install App for a PWA; links to the native iOS app page.
-
Finding: environment.json with Dockerfile build, install, start, terminals and snapshot; idempotent install per Build; failed Build does not replace active Build; secrets injected as env vars; multi-repo environments; default VM profile with limited CPU/memory, custom resource config "coming soon".
-
Finding: Tool execution on own machine while agent loop and inference stay in Cursor cloud; Team Pools need Enterprise; up to 200 workers per user and 1,000 per team; outbound HTTPS only; partner hosts such as AWS Lambda, Modal, Vercel; user pays for own infrastructure.
-
Finding: Flags for model, mode, sandbox, worktree, plugin-dir, MCP management and a "worker" command that starts a private cloud worker.
-
Finding: GitHub app permissions for PRs, checks, workflows and branch protection reads; Enterprise Server support; Protected Git Scopes.
-
Finding: Oct 6: remote control for local agents from iOS, on by default except Enterprise. Sep 23: Rollouts and Security Review (Teams/Enterprise; Security Review posts exploitable-bug PR findings). Sep 10: Projects beta.
cursor:ACC-05: Cursor exposes a CLI with headless mode, a public-beta Cloud Agents REST API, a TypeScript SDK, and signed status webhooks on the legacy API. cross-checked
cursor:ACC-05
Cursor exposes a CLI with headless mode, a public-beta Cloud Agents REST API, a TypeScript SDK, and signed status webhooks on the legacy API.
- Cloud Agents API v1 is in public beta and may change; webhooks are not yet in v1 (only statusChange in the legacy v0 API).
- SDK release stage and licence are not stated; Python SDK docs exist in the index but were not read.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Clients (iOS, web, desktop, Slack, GitHub/Bitbucket comments, Linear, API); paid plan required; managed VMs; environments; no stated concurrency limit; team visibility read-only unless admin enables follow-ups; billed at model API pricing; spend limit set at first use; screenshots, videos and logs as artifacts; draft/merge-ready PRs; Android: open cursor.com/agents in Chrome and tap Install App for a PWA; links to the native iOS app page.
-
Finding: Computer use, artifacts (screenshots, videos, log references) attached to PRs, remote desktop takeover, team MCP (HTTP/stdio), subscriptions up to 180 days, CI auto-fix for GitHub Actions on Teams only (max 10 follow-ups).
-
Finding: Schedule/cron, GitHub/GitLab/Bitbucket events, Slack, webhook, Linear, Sentry, PagerDuty triggers; PR, Slack and MCP tools; persistent Memories; billed as cloud agent usage; scheduled runs may be delayed.
-
Finding: Cloud Agents v1 remains public beta. Agent/run/stream/cancel, environment/secret and usage operations with API-key auth are documented. Current environment capacity is up to 300 repositories per October 16 release.
-
Finding: Only statusChange (ERROR or FINISHED) events; HMAC-SHA256 signature; retry policy not specified.
-
Finding: -p/--print non-interactive; --force needed to write files; text/json/stream-json output; CURSOR_API_KEY authentication.
-
Finding: Flags for model, mode, sandbox, worktree, plugin-dir, MCP management and a "worker" command that starts a private cloud worker.
-
Finding: Browser login or API key via CURSOR_API_KEY / --api-key.
-
Finding: Current CLI release index reread November 7 includes persistent sessions, worker improvements and durable goals with rollout gates. Old September version is not reported as latest.
-
Finding: TypeScript SDK page covers local and cloud runs, CURSOR_API_KEY or apiKey authentication, Agent.getUsage() for billed usage and dollar cost, and an SDK Bridge for other languages. Team Admin API keys are not yet supported. The docs navigation also lists a Python page, which was not read.
-
Finding: Newest November 6 voice chat; November 5 app deployment with explicit publish and security checks; October 20 Agent SDK generally available; October 16 environments API beta and up to 300 repos. Dates are release announcements, not tests.
cursor:ACC-06: Cursor is an MCP client with OAuth and admin allowlists, supports plugins and marketplaces, and reads skills, subagents and hooks from other tools' folders. cross-checked
cursor:ACC-06
Cursor is an MCP client with OAuth and admin allowlists, supports plugins and marketplaces, and reads skills, subagents and hooks from other tools' folders.
- CLAUDE.md import was not established in the pages read.
- Custom MCP servers in the Cloud Agents dashboard accept HTTP or stdio; SSE and mcp-remote are not supported there (cloud-capabilities). The API v1 mcpServers type lists sse, so SSE is documented for the API scope only.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Computer use, artifacts (screenshots, videos, log references) attached to PRs, remote desktop takeover, team MCP (HTTP/stdio), subscriptions up to 180 days, CI auto-fix for GitHub Actions on Teams only (max 10 follow-ups).
-
Finding: Auto-review, Allowlist and Run Everything modes; Seatbelt (macOS) and Landlock/seccomp (Linux) sandbox; classifier for out-of-sandbox calls; admins can restrict modes; Cloud Agents do not use run modes.
-
Finding: Flags for model, mode, sandbox, worktree, plugin-dir, MCP management and a "worker" command that starts a private cloud worker.
-
Finding: Live TypeScript SDK page lists no Python SDK page; it points to an SDK Bridge for other languages. Stored ACC-05 sdk note reads 'a Python SDK page is listed but not read'.
-
Finding: stdio, SSE and Streamable HTTP; tools, prompts, resources, roots, elicitation, Apps; OAuth with static client credentials; approval before tool calls; admin MCP allowlists and network controls.
-
Finding: Project, User and Team rules and AGENTS.md; team rules can be enforced; precedence Team, Project, User.
-
Finding: SKILL.md skills in .cursor, .agents, .claude and .codex directories; team distribution through marketplaces.
-
Finding: Built-in Explore/Bash/Browser and custom subagents (also read from .claude/.codex agent folders); per-subagent model; foreground, background and parallel; final summary returned; resumable by ID; cloud subagents via /in-cloud.
-
Finding: Lifecycle and tool hooks that can deny actions (exit code 2); Cloud Agents run command-based hooks from the repository (prompt-based and user-level hooks do not run); enterprise cloud distribution of hooks is Enterprise-only; loads Claude Code hook configs; failClosed option.
-
Finding: Plugins bundle rules, skills, agents, commands, hooks and MCP servers; marketplace plugins are open source and manually reviewed; team marketplaces on Teams/Enterprise.
-
Finding: GitHub app permissions for PRs, checks, workflows and branch protection reads; Enterprise Server support; Protected Git Scopes.
-
Finding: Delegate by assignee or @Cursor mention; real-time status in Linear; PR created on completion; replies and follow-ups in Linear; Linear ID linking not described.
-
Finding: Assign to Cursor or @Cursor in comments; progress and completion summary with PR link posted to Jira; requires Teams or Enterprise and usage-based pricing; ID linking not described.
-
Finding: @Cursor starts agents and follow-ups in threads; status reactions; plan posted in thread; routing rules; team follow-ups setting controls who can follow up.
cursor:PLAN-01: Work can start from prompts in the app, CLI or API, and from Slack, GitHub, Bitbucket, Linear and Jira mentions or assignments, with status posted back. cross-checked
cursor:PLAN-01
Work can start from prompts in the app, CLI or API, and from Slack, GitHub, Bitbucket, Linear and Jira mentions or assignments, with status posted back.
- Which system is the authoritative tracker, supported edits and stable external IDs are not documented; Linear and Jira pages do not describe issue-ID linking.
- Jira needs Cursor Teams or Enterprise with usage-based pricing.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
- The Linear docs page does not say a completion summary or PR link is posted back to the issue (retrieved 2026-10-09); Linear is described as showing status and creating PRs.
-
Finding: Clients (iOS, web, desktop, Slack, GitHub/Bitbucket comments, Linear, API); paid plan required; managed VMs; environments; no stated concurrency limit; team visibility read-only unless admin enables follow-ups; billed at model API pricing; spend limit set at first use; screenshots, videos and logs as artifacts; draft/merge-ready PRs; Android: open cursor.com/agents in Chrome and tap Install App for a PWA; links to the native iOS app page.
-
Finding: Schedule/cron, GitHub/GitLab/Bitbucket events, Slack, webhook, Linear, Sentry, PagerDuty triggers; PR, Slack and MCP tools; persistent Memories; billed as cloud agent usage; scheduled runs may be delayed.
-
Finding: Workspace for parallel agents across local, cloud and remote SSH; worktree isolation; handoff cloud to local and back; /in-cloud and /autopilot cloud subagents.
-
Finding: v1 API in public beta ("APIs may change before general availability"); webhooks not yet in v1; Basic/Bearer API-key auth; agents, runs, stream, cancel, artifacts, environments, secrets, usage endpoints; up to 20 repos; only one active run per agent.
-
Finding: CLI on macOS, Linux, WSL and Windows; agent, plan and ask modes; print mode; "&" prefix hands a conversation to a Cloud Agent; session resume; sandbox controls.
-
Finding: Delegate by assignee or @Cursor mention; real-time status in Linear; PR created on completion; replies and follow-ups in Linear; Linear ID linking not described.
-
Finding: Assign to Cursor or @Cursor in comments; progress and completion summary with PR link posted to Jira; requires Teams or Enterprise and usage-based pricing; ID linking not described.
-
Finding: @Cursor starts agents and follow-ups in threads; status reactions; plan posted in thread; routing rules; team follow-ups setting controls who can follow up.
cursor:PLAN-02: Plan Mode produces editable markdown plans before building; acceptance criteria, revision history and plan-to-work links are not documented. cross-checked
cursor:PLAN-02
Plan Mode produces editable markdown plans before building; acceptance criteria, revision history and plan-to-work links are not documented.
- Plans are files; revisions rely on git or on re-running a refined plan.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: v1 API in public beta ("APIs may change before general availability"); webhooks not yet in v1; Basic/Bearer API-key auth; agents, runs, stream, cancel, artifacts, environments, secrets, usage endpoints; up to 20 repos; only one active run per agent.
-
Finding: CLI on macOS, Linux, WSL and Windows; agent, plan and ask modes; print mode; "&" prefix hands a conversation to a Cloud Agent; session resume; sandbox controls.
-
Finding: Plans are editable markdown, saved to home directory by default and movable into the workspace; acceptance criteria not described.
cursor:PLAN-03: Projects let a coordinator agent plan larger bodies of work and delegate to cloud agents; task dependencies, readiness and blockers are not described.
cursor:PLAN-03
Projects let a coordinator agent plan larger bodies of work and delegate to cloud agents; task dependencies, readiness and blockers are not described.
- Dependencies, blockers and enforcement are not described on the Projects page.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Coordinator agent plans and delegates to cloud agents, shared context files sync across machines, subscriptions to Slack/PR/CI/schedules, runs in the cloud; paid plans only, not Privacy Mode (Legacy). Dependencies between tasks, plan view and cost visibility not described.
-
Finding: Newest November 6 voice chat; November 5 app deployment with explicit publish and security checks; October 20 Agent SDK generally available; October 16 environments API beta and up to 300 repos. Dates are release announcements, not tests.
cursor:PLAN-04: Rules, AGENTS.md, skills, team rules, automation memories and Project shared-context files carry instructions and knowledge across sessions. cross-checked
cursor:PLAN-04
Rules, AGENTS.md, skills, team rules, automation memories and Project shared-context files carry instructions and knowledge across sessions.
- Editor-side memories, provenance and freshness controls were not established.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Schedule/cron, GitHub/GitLab/Bitbucket events, Slack, webhook, Linear, Sentry, PagerDuty triggers; PR, Slack and MCP tools; persistent Memories; billed as cloud agent usage; scheduled runs may be delayed.
-
Finding: Coordinator agent plans and delegates to cloud agents, shared context files sync across machines, subscriptions to Slack/PR/CI/schedules, runs in the cloud; paid plans only, not Privacy Mode (Legacy). Dependencies between tasks, plan view and cost visibility not described.
-
Finding: Plans are editable markdown, saved to home directory by default and movable into the workspace; acceptance criteria not described.
-
Finding: Project, User and Team rules and AGENTS.md; team rules can be enforced; precedence Team, Project, User.
-
Finding: SKILL.md skills in .cursor, .agents, .claude and .codex directories; team distribution through marketplaces.
-
Finding: Read-only team or public transcript links with Fork to Cursor; Teams plans support both, Enterprise team links by default; not on individual plans; no comments.
cursor:PLAN-05: Agents search code with a local grep index, web search and MCP, and accept image inputs; embeddings-based indexing, citations and conflict detection are not documented. cross-checked
cursor:PLAN-05
Agents search code with a local grep index, web search and MCP, and accept image inputs; embeddings-based indexing, citations and conflict detection are not documented.
- Docs state Cursor does not store embeddings for search; freshness and ignore-file behaviour on that page are not covered.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: v1 API in public beta ("APIs may change before general availability"); webhooks not yet in v1; Basic/Bearer API-key auth; agents, runs, stream, cancel, artifacts, environments, secrets, usage endpoints; up to 20 repos; only one active run per agent.
-
Finding: stdio, SSE and Streamable HTTP; tools, prompts, resources, roots, elicitation, Apps; OAuth with static client credentials; approval before tool calls; admin MCP allowlists and network controls.
-
Finding: Search, web, terminal, browser, image generation tools; local checkpoints; queued and steering messages; side chats; conversation search; /goal; no limit on tool calls per task.
-
Finding: Built-in Explore/Bash/Browser and custom subagents (also read from .claude/.codex agent folders); per-subagent model; foreground, background and parallel; final summary returned; resumable by ID; cloud subagents via /in-cloud.
-
Finding: Local Instant Grep index; docs say Cursor does not store embeddings for search; multi-root workspaces not supported by Cloud Agents or worktrees.
-
Finding: Browser pane controlled by the agent with screenshots, console and network reads; recordings not mentioned.
cursor:PLAN-06: Projects has a persistent shared task list and board, distinct from member coordinator chats and worker agents. Projects requires Cloud Agents access; shared Projects requires a team account and Cursor 3.26.13 or later.
cursor:PLAN-06
Projects has a persistent shared task list and board, distinct from member coordinator chats and worker agents. Projects requires Cloud Agents access; shared Projects requires a team account and Cursor 3.26.13 or later.
- First-party documentation review only; no installed-product, account or multi-user scenario was executed. Unanswered atomic questions remain open.
-
Finding: Projects maintains work over months and exposes a shared task list and board distinct from each member own coordinator/workers. Team members work from the same tasks. Sharing needs a team account and Cursor 3.26.13+; Projects requires a Cloud Agents plan (Premium+, Teams, Enterprise or legacy paid plans).
cursor:PLAN-07: Agents interact with external Linear and Jira issues (assigned or mentioned) and report progress there; a native agent-readable issue store was not found. cross-checked
cursor:PLAN-07
Agents interact with external Linear and Jira issues (assigned or mentioned) and report progress there; a native agent-readable issue store was not found.
- This is integration with external trackers, not native issue operations; create/claim/dependency operations are not documented.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
- The Linear docs page says Cloud Agents show real-time status and create PRs; it does not say a completion summary or PR link is posted back to the issue (retrieved 2026-10-09).
-
Finding: Delegate by assignee or @Cursor mention; real-time status in Linear; PR created on completion; replies and follow-ups in Linear; Linear ID linking not described.
-
Finding: Assign to Cursor or @Cursor in comments; progress and completion summary with PR link posted to Jira; requires Teams or Enterprise and usage-based pricing; ID linking not described.
cursor:PLAN-08: Plan Mode saves editable Markdown plans for later reference and user-directed implementation. A dedicated living specification, revision ledger and drift detection remain unestablished. cross-checked
cursor:PLAN-08
Plan Mode saves editable Markdown plans for later reference and user-directed implementation. A dedicated living specification, revision ledger and drift detection remain unestablished.
- Plan files are documented; a separate specification object, exact revision governance and automated drift detection are not established.
-
Finding: Plans are editable through chat or Markdown, saved in the home directory by default, and can be moved into the workspace for later reference. Build starts on user choice.
cursor:COORD-01: Many agents can run in parallel across local worktrees, remote SSH machines, self-managed machines and the cloud; no numeric cloud concurrency limit is stated. cross-checked
cursor:COORD-01
Many agents can run in parallel across local worktrees, remote SSH machines, self-managed machines and the cloud; no numeric cloud concurrency limit is stated.
- Docs state "run as many agents as you want in parallel" without a number; spend limits and machine capacity still apply.
- Observed concurrency was not tested.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Clients (iOS, web, desktop, Slack, GitHub/Bitbucket comments, Linear, API); paid plan required; managed VMs; environments; no stated concurrency limit; team visibility read-only unless admin enables follow-ups; billed at model API pricing; spend limit set at first use; screenshots, videos and logs as artifacts; draft/merge-ready PRs; Android: open cursor.com/agents in Chrome and tap Install App for a PWA; links to the native iOS app page.
-
Finding: iPhone/iPad app (iOS/iPadOS 26+): start agents, follow-ups, PR review and merge actions, push notifications, Live Activities, Remote Control of local sessions; "Android is planned" with no timing; available on Start, Pro, Pro+, Ultra, Teams, Enterprise.
-
Finding: Tool execution on own machine while agent loop and inference stay in Cursor cloud; Team Pools need Enterprise; up to 200 workers per user and 1,000 per team; outbound HTTPS only; partner hosts such as AWS Lambda, Modal, Vercel; user pays for own infrastructure.
-
Finding: Workspace for parallel agents across local, cloud and remote SSH; worktree isolation; handoff cloud to local and back; /in-cloud and /autopilot cloud subagents.
-
Finding: Isolated Git checkouts in Agents Window; /worktree and /best-of-n in IDE; .cursor/worktrees.json setup scripts; default maximum 25 worktrees per machine with automatic cleanup.
cursor:COORD-02: Work is routed by Slack keyword rules and by Jira repository-name, service-name and keyword matching; Team Pool requests go to whichever worker claims them; ownership is the creating user, with follow-ups governed by a team setting. Linear routes by [repo=owner/name] text in the issue or comments, repository labels, the project label, the default repository, and triage rules that add labels and assign issues to Cursor. cross-checked
cursor:COORD-02
Work is routed by Slack keyword rules and by Jira repository-name, service-name and keyword matching; Team Pool requests go to whichever worker claims them; ownership is the creating user, with follow-ups governed by a team setting. Linear routes by [repo=owner/name] text in the issue or comments, repository labels, the project label, the default repository, and triage rules that add labels and assign issues to Cursor.
- Duplicate prevention and queues for humans were not established.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Tool execution on own machine while agent loop and inference stay in Cursor cloud; Team Pools need Enterprise; up to 200 workers per user and 1,000 per team; outbound HTTPS only; partner hosts such as AWS Lambda, Modal, Vercel; user pays for own infrastructure.
-
Finding: Team follow-ups: disabled (owner only), service accounts only, or all; admin-only toggles for summaries and long running agents; network mode configurable at user, team and environment level.
-
Finding: Coordinator agent plans and delegates to cloud agents, shared context files sync across machines, subscriptions to Slack/PR/CI/schedules, runs in the cloud; paid plans only, not Privacy Mode (Legacy). Dependencies between tasks, plan view and cost visibility not described.
-
Finding: v1 API in public beta ("APIs may change before general availability"); webhooks not yet in v1; Basic/Bearer API-key auth; agents, runs, stream, cancel, artifacts, environments, secrets, usage endpoints; up to 20 repos; only one active run per agent.
-
Finding: Delegate by assignee or @Cursor mention; real-time status in Linear; PR created on completion; replies and follow-ups in Linear; Linear ID linking not described.
-
Finding: Assign to Cursor or @Cursor in comments; progress and completion summary with PR link posted to Jira; requires Teams or Enterprise and usage-based pricing; ID linking not described.
-
Finding: @Cursor starts agents and follow-ups in threads; status reactions; plan posted in thread; routing rules; team follow-ups setting controls who can follow up.
cursor:COORD-03: Subagents (built-in and custom) and Project coordinators delegate work with isolated context and return final summaries; they can be resumed by ID. cross-checked
cursor:COORD-03
Subagents (built-in and custom) and Project coordinators delegate work with isolated context and return final summaries; they can be resumed by ID.
- Cross-harness delegation was not found; subagent definitions from .claude/.codex are read but run in Cursor.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: iPhone/iPad app (iOS/iPadOS 26+): start agents, follow-ups, PR review and merge actions, push notifications, Live Activities, Remote Control of local sessions; "Android is planned" with no timing; available on Start, Pro, Pro+, Ultra, Teams, Enterprise.
-
Finding: Coordinator agent plans and delegates to cloud agents, shared context files sync across machines, subscriptions to Slack/PR/CI/schedules, runs in the cloud; paid plans only, not Privacy Mode (Legacy). Dependencies between tasks, plan view and cost visibility not described.
-
Finding: Workspace for parallel agents across local, cloud and remote SSH; worktree isolation; handoff cloud to local and back; /in-cloud and /autopilot cloud subagents.
-
Finding: Live TypeScript SDK page lists no Python SDK page; it points to an SDK Bridge for other languages. Stored ACC-05 sdk note reads 'a Python SDK page is listed but not read'.
-
Finding: Built-in Explore/Bash/Browser and custom subagents (also read from .claude/.codex agent folders); per-subagent model; foreground, background and parallel; final summary returned; resumable by ID; cloud subagents via /in-cloud.
cursor:COORD-04: Automations run on schedules, source-control events, Slack, Linear, Sentry, PagerDuty and webhooks; Projects can subscribe to PRs, CI and schedules. cross-checked
cursor:COORD-04
Automations run on schedules, source-control events, Slack, Linear, Sentry, PagerDuty and webhooks; Projects can subscribe to PRs, CI and schedules.
- Scheduled triggers may run late but do not start before the indicated time; scheduler location, overlap policy and run history are not stated.
- Pull requests from forks fail for source-control triggers, except the pull-request-merged trigger.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Computer use, artifacts (screenshots, videos, log references) attached to PRs, remote desktop takeover, team MCP (HTTP/stdio), subscriptions up to 180 days, CI auto-fix for GitHub Actions on Teams only (max 10 follow-ups).
-
Finding: Schedule/cron, GitHub/GitLab/Bitbucket events, Slack, webhook, Linear, Sentry, PagerDuty triggers; PR, Slack and MCP tools; persistent Memories; billed as cloud agent usage; scheduled runs may be delayed.
-
Finding: Coordinator agent plans and delegates to cloud agents, shared context files sync across machines, subscriptions to Slack/PR/CI/schedules, runs in the cloud; paid plans only, not Privacy Mode (Legacy). Dependencies between tasks, plan view and cost visibility not described.
-
Finding: Bugbot Autofix's 3-attempt cap applies only in 'Commit to Existing Branch' mode, and requires on-demand pricing and storage enabled; stored claims state the cap without that qualifier.
cursor:COORD-05: Users can steer, queue, cancel, resume and follow agents from several clients; notifications reach mobile and Slack. cross-checked
cursor:COORD-05
Users can steer, queue, cancel, resume and follow agents from several clients; notifications reach mobile and Slack.
- Escalation paths and terminal/chat parity were not established.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: iPhone/iPad app (iOS/iPadOS 26+): start agents, follow-ups, PR review and merge actions, push notifications, Live Activities, Remote Control of local sessions; "Android is planned" with no timing; available on Start, Pro, Pro+, Ultra, Teams, Enterprise.
-
Finding: Firecracker microVM per agent in a separate AWS account; workspace recycled after idle; snapshots kept rolling 90 days inactive; conversation state kept indefinitely unless deleted via API; Privacy Mode (not Legacy); commands auto-run; draft PRs; HSM-signed commits; SOC 2 Type 2 statement; region not stated.
-
Finding: v1 API in public beta ("APIs may change before general availability"); webhooks not yet in v1; Basic/Bearer API-key auth; agents, runs, stream, cancel, artifacts, environments, secrets, usage endpoints; up to 20 repos; only one active run per agent.
-
Finding: CLI on macOS, Linux, WSL and Windows; agent, plan and ask modes; print mode; "&" prefix hands a conversation to a Cloud Agent; session resume; sandbox controls.
-
Finding: Live TypeScript SDK page lists no Python SDK page; it points to an SDK Bridge for other languages. Stored ACC-05 sdk note reads 'a Python SDK page is listed but not read'.
-
Finding: Search, web, terminal, browser, image generation tools; local checkpoints; queued and steering messages; side chats; conversation search; /goal; no limit on tool calls per task.
-
Finding: @Cursor starts agents and follow-ups in threads; status reactions; plan posted in thread; routing rules; team follow-ups setting controls who can follow up.
cursor:COORD-06: The Agents Window and the mobile app list agents; Projects show the agents their coordinator started. Dependency and workload views were not found. cross-checked
cursor:COORD-06
The Agents Window and the mobile app list agents; Projects show the agents their coordinator started. Dependency and workload views were not found.
- Fleet-scale behaviour was not tested.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: iPhone/iPad app (iOS/iPadOS 26+): start agents, follow-ups, PR review and merge actions, push notifications, Live Activities, Remote Control of local sessions; "Android is planned" with no timing; available on Start, Pro, Pro+, Ultra, Teams, Enterprise.
-
Finding: Coordinator agent plans and delegates to cloud agents, shared context files sync across machines, subscriptions to Slack/PR/CI/schedules, runs in the cloud; paid plans only, not Privacy Mode (Legacy). Dependencies between tasks, plan view and cost visibility not described.
-
Finding: Workspace for parallel agents across local, cloud and remote SSH; worktree isolation; handoff cloud to local and back; /in-cloud and /autopilot cloud subagents.
-
Finding: Search, web, terminal, browser, image generation tools; local checkpoints; queued and steering messages; side chats; conversation search; /goal; no limit on tool calls per task.
cursor:COORD-07: Cursor supports one-off tasks in Agent mode, longer coordinator-led Projects, Custom Modes with skills, and subagent definitions; in-run changes are limited to steering. cross-checked
cursor:COORD-07
Cursor supports one-off tasks in Agent mode, longer coordinator-led Projects, Custom Modes with skills, and subagent definitions; in-run changes are limited to steering.
- Discoverability and authentication friction were not evaluated.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Coordinator agent plans and delegates to cloud agents, shared context files sync across machines, subscriptions to Slack/PR/CI/schedules, runs in the cloud; paid plans only, not Privacy Mode (Legacy). Dependencies between tasks, plan view and cost visibility not described.
-
Finding: Search, web, terminal, browser, image generation tools; local checkpoints; queued and steering messages; side chats; conversation search; /goal; no limit on tool calls per task.
-
Finding: SKILL.md skills in .cursor, .agents, .claude and .codex directories; team distribution through marketplaces.
-
Finding: Built-in Explore/Bash/Browser and custom subagents (also read from .claude/.codex agent folders); per-subagent model; foreground, background and parallel; final summary returned; resumable by ID; cloud subagents via /in-cloud.
-
Finding: @Cursor starts agents and follow-ups in threads; status reactions; plan posted in thread; routing rules; team follow-ups setting controls who can follow up.
cursor:RUN-01: Work can run on the local machine, on remote SSH hosts, on Cursor-managed VMs, or on self-managed workers; managed cloud runs continue independently of the laptop. cross-checked
cursor:RUN-01
Work can run on the local machine, on remote SSH hosts, on Cursor-managed VMs, or on self-managed workers; managed cloud runs continue independently of the laptop.
- Default VM resources are limited; larger VMs need Enterprise support and custom resource configuration is "coming soon".
- Team Pools require Enterprise; My Machines plan requirement not stated.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Clients (iOS, web, desktop, Slack, GitHub/Bitbucket comments, Linear, API); paid plan required; managed VMs; environments; no stated concurrency limit; team visibility read-only unless admin enables follow-ups; billed at model API pricing; spend limit set at first use; screenshots, videos and logs as artifacts; draft/merge-ready PRs; Android: open cursor.com/agents in Chrome and tap Install App for a PWA; links to the native iOS app page.
-
Finding: iPhone/iPad app (iOS/iPadOS 26+): start agents, follow-ups, PR review and merge actions, push notifications, Live Activities, Remote Control of local sessions; "Android is planned" with no timing; available on Start, Pro, Pro+, Ultra, Teams, Enterprise.
-
Finding: environment.json with Dockerfile build, install, start, terminals and snapshot; idempotent install per Build; failed Build does not replace active Build; secrets injected as env vars; multi-repo environments; default VM profile with limited CPU/memory, custom resource config "coming soon".
-
Finding: Firecracker microVM per agent in a separate AWS account; workspace recycled after idle; snapshots kept rolling 90 days inactive; conversation state kept indefinitely unless deleted via API; Privacy Mode (not Legacy); commands auto-run; draft PRs; HSM-signed commits; SOC 2 Type 2 statement; region not stated.
-
Finding: Tool execution on own machine while agent loop and inference stay in Cursor cloud; Team Pools need Enterprise; up to 200 workers per user and 1,000 per team; outbound HTTPS only; partner hosts such as AWS Lambda, Modal, Vercel; user pays for own infrastructure.
-
Finding: Coordinator agent plans and delegates to cloud agents, shared context files sync across machines, subscriptions to Slack/PR/CI/schedules, runs in the cloud; paid plans only, not Privacy Mode (Legacy). Dependencies between tasks, plan view and cost visibility not described.
-
Finding: Workspace for parallel agents across local, cloud and remote SSH; worktree isolation; handoff cloud to local and back; /in-cloud and /autopilot cloud subagents.
-
Finding: Isolated Git checkouts in Agents Window; /worktree and /best-of-n in IDE; .cursor/worktrees.json setup scripts; default maximum 25 worktrees per machine with automatic cleanup.
-
Finding: Enterprise US-only data residency; some providers restrict models by location; Start plan India only.
-
Finding: Cloud Agents are the only feature requiring Cursor to store code; repo copies deleted after run; Enterprise US residency covers Cloud Agents with 10% uplift; CMEK for Cloud Agent data; telemetry and indexing not covered.
cursor:RUN-02: Local agents use worktrees and an OS sandbox; cloud agents run in per-agent Firecracker microVMs with configurable network egress and redacted secrets. cross-checked
cursor:RUN-02
Local agents use worktrees and an OS sandbox; cloud agents run in per-agent Firecracker microVMs with configurable network egress and redacted secrets.
- Isolation was not tested.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: environment.json with Dockerfile build, install, start, terminals and snapshot; idempotent install per Build; failed Build does not replace active Build; secrets injected as env vars; multi-repo environments; default VM profile with limited CPU/memory, custom resource config "coming soon".
-
Finding: Firecracker microVM per agent in a separate AWS account; workspace recycled after idle; snapshots kept rolling 90 days inactive; conversation state kept indefinitely unless deleted via API; Privacy Mode (not Legacy); commands auto-run; draft PRs; HSM-signed commits; SOC 2 Type 2 statement; region not stated.
-
Cloud Agent security and network
Finding: Three egress modes (allow all, default + allowlist, allowlist only); internet on by default; Enterprise can lock policy; runtime secrets redacted in transcripts and commits.
-
Finding: Isolated Git checkouts in Agents Window; /worktree and /best-of-n in IDE; .cursor/worktrees.json setup scripts; default maximum 25 worktrees per machine with automatic cleanup.
-
Finding: Auto-review, Allowlist and Run Everything modes; Seatbelt (macOS) and Landlock/seccomp (Linux) sandbox; classifier for out-of-sandbox calls; admins can restrict modes; Cloud Agents do not use run modes.
-
Finding: v1 API in public beta ("APIs may change before general availability"); webhooks not yet in v1; Basic/Bearer API-key auth; agents, runs, stream, cancel, artifacts, environments, secrets, usage endpoints; up to 20 repos; only one active run per agent.
-
Finding: Built-in Explore/Bash/Browser and custom subagents (also read from .claude/.codex agent folders); per-subagent model; foreground, background and parallel; final summary returned; resumable by ID; cloud subagents via /in-cloud.
cursor:RUN-03: Cloud agents accept follow-up runs and CLI sessions can be resumed. Current security docs say idle VM resources hibernate and are deleted on lifecycle timers, snapshots are deleted after a rolling 30 days of inactivity, and conversation state is kept until deleted. Worker crash and duplicate-action recovery remain unestablished.
cursor:RUN-03
Cloud agents accept follow-up runs and CLI sessions can be resumed. Current security docs say idle VM resources hibernate and are deleted on lifecycle timers, snapshots are deleted after a rolling 30 days of inactivity, and conversation state is kept until deleted. Worker crash and duplicate-action recovery remain unestablished.
- Worker crash, service restart and duplicate-action behaviour were not documented.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Team follow-ups: disabled (owner only), service accounts only, or all; admin-only toggles for summaries and long running agents; network mode configurable at user, team and environment level.
-
Finding: Week of Sep 21: 'Long-running mode removed from the web'; the Long-running option and team setting were removed. Also Origin CLI 'origin pr stack' removed Oct 1.
-
Finding: Clients (iOS, web, desktop, Slack, GitHub/Bitbucket comments, Linear, API); paid plan required; managed VMs; environments; no stated concurrency limit; team visibility read-only unless admin enables follow-ups; billed at model API pricing; spend limit set at first use; screenshots, videos and logs as artifacts; draft/merge-ready PRs; Android: open cursor.com/agents in Chrome and tap Install App for a PWA; links to the native iOS app page.
-
Finding: iPhone/iPad app (iOS/iPadOS 26+): start agents, follow-ups, PR review and merge actions, push notifications, Live Activities, Remote Control of local sessions; "Android is planned" with no timing; available on Start, Pro, Pro+, Ultra, Teams, Enterprise.
-
Finding: Coordinator agent plans and delegates to cloud agents, shared context files sync across machines, subscriptions to Slack/PR/CI/schedules, runs in the cloud; paid plans only, not Privacy Mode (Legacy). Dependencies between tasks, plan view and cost visibility not described.
-
Finding: Cloud Agents v1 remains public beta. Agent/run/stream/cancel, environment/secret and usage operations with API-key auth are documented. Current environment capacity is up to 300 repositories per October 16 release.
-
Finding: CLI on macOS, Linux, WSL and Windows; agent, plan and ask modes; print mode; "&" prefix hands a conversation to a Cloud Agent; session resume; sandbox controls.
-
Finding: Current CLI release index reread November 7 includes persistent sessions, worker improvements and durable goals with rollout gates. Old September version is not reported as latest.
-
Finding: Current Cloud security documentation says idle VM resources are hibernated and then deleted on lifecycle timers. It does not state a 90-day snapshot-retention period. This is documentation, not a recovery test.
-
Finding: Setup documentation says environment terminals run in shared tmux sessions and a failed Build does not replace the active successful Build. Build logs, debugging from failed Builds and selecting a successful Build are documented.
-
Finding: Workspace recycled after the run goes idle (timer refreshes on follow-ups); snapshots kept for a rolling 30 days of inactivity, extended by each start or resume, then deleted automatically and not deletable on demand; conversation state kept indefinitely by default and deletable on demand. Pull requests open ready for review unless a draft is requested, and nothing merges until a person reviews the change; terminal commands auto-run.
cursor:RUN-04: Cloud setup builds reproducible environments from agent-led setup, snapshots or Dockerfiles with install/start commands, secrets, forwarded ports and non-destructive Build failures. cross-checked
cursor:RUN-04
Cloud setup builds reproducible environments from agent-led setup, snapshots or Dockerfiles with install/start commands, secrets, forwarded ports and non-destructive Build failures.
- No dedicated cache feature or separate repair agent is described; collisions were not covered.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Clients (iOS, web, desktop, Slack, GitHub/Bitbucket comments, Linear, API); paid plan required; managed VMs; environments; no stated concurrency limit; team visibility read-only unless admin enables follow-ups; billed at model API pricing; spend limit set at first use; screenshots, videos and logs as artifacts; draft/merge-ready PRs; Android: open cursor.com/agents in Chrome and tap Install App for a PWA; links to the native iOS app page.
-
Finding: environment.json with Dockerfile build, install, start, terminals and snapshot; idempotent install per Build; failed Build does not replace active Build; secrets injected as env vars; multi-repo environments; default VM profile with limited CPU/memory, custom resource config "coming soon".
cursor:RUN-05: Spend limits bound cloud usage; time limits and timeouts are not stated. cross-checked
cursor:RUN-05
Spend limits bound cloud usage; time limits and timeouts are not stated.
- No run time limit was found; limits listed are spend, subscription lifetime, retry counts and context.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Clients (iOS, web, desktop, Slack, GitHub/Bitbucket comments, Linear, API); paid plan required; managed VMs; environments; no stated concurrency limit; team visibility read-only unless admin enables follow-ups; billed at model API pricing; spend limit set at first use; screenshots, videos and logs as artifacts; draft/merge-ready PRs; Android: open cursor.com/agents in Chrome and tap Install App for a PWA; links to the native iOS app page.
-
Finding: Computer use, artifacts (screenshots, videos, log references) attached to PRs, remote desktop takeover, team MCP (HTTP/stdio), subscriptions up to 180 days, CI auto-fix for GitHub Actions on Teams only (max 10 follow-ups).
-
Finding: Schedule/cron, GitHub/GitLab/Bitbucket events, Slack, webhook, Linear, Sentry, PagerDuty triggers; PR, Slack and MCP tools; persistent Memories; billed as cloud agent usage; scheduled runs may be delayed.
-
Finding: Search, web, terminal, browser, image generation tools; local checkpoints; queued and steering messages; side chats; conversation search; /goal; no limit on tool calls per task.
-
Finding: Bugbot Autofix has a maximum of 3 attempts per PR in Commit to Existing Branch mode, which prevents loops. Create New Branch is the recommended mode. Triggers, status checks and pricing are documented on the same page.
-
Finding: On-demand spend limits per user (individual), team (Teams), member overrides (Enterprise); enforcement not instant; cloud agent specific limits not described.
-
Finding: Email alerts on on-demand spend thresholds.
cursor:RUN-06: Logs, run states, per-run token usage and admin analytics exist; traces and approval-state reporting are not documented. cross-checked
cursor:RUN-06
Logs, run states, per-run token usage and admin analytics exist; traces and approval-state reporting are not documented.
- Reconnect consistency and responsiveness were not measured.
- Analytics API and Admin API availability is plan-dependent.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Clients (iOS, web, desktop, Slack, GitHub/Bitbucket comments, Linear, API); paid plan required; managed VMs; environments; no stated concurrency limit; team visibility read-only unless admin enables follow-ups; billed at model API pricing; spend limit set at first use; screenshots, videos and logs as artifacts; draft/merge-ready PRs; Android: open cursor.com/agents in Chrome and tap Install App for a PWA; links to the native iOS app page.
-
Finding: environment.json with Dockerfile build, install, start, terminals and snapshot; idempotent install per Build; failed Build does not replace active Build; secrets injected as env vars; multi-repo environments; default VM profile with limited CPU/memory, custom resource config "coming soon".
-
Finding: Computer use, artifacts (screenshots, videos, log references) attached to PRs, remote desktop takeover, team MCP (HTTP/stdio), subscriptions up to 180 days, CI auto-fix for GitHub Actions on Teams only (max 10 follow-ups).
-
Finding: v1 API in public beta ("APIs may change before general availability"); webhooks not yet in v1; Basic/Bearer API-key auth; agents, runs, stream, cancel, artifacts, environments, secrets, usage endpoints; up to 20 repos; only one active run per agent.
-
Finding: Live TypeScript SDK page lists no Python SDK page; it points to an SDK Bridge for other languages. Stored ACC-05 sdk note reads 'a Python SDK page is listed but not read'.
-
Finding: Per-user and per-model usage, Cloud Agent starts and PRs, CSV export; cost per agent not covered.
cursor:RUN-07: Token usage is available per agent and per run through the API and SDK; billed dollar cost per agent is available through the SDK (agent.getUsage()); allowances and spend are shown in the dashboard. Per-epic and child roll-ups were not found. cross-checked
cursor:RUN-07
Token usage is available per agent and per run through the API and SDK; billed dollar cost per agent is available through the SDK (agent.getUsage()); allowances and spend are shown in the dashboard. Per-epic and child roll-ups were not found.
- SDK local-agent usage reporting needs per-account enablement.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
- Analytics Admin API access is Enterprise-only; it is distinct from per-agent usage reporting.
-
Finding: Clients (iOS, web, desktop, Slack, GitHub/Bitbucket comments, Linear, API); paid plan required; managed VMs; environments; no stated concurrency limit; team visibility read-only unless admin enables follow-ups; billed at model API pricing; spend limit set at first use; screenshots, videos and logs as artifacts; draft/merge-ready PRs; Android: open cursor.com/agents in Chrome and tap Install App for a PWA; links to the native iOS app page.
-
Finding: Tool execution on own machine while agent loop and inference stay in Cursor cloud; Team Pools need Enterprise; up to 200 workers per user and 1,000 per team; outbound HTTPS only; partner hosts such as AWS Lambda, Modal, Vercel; user pays for own infrastructure.
-
Finding: Cloud Agents v1 remains public beta. Agent/run/stream/cancel, environment/secret and usage operations with API-key auth are documented. Current environment capacity is up to 300 repositories per October 16 release.
-
Finding: TypeScript SDK page covers local and cloud runs, CURSOR_API_KEY or apiKey authentication, Agent.getUsage() for billed usage and dollar cost, and an SDK Bridge for other languages. Team Admin API keys are not yet supported. The docs navigation also lists a Python page, which was not read.
-
Finding: Usage docs separate weekly pools/prepaid credits on X Subscription from legacy Cursor monthly dual pools/on-demand. Dashboard tracks usage and reset; no fixed workload allowance established.
-
Finding: Built-in Explore/Bash/Browser and custom subagents (also read from .claude/.codex agent folders); per-subagent model; foreground, background and parallel; final summary returned; resumable by ID; cloud subagents via /in-cloud.
-
Finding: Per-user and per-model usage, Cloud Agent starts and PRs, CSV export; cost per agent not covered.
-
Finding: Email alerts on on-demand spend thresholds.
cursor:RUN-08: Team Pools queue requests for workers and a controller scales machines; other compute-aware scheduling and test deduplication were not found. cross-checked
cursor:RUN-08
Team Pools queue requests for workers and a controller scales machines; other compute-aware scheduling and test deduplication were not found.
- Fairness, host awareness and verification reuse were not found.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: environment.json with Dockerfile build, install, start, terminals and snapshot; idempotent install per Build; failed Build does not replace active Build; secrets injected as env vars; multi-repo environments; default VM profile with limited CPU/memory, custom resource config "coming soon".
-
Finding: Tool execution on own machine while agent loop and inference stay in Cursor cloud; Team Pools need Enterprise; up to 200 workers per user and 1,000 per team; outbound HTTPS only; partner hosts such as AWS Lambda, Modal, Vercel; user pays for own infrastructure.
-
Finding: v1 API in public beta ("APIs may change before general availability"); webhooks not yet in v1; Basic/Bearer API-key auth; agents, runs, stream, cancel, artifacts, environments, secrets, usage endpoints; up to 20 repos; only one active run per agent.
-
Finding: Oct 6: remote control for local agents from iOS, on by default except Enterprise. Sep 23: Rollouts and Security Review (Teams/Enterprise; Security Review posts exploitable-bug PR findings). Sep 10: Projects beta.
cursor:TEAM-01: Agents are visible to members of the team they were started under and are read-only for viewers; follow-ups are owner-only unless an admin enables them for service accounts or everyone. Each viewer connects their own source-control account and needs access to the repository; team membership alone does not grant access. cross-checked
cursor:TEAM-01
Agents are visible to members of the team they were started under and are read-only for viewers; follow-ups are owner-only unless an admin enables them for service accounts or everyone. Each viewer connects their own source-control account and needs access to the repository; team membership alone does not grant access.
- Presence, ordered co-prompting and live comments were not found; shared transcripts have no comments.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Clients (iOS, web, desktop, Slack, GitHub/Bitbucket comments, Linear, API); paid plan required; managed VMs; environments; no stated concurrency limit; team visibility read-only unless admin enables follow-ups; billed at model API pricing; spend limit set at first use; screenshots, videos and logs as artifacts; draft/merge-ready PRs; Android: open cursor.com/agents in Chrome and tap Install App for a PWA; links to the native iOS app page.
-
Finding: Team follow-ups: disabled (owner only), service accounts only, or all; admin-only toggles for summaries and long running agents; network mode configurable at user, team and environment level.
-
Finding: Agent, owner, turn (who submitted the turn) and workspace identifiers readable from the VM; no PR number key documented.
-
Finding: v1 API in public beta ("APIs may change before general availability"); webhooks not yet in v1; Basic/Bearer API-key auth; agents, runs, stream, cancel, artifacts, environments, secrets, usage endpoints; up to 20 repos; only one active run per agent.
-
Finding: Read-only team or public transcript links with Fork to Cursor; Teams plans support both, Enterprise team links by default; not on individual plans; no comments.
-
Finding: @Cursor starts agents and follow-ups in threads; status reactions; plan posted in thread; routing rules; team follow-ups setting controls who can follow up.
cursor:TEAM-02: Teams have Member, Admin and Unpaid Admin roles with SAML SSO; Enterprise adds SCIM, organization groups and MDM restrictions. cross-checked
cursor:TEAM-02
Teams have Member, Admin and Unpaid Admin roles with SAML SSO; Enterprise adds SCIM, organization groups and MDM restrictions.
- Project- and task-level roles were not found.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Hobby free (no card, limited Agent requests); Individual from $20/mo; Teams $40/user/mo; Enterprise custom; on-demand usage billed in arrears; prices exclusive of taxes; yearly toggle without shown prices.
-
Finding: Firecracker microVM per agent in a separate AWS account; workspace recycled after idle; snapshots kept rolling 90 days inactive; conversation state kept indefinitely unless deleted via API; Privacy Mode (not Legacy); commands auto-run; draft PRs; HSM-signed commits; SOC 2 Type 2 statement; region not stated.
-
Finding: GitHub app permissions for PRs, checks, workflows and branch protection reads; Enterprise Server support; Protected Git Scopes.
-
Finding: Member, Admin and Unpaid Admin roles; invite, link, SSO and domain join; SAML 2.0 SSO on Team plans; seats $40 Standard and $120 Premium per user per month; yearly saves 20%; removal does not delete the member account. SCIM is covered on the IAM page, not here.
-
Identity and access management
Finding: SAML 2.0, SCIM 2.0 (Enterprise with SSO), organization groups, MDM team ID restrictions.
-
Finding: Enterprise audit log of admin and configuration events (not prompts or agent output); CSV and Admin API export; SIEM streaming; retention and integrity controls not described.
cursor:TEAM-03: Run modes, sandbox rules, MCP approvals and hooks control agent actions locally; cloud agents run commands without per-step approval and rely on draft PRs and network policy. cross-checked
cursor:TEAM-03
Run modes, sandbox rules, MCP approvals and hooks control agent actions locally; cloud agents run commands without per-step approval and rely on draft PRs and network policy.
- Approver identity, revocation and temporary secret grants were not documented.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
- Run-modes page: Ask Every Time deprecated in 3.5 (2026-05-22); Auto-review became the recommended default in 3.6 (2026-05-29).
-
Finding: Firecracker microVM per agent in a separate AWS account; workspace recycled after idle; snapshots kept rolling 90 days inactive; conversation state kept indefinitely unless deleted via API; Privacy Mode (not Legacy); commands auto-run; draft PRs; HSM-signed commits; SOC 2 Type 2 statement; region not stated.
-
Cloud Agent security and network
Finding: Three egress modes (allow all, default + allowlist, allowlist only); internet on by default; Enterprise can lock policy; runtime secrets redacted in transcripts and commits.
-
Finding: Auto-review, Allowlist and Run Everything modes; Seatbelt (macOS) and Landlock/seccomp (Linux) sandbox; classifier for out-of-sandbox calls; admins can restrict modes; Cloud Agents do not use run modes.
-
Finding: stdio, SSE and Streamable HTTP; tools, prompts, resources, roots, elicitation, Apps; OAuth with static client credentials; approval before tool calls; admin MCP allowlists and network controls.
-
Finding: Lifecycle and tool hooks that can deny actions (exit code 2); Cloud Agents run command-based hooks from the repository (prompt-based and user-level hooks do not run); enterprise cloud distribution of hooks is Enterprise-only; loads Claude Code hook configs; failClosed option.
-
Finding: Agent routes PRs and can approve low-risk PRs per policy files; GitHub and Origin only; does not replace full code review.
-
Finding: agent acp exposes Cursor via Agent Client Protocol to JetBrains, Neovim, Zed and custom clients; modes, permissions, MCP; maturity not stated.
cursor:TEAM-04: Enterprise audit logs record admin and configuration events with CSV, API and SIEM export; conversations persist across sessions and can be shared read-only. cross-checked
cursor:TEAM-04
Enterprise audit logs record admin and configuration events with CSV, API and SIEM export; conversations persist across sessions and can be shared read-only.
- Audit logs exclude prompts and agent output; retention and tamper-evidence are not stated.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Clients (iOS, web, desktop, Slack, GitHub/Bitbucket comments, Linear, API); paid plan required; managed VMs; environments; no stated concurrency limit; team visibility read-only unless admin enables follow-ups; billed at model API pricing; spend limit set at first use; screenshots, videos and logs as artifacts; draft/merge-ready PRs; Android: open cursor.com/agents in Chrome and tap Install App for a PWA; links to the native iOS app page.
-
Finding: Firecracker microVM per agent in a separate AWS account; workspace recycled after idle; snapshots kept rolling 90 days inactive; conversation state kept indefinitely unless deleted via API; Privacy Mode (not Legacy); commands auto-run; draft PRs; HSM-signed commits; SOC 2 Type 2 statement; region not stated.
-
Cloud Agent security and network
Finding: Three egress modes (allow all, default + allowlist, allowlist only); internet on by default; Enterprise can lock policy; runtime secrets redacted in transcripts and commits.
-
Finding: Read-only team or public transcript links with Fork to Cursor; Teams plans support both, Enterprise team links by default; not on individual plans; no comments.
-
Finding: Enterprise audit log of admin and configuration events (not prompts or agent output); CSV and Admin API export; SIEM streaming; retention and integrity controls not described.
cursor:TEAM-05: Agent Review, Bugbot, a diffs view and mobile PR review support reviewing agent output; shipping guidance and risk views were not found. cross-checked
cursor:TEAM-05
Agent Review, Bugbot, a diffs view and mobile PR review support reviewing agent output; shipping guidance and risk views were not found.
- Review ergonomics (keyboard, effort, missed evidence) were not evaluated.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: iPhone/iPad app (iOS/iPadOS 26+): start agents, follow-ups, PR review and merge actions, push notifications, Live Activities, Remote Control of local sessions; "Android is planned" with no timing; available on Start, Pro, Pro+, Ultra, Teams, Enterprise.
-
Finding: Computer use, artifacts (screenshots, videos, log references) attached to PRs, remote desktop takeover, team MCP (HTTP/stdio), subscriptions up to 180 days, CI auto-fix for GitHub Actions on Teams only (max 10 follow-ups).
-
Finding: Schedule/cron, GitHub/GitLab/Bitbucket events, Slack, webhook, Linear, Sentry, PagerDuty triggers; PR, Slack and MCP tools; persistent Memories; billed as cloud agent usage; scheduled runs may be delayed.
-
Finding: Workspace for parallel agents across local, cloud and remote SSH; worktree isolation; handoff cloud to local and back; /in-cloud and /autopilot cloud subagents.
-
Finding: Automatic, /agent-review or Source Control tab; Quick and Deep depth; rules from BUGBOT.md.
-
Finding: Bugbot Autofix's 3-attempt cap applies only in 'Commit to Existing Branch' mode, and requires on-demand pricing and storage enabled; stored claims state the cap without that qualifier.
-
Finding: Agent routes PRs and can approve low-risk PRs per policy files; GitHub and Origin only; does not replace full code review.
cursor:TEAM-06: Teammates can continue shared cloud runs when an admin enables team follow-ups, take over the agent desktop, or fork shared transcripts into their own editor. cross-checked
cursor:TEAM-06
Teammates can continue shared cloud runs when an admin enables team follow-ups, take over the agent desktop, or fork shared transcripts into their own editor.
- Team follow-ups require admin enablement (disabled, service accounts only, or all).
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Clients (iOS, web, desktop, Slack, GitHub/Bitbucket comments, Linear, API); paid plan required; managed VMs; environments; no stated concurrency limit; team visibility read-only unless admin enables follow-ups; billed at model API pricing; spend limit set at first use; screenshots, videos and logs as artifacts; draft/merge-ready PRs; Android: open cursor.com/agents in Chrome and tap Install App for a PWA; links to the native iOS app page.
-
Finding: Firecracker microVM per agent in a separate AWS account; workspace recycled after idle; snapshots kept rolling 90 days inactive; conversation state kept indefinitely unless deleted via API; Privacy Mode (not Legacy); commands auto-run; draft PRs; HSM-signed commits; SOC 2 Type 2 statement; region not stated.
-
Finding: Computer use, artifacts (screenshots, videos, log references) attached to PRs, remote desktop takeover, team MCP (HTTP/stdio), subscriptions up to 180 days, CI auto-fix for GitHub Actions on Teams only (max 10 follow-ups).
-
Finding: Team follow-ups: disabled (owner only), service accounts only, or all; admin-only toggles for summaries and long running agents; network mode configurable at user, team and environment level.
-
Finding: Workspace for parallel agents across local, cloud and remote SSH; worktree isolation; handoff cloud to local and back; /in-cloud and /autopilot cloud subagents.
-
Finding: Read-only team or public transcript links with Fork to Cursor; Teams plans support both, Enterprise team links by default; not on individual plans; no comments.
cursor:SHIP-01: Cloud runs produce diffs, PRs and artifacts (screenshots, videos, log references) with remote desktop takeover and port previews. cross-checked
cursor:SHIP-01
Cloud runs produce diffs, PRs and artifacts (screenshots, videos, log references) with remote desktop takeover and port previews.
- Artifacts are kept with the agent transcript: indefinite by default and deletable; the Delete Agent API removes the transcript and artifacts (cloud-security retention table, retrieved 2026-10-09).
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Clients (iOS, web, desktop, Slack, GitHub/Bitbucket comments, Linear, API); paid plan required; managed VMs; environments; no stated concurrency limit; team visibility read-only unless admin enables follow-ups; billed at model API pricing; spend limit set at first use; screenshots, videos and logs as artifacts; draft/merge-ready PRs; Android: open cursor.com/agents in Chrome and tap Install App for a PWA; links to the native iOS app page.
-
Finding: iPhone/iPad app (iOS/iPadOS 26+): start agents, follow-ups, PR review and merge actions, push notifications, Live Activities, Remote Control of local sessions; "Android is planned" with no timing; available on Start, Pro, Pro+, Ultra, Teams, Enterprise.
-
Finding: environment.json with Dockerfile build, install, start, terminals and snapshot; idempotent install per Build; failed Build does not replace active Build; secrets injected as env vars; multi-repo environments; default VM profile with limited CPU/memory, custom resource config "coming soon".
-
Finding: Firecracker microVM per agent in a separate AWS account; workspace recycled after idle; snapshots kept rolling 90 days inactive; conversation state kept indefinitely unless deleted via API; Privacy Mode (not Legacy); commands auto-run; draft PRs; HSM-signed commits; SOC 2 Type 2 statement; region not stated.
-
Finding: Computer use, artifacts (screenshots, videos, log references) attached to PRs, remote desktop takeover, team MCP (HTTP/stdio), subscriptions up to 180 days, CI auto-fix for GitHub Actions on Teams only (max 10 follow-ups).
-
Finding: Workspace for parallel agents across local, cloud and remote SSH; worktree isolation; handoff cloud to local and back; /in-cloud and /autopilot cloud subagents.
-
Finding: v1 API in public beta ("APIs may change before general availability"); webhooks not yet in v1; Basic/Bearer API-key auth; agents, runs, stream, cancel, artifacts, environments, secrets, usage endpoints; up to 20 repos; only one active run per agent.
-
Finding: Bugbot Autofix's 3-attempt cap applies only in 'Commit to Existing Branch' mode, and requires on-demand pricing and storage enabled; stored claims state the cap without that qualifier.
cursor:SHIP-02: Independent AI review (Agent Review, Bugbot) and policy files exist, and Bugbot can fail a status check; lint, type and spec gates are not Cursor-specific features in the pages read. cross-checked
cursor:SHIP-02
Independent AI review (Agent Review, Bugbot) and policy files exist, and Bugbot can fail a status check; lint, type and spec gates are not Cursor-specific features in the pages read.
- Findings default to a neutral Bugbot check, so merge blocking needs fail-on-unresolved-issues and branch protection.
- CI auto-fix is GitHub Actions only and Teams-only.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
- Security Review (Teams and Enterprise, 2026-09-23) posts exploitable-bug findings on pull requests (changelog, retrieved 2026-10-09).
-
Finding: Computer use, artifacts (screenshots, videos, log references) attached to PRs, remote desktop takeover, team MCP (HTTP/stdio), subscriptions up to 180 days, CI auto-fix for GitHub Actions on Teams only (max 10 follow-ups).
-
Finding: Automatic, /agent-review or Source Control tab; Quick and Deep depth; rules from BUGBOT.md.
-
Finding: Bugbot Autofix's 3-attempt cap applies only in 'Commit to Existing Branch' mode, and requires on-demand pricing and storage enabled; stored claims state the cap without that qualifier.
-
Finding: Agent routes PRs and can approve low-risk PRs per policy files; GitHub and Origin only; does not replace full code review.
-
Finding: Oct 6: remote control for local agents from iOS, on by default except Enterprise. Sep 23: Rollouts and Security Review (Teams/Enterprise; Security Review posts exploitable-bug PR findings). Sep 10: Projects beta.
cursor:SHIP-03: Cloud agents push branches, make signed commits, open draft PRs and respond to PR comments; local work uses worktrees. cross-checked
cursor:SHIP-03
Cloud agents push branches, make signed commits, open draft PRs and respond to PR comments; local work uses worktrees.
- Rebase and conflict handling were not documented beyond mobile "update branch".
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Clients (iOS, web, desktop, Slack, GitHub/Bitbucket comments, Linear, API); paid plan required; managed VMs; environments; no stated concurrency limit; team visibility read-only unless admin enables follow-ups; billed at model API pricing; spend limit set at first use; screenshots, videos and logs as artifacts; draft/merge-ready PRs; Android: open cursor.com/agents in Chrome and tap Install App for a PWA; links to the native iOS app page.
-
Finding: iPhone/iPad app (iOS/iPadOS 26+): start agents, follow-ups, PR review and merge actions, push notifications, Live Activities, Remote Control of local sessions; "Android is planned" with no timing; available on Start, Pro, Pro+, Ultra, Teams, Enterprise.
-
Finding: Firecracker microVM per agent in a separate AWS account; workspace recycled after idle; snapshots kept rolling 90 days inactive; conversation state kept indefinitely unless deleted via API; Privacy Mode (not Legacy); commands auto-run; draft PRs; HSM-signed commits; SOC 2 Type 2 statement; region not stated.
-
Finding: Computer use, artifacts (screenshots, videos, log references) attached to PRs, remote desktop takeover, team MCP (HTTP/stdio), subscriptions up to 180 days, CI auto-fix for GitHub Actions on Teams only (max 10 follow-ups).
-
Finding: Isolated Git checkouts in Agents Window; /worktree and /best-of-n in IDE; .cursor/worktrees.json setup scripts; default maximum 25 worktrees per machine with automatic cleanup.
-
Finding: v1 API in public beta ("APIs may change before general availability"); webhooks not yet in v1; Basic/Bearer API-key auth; agents, runs, stream, cancel, artifacts, environments, secrets, usage endpoints; up to 20 repos; only one active run per agent.
-
Finding: GitHub app permissions for PRs, checks, workflows and branch protection reads; Enterprise Server support; Protected Git Scopes.
cursor:SHIP-04: Agents open pull requests (ready for review unless a draft is requested). PR Routing and Approval can approve low-risk PRs that meet configured criteria (GitHub and Origin) but does not replace full code review; the mobile app lets a person squash-merge or toggle auto-merge, and the Cloud security docs say nothing merges until a person reviews the change.
cursor:SHIP-04
Agents open pull requests (ready for review unless a draft is requested). PR Routing and Approval can approve low-risk PRs that meet configured criteria (GitHub and Origin) but does not replace full code review; the mobile app lets a person squash-merge or toggle auto-merge, and the Cloud security docs say nothing merges until a person reviews the change.
- Release, deploy and exact-revision approval were not documented.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: iPhone/iPad app (iOS/iPadOS 26+): start agents, follow-ups, PR review and merge actions, push notifications, Live Activities, Remote Control of local sessions; "Android is planned" with no timing; available on Start, Pro, Pro+, Ultra, Teams, Enterprise.
-
Finding: Firecracker microVM per agent in a separate AWS account; workspace recycled after idle; snapshots kept rolling 90 days inactive; conversation state kept indefinitely unless deleted via API; Privacy Mode (not Legacy); commands auto-run; draft PRs; HSM-signed commits; SOC 2 Type 2 statement; region not stated.
-
Finding: GitHub app permissions for PRs, checks, workflows and branch protection reads; Enterprise Server support; Protected Git Scopes.
-
Finding: Agent routes PRs and can approve low-risk PRs per policy files; GitHub and Origin only; does not replace full code review.
-
Finding: Cloud Agents build and host apps from Origin or GitHub. A user chooses publish and a URL; security checks must pass before a live link is returned. Local agents do not support this feature.
-
Finding: Workspace recycled after the run goes idle (timer refreshes on follow-ups); snapshots kept for a rolling 30 days of inactivity, extended by each start or resume, then deleted automatically and not deletable on demand; conversation state kept indefinitely by default and deletable on demand. Pull requests open ready for review unless a draft is requested, and nothing merges until a person reviews the change; terminal commands auto-run.
cursor:SHIP-05: PR, deployment and check status appear in mobile, and Rollouts monitors deploys for regressions and notifies the author; Rollouts does not roll back on its own, and commit-to-task links were not found.
cursor:SHIP-05
PR, deployment and check status appear in mobile, and Rollouts monitors deploys for regressions and notifies the author; Rollouts does not roll back on its own, and commit-to-task links were not found.
- Rollouts names a suspected change and notifies its author but does not merge or roll back automatically (Sep 23, 2026 changelog).
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: iPhone/iPad app (iOS/iPadOS 26+): start agents, follow-ups, PR review and merge actions, push notifications, Live Activities, Remote Control of local sessions; "Android is planned" with no timing; available on Start, Pro, Pro+, Ultra, Teams, Enterprise.
-
Finding: Assign to Cursor or @Cursor in comments; progress and completion summary with PR link posted to Jira; requires Teams or Enterprise and usage-based pricing; ID linking not described.
-
Finding: Oct 6: remote control for local agents from iOS, on by default except Enterprise. Sep 23: Rollouts and Security Review (Teams/Enterprise; Security Review posts exploitable-bug PR findings). Sep 10: Projects beta.
-
Finding: Plans are editable through chat or Markdown, saved in the home directory by default, and can be moved into the workspace for later reference. Build starts on user choice.
-
Finding: When Rollouts detects a regression it names the change it suspects and notifies the author; the entry states Rollouts does not merge or roll back on its own today.
cursor:SHIP-06: High-volume batch review views were not found; mobile and PR routing give partial support. cross-checked
cursor:SHIP-06
High-volume batch review views were not found; mobile and PR routing give partial support.
- No batch queue, risk filters or exact-revision approval documented in pages read.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: iPhone/iPad app (iOS/iPadOS 26+): start agents, follow-ups, PR review and merge actions, push notifications, Live Activities, Remote Control of local sessions; "Android is planned" with no timing; available on Start, Pro, Pro+, Ultra, Teams, Enterprise.
-
Finding: Agent routes PRs and can approve low-risk PRs per policy files; GitHub and Origin only; does not replace full code review.
cursor:OWN-01: The Terms grant a limited right to access and use the Service and reserve all rights not granted; the public cursor/cursor repository holds no product source. cross-checked
cursor:OWN-01
The Terms grant a limited right to access and use the Service and reserve all rights not granted; the public cursor/cursor repository holds no product source.
- Terms were read at a high level (sections 1.1, 1.5, 5.1); an exact client licence file or EULA was not located.
- Marketplace plugins must be open source, but that does not apply to the client.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Plugins bundle rules, skills, agents, commands, hooks and MCP servers; marketplace plugins are open source and manually reviewed; team marketplaces on Teams/Enterprise.
-
Finding: Public repository with issue templates, README and SECURITY.md only; bugs and requests are directed to the forum; no product source present; 33.3k stars at retrieval.
-
Finding: Limited right to access and use the Service (1.1); reverse engineering and derivative works prohibited (1.5); no implied licences and all rights reserved (5.1); Beta Services may be discontinued at any time (1.6); Anysphere may modify, suspend or discontinue the Services (9, 10). Last updated Sep 3, 2026.
cursor:OWN-02: Cursor is a vendor-hosted service; only tool execution can be self-hosted, and cancellation reverts the account to Hobby. cross-checked
cursor:OWN-02
Cursor is a vendor-hosted service; only tool execution can be self-hosted, and cancellation reverts the account to Hobby.
- Export formats and history portability beyond API/CSV were not established.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Tool execution on own machine while agent loop and inference stay in Cursor cloud; Team Pools need Enterprise; up to 200 workers per user and 1,000 per team; outbound HTTPS only; partner hosts such as AWS Lambda, Modal, Vercel; user pays for own infrastructure.
-
Finding: v1 API in public beta ("APIs may change before general availability"); webhooks not yet in v1; Basic/Bearer API-key auth; agents, runs, stream, cancel, artifacts, environments, secrets, usage endpoints; up to 20 repos; only one active run per agent.
-
Finding: Privacy Mode: no training by Cursor or providers; on by default for Teams and Enterprise; ZDR does not apply to own API keys; subprocessor list at trust.cursor.com.
-
Finding: Enterprise audit log of admin and configuration events (not prompts or agent output); CSV and Admin API export; SIEM streaming; retention and integrity controls not described.
-
Finding: Per-user and per-model usage, Cloud Agent starts and PRs, CSV export; cost per agent not covered.
-
Finding: Cancel via billing portal; access continues to period end; account reverts to Hobby and is preserved.
cursor:OWN-03: Privacy Mode prevents training. Cloud Agents store repository code with documented retention; local agents send model context, and Cursor Origin separately hosts repositories. Usage analytics and Enterprise Conversation Insights are documented; broader telemetry and consent behavior were not observed. cross-checked
cursor:OWN-03
Privacy Mode prevents training. Cloud Agents store repository code with documented retention; local agents send model context, and Cursor Origin separately hosts repositories. Usage analytics and Enterprise Conversation Insights are documented; broader telemetry and consent behavior were not observed.
- Usage analytics are collected from client version 1.5 onward; AI code detection runs on-device and line counts are stored. Enterprise Conversation Insights is enabled by default and can be disabled in team settings. Broader telemetry defaults and consent prompts were not observed. See analytics.
- Privacy Mode default for individual plans is not stated; Teams and Enterprise default on.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Firecracker microVM per agent in a separate AWS account; workspace recycled after idle; snapshots kept rolling 90 days inactive; conversation state kept indefinitely unless deleted via API; Privacy Mode (not Legacy); commands auto-run; draft PRs; HSM-signed commits; SOC 2 Type 2 statement; region not stated.
-
Cloud Agent security and network
Finding: Three egress modes (allow all, default + allowlist, allowlist only); internet on by default; Enterprise can lock policy; runtime secrets redacted in transcripts and commits.
-
Finding: v1 API in public beta ("APIs may change before general availability"); webhooks not yet in v1; Basic/Bearer API-key auth; agents, runs, stream, cancel, artifacts, environments, secrets, usage endpoints; up to 20 repos; only one active run per agent.
-
Finding: OpenAI, Anthropic, Google, Azure OpenAI and AWS Bedrock keys; chat models only; Tab unaffected; whether Agent/Cloud Agents accept keys is not stated.
-
Finding: Privacy Mode: no training by Cursor or providers; on by default for Teams and Enterprise; ZDR does not apply to own API keys; subprocessor list at trust.cursor.com.
-
Finding: Enterprise US-only data residency; some providers restrict models by location; Start plan India only.
-
Finding: Cloud Agents are the only feature requiring Cursor to store code; repo copies deleted after run; Enterprise US residency covers Cloud Agents with 10% uplift; CMEK for Cloud Agent data; telemetry and indexing not covered.
-
Finding: Per-user and per-model usage, Cloud Agent starts and PRs, CSV export; cost per agent not covered.
-
Finding: Cursor git forge in early beta with repositories and pull requests; issue tracking not mentioned.
cursor:OWN-04: New subscriptions cost $8 Plus, $30 Premium, $100 Super or $200 Ultra monthly; Teams seats $30/$100/$200. Legacy rates continue for existing accounts. Third-party models are billed separately. cross-checked
cursor:OWN-04
New subscriptions cost $8 Plus, $30 Premium, $100 Super or $200 Ultra monthly; Teams seats $30/$100/$200. Legacy rates continue for existing accounts. Third-party models are billed separately.
- No workload, billing or authenticated client test was performed; unestablished details remain unknown.
-
Finding: New-customer X Subscription: Free $0, Plus $8, Premium $30, Super $100, Ultra $200 monthly; Teams tiers $30/$100/$200 per user. Third-party model use excluded and separately billed. Legacy Cursor plans continue only for existing customers. Taxes extra.
-
Finding: New X Subscription uses weekly Cursor model pools and prepaid credits for third-party models. Legacy monthly dual pools/on-demand remain for existing accounts. Eligible Teams/Enterprise third-party requests add $0.25/million Cursor Token Rate with stated exemptions.
-
Finding: New X Subscription yearly prices: Plus $80, Premium $300, Super $1,000, Ultra $2,000. Weekly Cursor/Grok usage pool; Plus excludes Cloud and third-party models. Legacy plans remain opt-in migration, not available to new customers.
cursor:OWN-05: Fixed new-plan examples only: one Premium user $30/month; two Premium Teams seats $60; five $150. These totals exclude model/prepaid usage, taxes and user infrastructure. cross-checked
cursor:OWN-05
Fixed new-plan examples only: one Premium user $30/month; two Premium Teams seats $60; five $150. These totals exclude model/prepaid usage, taxes and user infrastructure.
- No workload, billing or authenticated client test was performed; unestablished details remain unknown.
-
Finding: New X Subscription uses weekly Cursor model pools and prepaid credits for third-party models. Legacy monthly dual pools/on-demand remain for existing accounts. Eligible Teams/Enterprise third-party requests add $0.25/million Cursor Token Rate with stated exemptions.
-
Finding: New X Subscription yearly prices: Plus $80, Premium $300, Super $1,000, Ultra $2,000. Weekly Cursor/Grok usage pool; Plus excludes Cloud and third-party models. Legacy plans remain opt-in migration, not available to new customers.
cursor:OWN-06: Cursor announced its completed acquisition by SpaceX on August 14, 2026 and described continued product development. The current release feed includes November 6 voice chat and November 5 app deployment; Cloud API v1 and Origin retain separate beta labels. cross-checked
cursor:OWN-06
Cursor announced its completed acquisition by SpaceX on August 14, 2026 and described continued product development. The current release feed includes November 6 voice chat and November 5 app deployment; Cloud API v1 and Origin retain separate beta labels.
- Status history and response-time commitments were not found beyond Enterprise priority support.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Clients (iOS, web, desktop, Slack, GitHub/Bitbucket comments, Linear, API); paid plan required; managed VMs; environments; no stated concurrency limit; team visibility read-only unless admin enables follow-ups; billed at model API pricing; spend limit set at first use; screenshots, videos and logs as artifacts; draft/merge-ready PRs; Android: open cursor.com/agents in Chrome and tap Install App for a PWA; links to the native iOS app page.
-
Finding: New-customer X Subscription: Free $0, Plus $8, Premium $30, Super $100, Ultra $200 monthly; Teams tiers $30/$100/$200 per user. Third-party model use excluded and separately billed. Legacy Cursor plans continue only for existing customers. Taxes extra.
-
Finding: iPhone/iPad app (iOS/iPadOS 26+): start agents, follow-ups, PR review and merge actions, push notifications, Live Activities, Remote Control of local sessions; "Android is planned" with no timing; available on Start, Pro, Pro+, Ultra, Teams, Enterprise.
-
Finding: Coordinator agent plans and delegates to cloud agents, shared context files sync across machines, subscriptions to Slack/PR/CI/schedules, runs in the cloud; paid plans only, not Privacy Mode (Legacy). Dependencies between tasks, plan view and cost visibility not described.
-
Finding: Cloud Agents v1 remains public beta. Agent/run/stream/cancel, environment/secret and usage operations with API-key auth are documented. Current environment capacity is up to 300 repositories per October 16 release.
-
Finding: Current CLI release index reread November 7 includes persistent sessions, worker improvements and durable goals with rollout gates. Old September version is not reported as latest.
-
Finding: Public repository with issue templates, README and SECURITY.md only; bugs and requests are directed to the forum; no product source present; 33.3k stars at retrieval.
-
Finding: Limited right to access and use the Service (1.1); reverse engineering and derivative works prohibited (1.5); no implied licences and all rights reserved (5.1); Beta Services may be discontinued at any time (1.6); Anysphere may modify, suspend or discontinue the Services (9, 10). Last updated Sep 3, 2026.
-
Finding: New X Subscription yearly prices: Plus $80, Premium $300, Super $1,000, Ultra $2,000. Weekly Cursor/Grok usage pool; Plus excludes Cloud and third-party models. Legacy plans remain opt-in migration, not available to new customers.
-
Finding: Newest November 6 voice chat; November 5 app deployment with explicit publish and security checks; October 20 Agent SDK generally available; October 16 environments API beta and up to 300 repos. Dates are release announcements, not tests.
-
Finding: Week of Sep 21: 'Long-running mode removed from the web'; the Long-running option and team setting were removed. Also Origin CLI 'origin pr stack' removed Oct 1.
-
Finding: Cursor git forge in early beta with repositories and pull requests; issue tracking not mentioned.
-
Cursor is now a part of SpaceX
Finding: Cursor Team states that SpaceX has completed its acquisition of Cursor. The post describes continuing product and model development with SpaceX and states no sunset or service end date.
cursor:OWN-07: Cursor's pricing footer displays "SOC 2 | ISO27001 | ISO42001 | AIUC-1 Certified" (vendor statement, pricing page retrieved 2026-10-09). Cursor's docs state SOC 2 Type II and point to its Trust Center. Scope, issuer, audit period and report date are not verified, and no independent certification is recorded. cross-checked
cursor:OWN-07
Cursor's pricing footer displays "SOC 2 | ISO27001 | ISO42001 | AIUC-1 Certified" (vendor statement, pricing page retrieved 2026-10-09). Cursor's docs state SOC 2 Type II and point to its Trust Center. Scope, issuer, audit period and report date are not verified, and no independent certification is recorded.
- Audit period, scope and report date are not stated; trust.cursor.com rendered no content to the fetch tool.
- Documentation and changelog pass only; no product was installed or run and no shared scenario was executed. Unlisted atomic questions stay unknown.
-
Finding: Hobby free (no card, limited Agent requests); Individual from $20/mo; Teams $40/user/mo; Enterprise custom; on-demand usage billed in arrears; prices exclusive of taxes; yearly toggle without shown prices.
-
Finding: Firecracker microVM per agent in a separate AWS account; workspace recycled after idle; snapshots kept rolling 90 days inactive; conversation state kept indefinitely unless deleted via API; Privacy Mode (not Legacy); commands auto-run; draft PRs; HSM-signed commits; SOC 2 Type 2 statement; region not stated.
-
Finding: States SOC 2 Type II compliance without audit period; reports requested via trust.cursor.com; ISO not mentioned.
-
Finding: Enterprise audit log of admin and configuration events (not prompts or agent output); CSV and Admin API export; SIEM streaming; retention and integrity controls not described.
cursor:OWN-08: Free has limited core use. Plus $8 lacks Cloud Agents; Premium $30 and higher add Cloud access. Existing legacy paid accounts keep their access. cross-checked
cursor:OWN-08
Free has limited core use. Plus $8 lacks Cloud Agents; Premium $30 and higher add Cloud access. Existing legacy paid accounts keep their access.
- No workload, billing or authenticated client test was performed; unestablished details remain unknown.
-
Finding: New-customer X Subscription: Free $0, Plus $8, Premium $30, Super $100, Ultra $200 monthly; Teams tiers $30/$100/$200 per user. Third-party model use excluded and separately billed. Legacy Cursor plans continue only for existing customers. Taxes extra.
-
Finding: New X Subscription yearly prices: Plus $80, Premium $300, Super $1,000, Ultra $2,000. Weekly Cursor/Grok usage pool; Plus excludes Cloud and third-party models. Legacy plans remain opt-in migration, not available to new customers.
cursor:OWN-09: New-plan seat rates and annual prices are public; usage pools are weekly and third-party models prepaid. Legacy billing remains separate. cross-checked
cursor:OWN-09
New-plan seat rates and annual prices are public; usage pools are weekly and third-party models prepaid. Legacy billing remains separate.
- No workload, billing or authenticated client test was performed; unestablished details remain unknown.
-
Finding: New-customer X Subscription: Free $0, Plus $8, Premium $30, Super $100, Ultra $200 monthly; Teams tiers $30/$100/$200 per user. Third-party model use excluded and separately billed. Legacy Cursor plans continue only for existing customers. Taxes extra.
-
Finding: New X Subscription uses weekly Cursor model pools and prepaid credits for third-party models. Legacy monthly dual pools/on-demand remain for existing accounts. Eligible Teams/Enterprise third-party requests add $0.25/million Cursor Token Rate with stated exemptions.
-
Finding: Cancellation keeps access to period end; returns to Free, or Hobby on legacy accounts. Billing problems can block cancellation; no hands-on cancellation performed.
-
Finding: New X Subscription yearly prices: Plus $80, Premium $300, Super $1,000, Ultra $2,000. Weekly Cursor/Grok usage pool; Plus excludes Cloud and third-party models. Legacy plans remain opt-in migration, not available to new customers.