# OpenHands

Podium publishes this catalog and sells an agent development environment. Podium chooses the products, criteria, sources and wording, so treat these pages as vendor-authored research. Check important claims against the linked primary sources. The same evidence rules apply to Podium and its competitors, but they do not remove this conflict of interest.

Page: https://adecomparison.com/ade/openhands/
Website: https://www.openhands.dev/
Research: covered

Method: https://adecomparison.com/method/

Rank by must-haves met, then preferences, then name. Read each finding's checked date. Findings stay published after their scheduled recheck. Checks over six months old may be outdated.

## openhands:ACC-01

Finding: Canvas provides desktop installation paths, browser UI and documented phone/tablet browser access. A separate public source-built native mobile MVP supports iOS/Android direct Agent Server access, including conversation listing and message sending; Cloud login and Canvas pairing remain planned. The product-owned agent-canvas command launches the local stack.

Verification: sourced
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Canvas desktop/browser and launcher; native mobile is limited to direct Agent Server access, without Cloud login or Canvas pairing.
Surface: Agent Canvas; OpenHands Agent Server backends; Public source-built OpenHands native mobile MVP for direct Agent Server access
Version: e14b76cca01a71597cee242c202eed730e745ce3 Canvas launcher; source-built native mobile MVP 47f815b1d111b6a71b405dcb77acb3682da59b85

Only the named atomic answers are established by this source pass. Remaining atomic answers retain their documented gaps.

Source: https://github.com/OpenHands/OpenHands/blob/b811a4c33979ca4b5d241730b995b56362f4ab34/README.md
Locator: Overview, setup, features and component boundaries
Finding: Canvas launches local Agent Server/automation services with Node 24 and uv, or optional Docker runtimes. Backends can live on laptop, remote VM or managed cloud. Same host workspace mounts still share files.
Retrieved: 2026-11-07

Source: https://www.openhands.dev/product/canvas
Locator: ACP; Where your agents run; Get started
Finding: Vendor describes desktop installs, ACP agents with existing subscriptions, extensions and local, self-hosted or managed backends.
Retrieved: 2026-11-07

Source: https://docs.openhands.dev/openhands/usage/agent-canvas/mobile-access
Locator: Tailscale; ngrok
Finding: Canvas documents phone/tablet browser access through Tailscale or authenticated public-mode ngrok; it does not establish native mobile apps.
Retrieved: 2026-11-07

Source: https://github.com/OpenHands/openhands-mobile/blob/47f815b1d111b6a71b405dcb77acb3682da59b85/README.md
Locator: What works today; Prerequisites; Install
Finding: The public native Expo MVP can list conversations, stream history/events and send messages to a reachable Agent Server using a session API key. Source-build instructions cover Android and iOS with native toolchains. Cloud login, QR pairing and a laptop remote-session facade are still planned; no store/release binary is asserted.
Retrieved: 2026-11-13

Source: https://github.com/OpenHands/openhands-mobile/blob/47f815b1d111b6a71b405dcb77acb3682da59b85/src/api/agent-server.ts
Locator: conversation search/create; sendMessage, lines 220-238; src/screens/ChatScreen.tsx
Finding: The native mobile client implements REST conversation listing/history and sends message events to /api/conversations/{id}/events; ChatScreen calls that sendMessage path, and useConversationSocket streams events through WebSocket. app.json declares iOS and Android native platforms.
Retrieved: 2026-11-13

Source: https://github.com/OpenHands/OpenHands/blob/e14b76cca01a71597cee242c202eed730e745ce3/bin/agent-canvas.mjs
Locator: CLI entry point; help/info; local stack startup
Finding: The product-owned agent-canvas CLI starts the local Agent Server, automation backend and pre-built frontend. It exposes version, stack-info, public/frontend-only/backend-only and help flags. This establishes a user-runnable launcher CLI, without asserting a separate task-operation CLI.
Retrieved: 2026-11-13

## openhands:ACC-02

Finding: Canvas integrates OpenHands and ACP agents with custom stdio ACP commands.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Not specified
Surface: Agent Canvas; OpenHands Agent Server backends; OpenHands Cloud
Version: First-party sources checked 2026-11-07; pinned implementations retain their cited revision.

The remaining atomic questions and the rubric scenario are unverified. Documentation is not a successful product test.
Only the named atomic facts are established by documentation. Remaining rubric scenarios and unknown answers were not demonstrated.

Source: https://github.com/OpenHands/OpenHands/blob/b811a4c33979ca4b5d241730b995b56362f4ab34/README.md
Locator: Overview, setup, features and component boundaries
Finding: Canvas launches local Agent Server/automation services with Node 24 and uv, or optional Docker runtimes. Backends can live on laptop, remote VM or managed cloud. Same host workspace mounts still share files.
Retrieved: 2026-11-07

Source: https://www.openhands.dev/product/canvas
Locator: ACP; Where your agents run; Get started
Finding: Vendor describes desktop installs, ACP agents with existing subscriptions, extensions and local, self-hosted or managed backends.
Retrieved: 2026-11-07

Source: https://github.com/OpenHands/OpenHands/blob/b811a4c33979ca4b5d241730b995b56362f4ab34/docs/ACP_AGENTS.md
Locator: Authentication; Per-conversation isolation; Model switching
Finding: Local ACP agents reuse provider-native logins; clean containers/cloud need credentials. Pi/OpenCode are offered only on local/self-hosted backends. Same-provider conversations sharing container HOME can race on auth/config/lock files; Canvas cannot yet send SDK per-conversation data-dir isolation. Agent choice is per backend and running conversations retain their original agent.
Retrieved: 2026-11-07

## openhands:ACC-03

Finding: Local ACP providers can reuse native logins, while clean containers/cloud require credentials and have provider limits.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Not specified
Surface: Agent Canvas; OpenHands Agent Server backends; OpenHands Cloud
Version: First-party sources checked 2026-11-07; pinned implementations retain their cited revision.

The remaining atomic questions and the rubric scenario are unverified. Documentation is not a successful product test.
Only the named atomic facts are established by documentation. Remaining rubric scenarios and unknown answers were not demonstrated.

Source: https://www.openhands.dev/product/canvas
Locator: ACP; Where your agents run; Get started
Finding: Vendor describes desktop installs, ACP agents with existing subscriptions, extensions and local, self-hosted or managed backends.
Retrieved: 2026-11-07

Source: https://github.com/OpenHands/OpenHands/blob/b811a4c33979ca4b5d241730b995b56362f4ab34/docs/ACP_AGENTS.md
Locator: Authentication; Per-conversation isolation; Model switching
Finding: Local ACP agents reuse provider-native logins; clean containers/cloud need credentials. Pi/OpenCode are offered only on local/self-hosted backends. Same-provider conversations sharing container HOME can race on auth/config/lock files; Canvas cannot yet send SDK per-conversation data-dir isolation. Agent choice is per backend and running conversations retain their original agent.
Retrieved: 2026-11-07

## openhands:ACC-04

Finding: Local stack requires Node 24 and uv; Docker and hardened VM routes are separate.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Not specified
Surface: Agent Canvas; OpenHands Agent Server backends; OpenHands Cloud
Version: First-party sources checked 2026-11-07; pinned implementations retain their cited revision.

The remaining atomic questions and the rubric scenario are unverified. Documentation is not a successful product test.

Source: https://github.com/OpenHands/OpenHands/blob/b811a4c33979ca4b5d241730b995b56362f4ab34/README.md
Locator: Overview, setup, features and component boundaries
Finding: Canvas launches local Agent Server/automation services with Node 24 and uv, or optional Docker runtimes. Backends can live on laptop, remote VM or managed cloud. Same host workspace mounts still share files.
Retrieved: 2026-11-07

Source: https://github.com/OpenHands/OpenHands/blob/b811a4c33979ca4b5d241730b995b56362f4ab34/docs/SELF_HOSTING.md
Locator: Quickstart; public mode; security posture
Finding: User provisions and hardens a VM; public mode requires an API key, with TLS/firewall setup and shared-origin editor caveats.
Retrieved: 2026-11-07

## openhands:ACC-05

Finding: Canvas uses the Agent Server REST API; SDK ownership is separate.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Not specified
Surface: Agent Canvas; OpenHands Agent Server backends; OpenHands Cloud
Version: First-party sources checked 2026-11-07; pinned implementations retain their cited revision.

The remaining atomic questions and the rubric scenario are unverified. Documentation is not a successful product test.

Source: https://github.com/OpenHands/OpenHands/blob/b811a4c33979ca4b5d241730b995b56362f4ab34/README.md
Locator: Overview, setup, features and component boundaries
Finding: Canvas launches local Agent Server/automation services with Node 24 and uv, or optional Docker runtimes. Backends can live on laptop, remote VM or managed cloud. Same host workspace mounts still share files.
Retrieved: 2026-11-07

## openhands:ACC-06

Finding: Canvas manages backend-scoped MCP, skills and plugins, with health and credential checks.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Not specified
Surface: Agent Canvas; OpenHands Agent Server backends; OpenHands Cloud
Version: First-party sources checked 2026-11-07; pinned implementations retain their cited revision.

The remaining atomic questions and the rubric scenario are unverified. Documentation is not a successful product test.
Only the named atomic facts are established by documentation. Remaining rubric scenarios and unknown answers were not demonstrated.

Source: https://www.openhands.dev/product/canvas
Locator: ACP; Where your agents run; Get started
Finding: Vendor describes desktop installs, ACP agents with existing subscriptions, extensions and local, self-hosted or managed backends.
Retrieved: 2026-11-07

Source: https://docs.openhands.dev/openhands/usage/agent-canvas/customize-and-settings
Locator: Persistent memory; MCP health; Per-backend settings
Finding: Canvas loads workspace/user persistent memory when enabled. Skills/MCP/secrets are backend-scoped. MCP health distinguishes reachable, rejected credentials and connection failure; public-server reachability does not verify credentials. Skills load on new conversation start.
Retrieved: 2026-11-07

## openhands:ACC-07

Finding: Canvas is a React/TypeScript web frontend.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2027-02-05
Plan: Not specified
Surface: Agent Canvas
Version: First-party sources checked 2026-11-07; pinned implementations retain their cited revision.

The remaining atomic questions and the rubric scenario are unverified. Documentation is not a successful product test.

Source: https://github.com/OpenHands/OpenHands/blob/b811a4c33979ca4b5d241730b995b56362f4ab34/docs/architecture.md
Locator: System boundaries; Runtime modes; Packaging
Finding: Canvas renders conversations and controls backend APIs; execution, isolation and automation belong to separate runtime services.
Retrieved: 2026-11-07

## openhands:PLAN-04

Finding: Canvas documents skills and a persistent workspace/user memory setting that can be enabled or disabled.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Agent Canvas unless answer explicitly scopes paid Teams/Enterprise
Surface: Agent Canvas; OpenHands Agent Server backends
Version: First-party documentation/source checked 2026-11-07; exact pinned source revision is recorded in each citation. No installed-product fixture.

Only the named atomic facts are established by documentation. Remaining rubric scenarios and unknown answers were not demonstrated.

Source: https://docs.openhands.dev/openhands/usage/agent-canvas/customize-and-settings
Locator: Persistent memory; MCP health; Per-backend settings
Finding: Canvas loads workspace/user persistent memory when enabled. Skills/MCP/secrets are backend-scoped. MCP health distinguishes reachable, rejected credentials and connection failure; public-server reachability does not verify credentials. Skills load on new conversation start.
Retrieved: 2026-11-07

## openhands:PLAN-06

Finding: The Canvas Task List renders the latest TaskTrackerObservation from the current conversation, with position-derived IDs and todo/in-progress/done states. Bounded inspection found no independent native issue/backlog work-item store; saved automations are trigger configurations.

Verification: sourced
Checked: 2026-11-13
Scheduled recheck: 2027-03-13
Plan: Not specified
Surface: Agent Canvas
Version: e14b76cca01a71597cee242c202eed730e745ce3 native Canvas Task List and application stores/routes.

Only the named atomic answers are established by this source pass. Remaining atomic answers retain their documented gaps.

Source: https://github.com/OpenHands/OpenHands/blob/e14b76cca01a71597cee242c202eed730e745ce3/src/hooks/use-task-list.ts
Locator: getTaskListFromEvent; useTaskList; routes/task-list-tab.tsx
Finding: The Canvas Task List derives its entries from the latest TaskTrackerObservation in the current conversation event store. IDs are generated from list position and statuses are todo/in_progress/done. Inspection of the native task-list route and API/store surfaces found no independent issue/backlog work-item entity; automations are separate trigger definitions.
Retrieved: 2026-11-13

## openhands:COORD-01

Finding: Parallel conversations can use separate Docker runtimes.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Not specified
Surface: Agent Canvas; OpenHands Agent Server backends; OpenHands Cloud
Version: First-party sources checked 2026-11-07; pinned implementations retain their cited revision.

The remaining atomic questions and the rubric scenario are unverified. Documentation is not a successful product test.

Source: https://github.com/OpenHands/OpenHands/blob/b811a4c33979ca4b5d241730b995b56362f4ab34/README.md
Locator: Overview, setup, features and component boundaries
Finding: Canvas launches local Agent Server/automation services with Node 24 and uv, or optional Docker runtimes. Backends can live on laptop, remote VM or managed cloud. Same host workspace mounts still share files.
Retrieved: 2026-11-07

## openhands:COORD-03

Finding: Canvas launches linked child conversations and returns their results to the parent.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Agent Canvas unless answer explicitly scopes paid Teams/Enterprise
Surface: Agent Canvas; OpenHands Agent Server backends; OpenHands Cloud
Version: First-party documentation/source checked 2026-11-07; exact pinned source revision is recorded in each citation. No installed-product fixture.

Only the named atomic facts are established by documentation. Remaining rubric scenarios and unknown answers were not demonstrated.

Source: https://docs.openhands.dev/openhands/usage/agent-canvas/conversations
Locator: Child conversations; Commits drawer; Goals; Export; Archive
Finding: Canvas child conversations stay linked and return results; local children may use isolated worktrees or shared workspace. Commit drawer shows diffs and agent Git/PR actions. Goals use a judge with round caps and user-message interruption. Transcripts export locally as Markdown/HTML with tool details; archive keeps backend history.
Retrieved: 2026-11-07

## openhands:COORD-04

Finding: Automation Server owns schedules/events and Canvas exposes lifecycle/run history.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Not specified
Surface: Agent Canvas; OpenHands Agent Server backends; OpenHands Cloud
Version: First-party sources checked 2026-11-07; pinned implementations retain their cited revision.

The remaining atomic questions and the rubric scenario are unverified. Documentation is not a successful product test.
Only the named atomic facts are established by documentation. Remaining rubric scenarios and unknown answers were not demonstrated.

Source: https://github.com/OpenHands/OpenHands/blob/b811a4c33979ca4b5d241730b995b56362f4ab34/README.md
Locator: Overview, setup, features and component boundaries
Finding: Canvas launches local Agent Server/automation services with Node 24 and uv, or optional Docker runtimes. Backends can live on laptop, remote VM or managed cloud. Same host workspace mounts still share files.
Retrieved: 2026-11-07

Source: https://github.com/OpenHands/OpenHands/blob/b811a4c33979ca4b5d241730b995b56362f4ab34/docs/architecture.md
Locator: System boundaries; Runtime modes; Packaging
Finding: Canvas renders conversations and controls backend APIs; execution, isolation and automation belong to separate runtime services.
Retrieved: 2026-11-07

Source: https://docs.openhands.dev/openhands/usage/agent-canvas/managing-automations
Locator: Run statuses; Activity costs; Shared automation conversations; Export/import
Finding: Automations have pending/running/completed/failed/cancelled/skipped status; concurrency limits may skip runs. Activity exports CSV/JSON with numeric USD cost or null. Automations export/import versioned JSON; disabling preserves configuration. Other Cloud organization users can view run conversations read-only but cannot send messages.
Retrieved: 2026-11-07

## openhands:COORD-05

Finding: Canvas supports goal interruption and preserves failed-message retry controls.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Agent Canvas unless answer explicitly scopes paid Teams/Enterprise
Surface: Agent Canvas; OpenHands Agent Server backends
Version: First-party documentation/source checked 2026-11-07; exact pinned source revision is recorded in each citation. No installed-product fixture.

Only the named atomic facts are established by documentation. Remaining rubric scenarios and unknown answers were not demonstrated.

Source: https://docs.openhands.dev/openhands/usage/agent-canvas/conversations
Locator: Child conversations; Commits drawer; Goals; Export; Archive
Finding: Canvas child conversations stay linked and return results; local children may use isolated worktrees or shared workspace. Commit drawer shows diffs and agent Git/PR actions. Goals use a judge with round caps and user-message interruption. Transcripts export locally as Markdown/HTML with tool details; archive keeps backend history.
Retrieved: 2026-11-07

## openhands:COORD-06

Finding: Canvas shows linked children and grouped conversation metadata. Fleet fixture remains untested.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Agent Canvas unless answer explicitly scopes paid Teams/Enterprise
Surface: Agent Canvas; OpenHands Agent Server backends
Version: First-party documentation/source checked 2026-11-07; exact pinned source revision is recorded in each citation. No installed-product fixture.

Only the named atomic facts are established by documentation. Remaining rubric scenarios and unknown answers were not demonstrated.

Source: https://docs.openhands.dev/openhands/usage/agent-canvas/conversations
Locator: Child conversations; Commits drawer; Goals; Export; Archive
Finding: Canvas child conversations stay linked and return results; local children may use isolated worktrees or shared workspace. Commit drawer shows diffs and agent Git/PR actions. Goals use a judge with round caps and user-message interruption. Transcripts export locally as Markdown/HTML with tool details; archive keeps backend history.
Retrieved: 2026-11-07

## openhands:COORD-07

Finding: Canvas has reusable profiles, goals and backend-specific configuration.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Agent Canvas unless answer explicitly scopes paid Teams/Enterprise
Surface: Agent Canvas; OpenHands Agent Server backends
Version: First-party documentation/source checked 2026-11-07; exact pinned source revision is recorded in each citation. No installed-product fixture.

Only the named atomic facts are established by documentation. Remaining rubric scenarios and unknown answers were not demonstrated.

Source: https://docs.openhands.dev/openhands/usage/agent-canvas/conversations
Locator: Child conversations; Commits drawer; Goals; Export; Archive
Finding: Canvas child conversations stay linked and return results; local children may use isolated worktrees or shared workspace. Commit drawer shows diffs and agent Git/PR actions. Goals use a judge with round caps and user-message interruption. Transcripts export locally as Markdown/HTML with tool details; archive keeps backend history.
Retrieved: 2026-11-07

Source: https://docs.openhands.dev/openhands/usage/agent-canvas/customize-and-settings
Locator: Persistent memory; MCP health; Per-backend settings
Finding: Canvas loads workspace/user persistent memory when enabled. Skills/MCP/secrets are backend-scoped. MCP health distinguishes reachable, rejected credentials and connection failure; public-server reachability does not verify credentials. Skills load on new conversation start.
Retrieved: 2026-11-07

## openhands:RUN-01

Finding: Canvas switches between local, self-hosted remote and optional cloud backends.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Not specified
Surface: Agent Canvas; OpenHands Agent Server backends; OpenHands Cloud
Version: First-party sources checked 2026-11-07; pinned implementations retain their cited revision.

The remaining atomic questions and the rubric scenario are unverified. Documentation is not a successful product test.

Source: https://github.com/OpenHands/OpenHands/blob/b811a4c33979ca4b5d241730b995b56362f4ab34/README.md
Locator: Overview, setup, features and component boundaries
Finding: Canvas launches local Agent Server/automation services with Node 24 and uv, or optional Docker runtimes. Backends can live on laptop, remote VM or managed cloud. Same host workspace mounts still share files.
Retrieved: 2026-11-07

Source: https://www.openhands.dev/product/canvas
Locator: ACP; Where your agents run; Get started
Finding: Vendor describes desktop installs, ACP agents with existing subscriptions, extensions and local, self-hosted or managed backends.
Retrieved: 2026-11-07

## openhands:RUN-02

Finding: Optional containers and worktrees separate execution/files, with documented shared-HOME and shared-workspace limits.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Not specified
Surface: Agent Canvas; OpenHands Agent Server backends; OpenHands Cloud
Version: First-party sources checked 2026-11-07; pinned implementations retain their cited revision.

The remaining atomic questions and the rubric scenario are unverified. Documentation is not a successful product test.
Only the named atomic facts are established by documentation. Remaining rubric scenarios and unknown answers were not demonstrated.

Source: https://github.com/OpenHands/OpenHands/blob/b811a4c33979ca4b5d241730b995b56362f4ab34/README.md
Locator: Overview, setup, features and component boundaries
Finding: Canvas launches local Agent Server/automation services with Node 24 and uv, or optional Docker runtimes. Backends can live on laptop, remote VM or managed cloud. Same host workspace mounts still share files.
Retrieved: 2026-11-07

Source: https://docs.openhands.dev/openhands/usage/agent-canvas/conversations
Locator: Child conversations; Commits drawer; Goals; Export; Archive
Finding: Canvas child conversations stay linked and return results; local children may use isolated worktrees or shared workspace. Commit drawer shows diffs and agent Git/PR actions. Goals use a judge with round caps and user-message interruption. Transcripts export locally as Markdown/HTML with tool details; archive keeps backend history.
Retrieved: 2026-11-07

Source: https://github.com/OpenHands/OpenHands/blob/b811a4c33979ca4b5d241730b995b56362f4ab34/docs/ACP_AGENTS.md
Locator: Authentication; Per-conversation isolation; Model switching
Finding: Local ACP agents reuse provider-native logins; clean containers/cloud need credentials. Pi/OpenCode are offered only on local/self-hosted backends. Same-provider conversations sharing container HOME can race on auth/config/lock files; Canvas cannot yet send SDK per-conversation data-dir isolation. Agent choice is per backend and running conversations retain their original agent.
Retrieved: 2026-11-07

## openhands:RUN-03

Finding: Canvas reloads backend conversation history and keeps archived conversations without deleting them.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Not specified
Surface: Agent Canvas; OpenHands Agent Server backends; OpenHands Cloud
Version: First-party sources checked 2026-11-07; pinned implementations retain their cited revision.

The remaining atomic questions and the rubric scenario are unverified. Documentation is not a successful product test.
Only the named atomic facts are established by documentation. Remaining rubric scenarios and unknown answers were not demonstrated.

Source: https://github.com/OpenHands/OpenHands/blob/b811a4c33979ca4b5d241730b995b56362f4ab34/README.md
Locator: Overview, setup, features and component boundaries
Finding: Canvas launches local Agent Server/automation services with Node 24 and uv, or optional Docker runtimes. Backends can live on laptop, remote VM or managed cloud. Same host workspace mounts still share files.
Retrieved: 2026-11-07

Source: https://docs.openhands.dev/openhands/usage/agent-canvas/conversations
Locator: Child conversations; Commits drawer; Goals; Export; Archive
Finding: Canvas child conversations stay linked and return results; local children may use isolated worktrees or shared workspace. Commit drawer shows diffs and agent Git/PR actions. Goals use a judge with round caps and user-message interruption. Transcripts export locally as Markdown/HTML with tool details; archive keeps backend history.
Retrieved: 2026-11-07

## openhands:RUN-04

Finding: The user chooses workspace mounts and runtime configuration.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Not specified
Surface: Agent Canvas; OpenHands Agent Server backends; OpenHands Cloud
Version: First-party sources checked 2026-11-07; pinned implementations retain their cited revision.

The remaining atomic questions and the rubric scenario are unverified. Documentation is not a successful product test.

Source: https://github.com/OpenHands/OpenHands/blob/b811a4c33979ca4b5d241730b995b56362f4ab34/README.md
Locator: Overview, setup, features and component boundaries
Finding: Canvas launches local Agent Server/automation services with Node 24 and uv, or optional Docker runtimes. Backends can live on laptop, remote VM or managed cloud. Same host workspace mounts still share files.
Retrieved: 2026-11-07

## openhands:RUN-05

Finding: Goals and experimental critic refinement have explicit iteration limits.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Agent Canvas unless answer explicitly scopes paid Teams/Enterprise
Surface: Agent Canvas; OpenHands Agent Server backends
Version: First-party documentation/source checked 2026-11-07; exact pinned source revision is recorded in each citation. No installed-product fixture.

Only the named atomic facts are established by documentation. Remaining rubric scenarios and unknown answers were not demonstrated.

Source: https://docs.openhands.dev/openhands/usage/agent-canvas/conversations
Locator: Child conversations; Commits drawer; Goals; Export; Archive
Finding: Canvas child conversations stay linked and return results; local children may use isolated worktrees or shared workspace. Commit drawer shows diffs and agent Git/PR actions. Goals use a judge with round caps and user-message interruption. Transcripts export locally as Markdown/HTML with tool details; archive keeps backend history.
Retrieved: 2026-11-07

Source: https://docs.openhands.dev/openhands/usage/agent-canvas/critic
Locator: Experimental critic; Iterative refinement
Finding: Experimental critic applies to OpenHands-agent conversations; ACP agents may not expose the path. Default threshold 0.6 and maximum three refinement attempts. Results expose score and detected issue categories; this is additional evaluation, not a successful acceptance-test result.
Retrieved: 2026-11-07

Source: https://docs.openhands.dev/openhands/usage/cloud/organizations/budgets
Locator: Enterprise managed gateway budgets; Usage; Why totals differ
Finding: Enterprise managed-gateway budgets cap organization/user monthly spend. Gateway admits requests and can block at limits; direct provider calls are outside enforcement. Dashboard conversation estimates and gateway budget counters differ in scope/freshness. Spend may be live/stale/unavailable, not zero when unavailable; threshold alerts go to admins/owners.
Retrieved: 2026-11-07

## openhands:RUN-06

Finding: Canvas displays automation phase/status and exports reported costs with null for unavailable cost.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Not specified
Surface: Agent Canvas; OpenHands Agent Server backends; OpenHands Cloud
Version: First-party sources checked 2026-11-07; pinned implementations retain their cited revision.

The remaining atomic questions and the rubric scenario are unverified. Documentation is not a successful product test.
Only the named atomic facts are established by documentation. Remaining rubric scenarios and unknown answers were not demonstrated.

Source: https://github.com/OpenHands/OpenHands/blob/b811a4c33979ca4b5d241730b995b56362f4ab34/README.md
Locator: Overview, setup, features and component boundaries
Finding: Canvas launches local Agent Server/automation services with Node 24 and uv, or optional Docker runtimes. Backends can live on laptop, remote VM or managed cloud. Same host workspace mounts still share files.
Retrieved: 2026-11-07

Source: https://docs.openhands.dev/openhands/usage/agent-canvas/managing-automations
Locator: Run statuses; Activity costs; Shared automation conversations; Export/import
Finding: Automations have pending/running/completed/failed/cancelled/skipped status; concurrency limits may skip runs. Activity exports CSV/JSON with numeric USD cost or null. Automations export/import versioned JSON; disabling preserves configuration. Other Cloud organization users can view run conversations read-only but cannot send messages.
Retrieved: 2026-11-07

## openhands:RUN-07

Finding: Canvas automation history exposes reported per-run LLM cost with explicit missing values.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Agent Canvas unless answer explicitly scopes paid Teams/Enterprise
Surface: Agent Canvas; OpenHands Agent Server backends; OpenHands Cloud
Version: First-party documentation/source checked 2026-11-07; exact pinned source revision is recorded in each citation. No installed-product fixture.

Only the named atomic facts are established by documentation. Remaining rubric scenarios and unknown answers were not demonstrated.

Source: https://docs.openhands.dev/openhands/usage/agent-canvas/managing-automations
Locator: Run statuses; Activity costs; Shared automation conversations; Export/import
Finding: Automations have pending/running/completed/failed/cancelled/skipped status; concurrency limits may skip runs. Activity exports CSV/JSON with numeric USD cost or null. Automations export/import versioned JSON; disabling preserves configuration. Other Cloud organization users can view run conversations read-only but cannot send messages.
Retrieved: 2026-11-07

Source: https://docs.openhands.dev/openhands/usage/cloud/organizations/budgets
Locator: Enterprise managed gateway budgets; Usage; Why totals differ
Finding: Enterprise managed-gateway budgets cap organization/user monthly spend. Gateway admits requests and can block at limits; direct provider calls are outside enforcement. Dashboard conversation estimates and gateway budget counters differ in scope/freshness. Spend may be live/stale/unavailable, not zero when unavailable; threshold alerts go to admins/owners.
Retrieved: 2026-11-07

## openhands:RUN-08

Finding: Backend concurrency caps can skip automation dispatch; resource-aware fairness remains unknown.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Agent Canvas unless answer explicitly scopes paid Teams/Enterprise
Surface: Agent Canvas; OpenHands Agent Server backends; OpenHands Cloud
Version: First-party documentation/source checked 2026-11-07; exact pinned source revision is recorded in each citation. No installed-product fixture.

Only the named atomic facts are established by documentation. Remaining rubric scenarios and unknown answers were not demonstrated.

Source: https://docs.openhands.dev/openhands/usage/agent-canvas/managing-automations
Locator: Run statuses; Activity costs; Shared automation conversations; Export/import
Finding: Automations have pending/running/completed/failed/cancelled/skipped status; concurrency limits may skip runs. Activity exports CSV/JSON with numeric USD cost or null. Automations export/import versioned JSON; disabling preserves configuration. Other Cloud organization users can view run conversations read-only but cannot send messages.
Retrieved: 2026-11-07

## openhands:TEAM-01

Finding: Cloud organization automation conversations support read-only team visibility. Same-agent co-prompting remains unestablished.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Cloud organization automation conversations
Surface: OpenHands Cloud
Version: Not specified

Only the named atomic facts are established by documentation. Remaining rubric scenarios and unknown answers were not demonstrated.

Source: https://docs.openhands.dev/openhands/usage/agent-canvas/managing-automations
Locator: Run statuses; Activity costs; Shared automation conversations; Export/import
Finding: Automations have pending/running/completed/failed/cancelled/skipped status; concurrency limits may skip runs. Activity exports CSV/JSON with numeric USD cost or null. Automations export/import versioned JSON; disabling preserves configuration. Other Cloud organization users can view run conversations read-only but cannot send messages.
Retrieved: 2026-11-07

## openhands:TEAM-02

Finding: Organizations document Member/Admin/Owner permissions; paid Teams/Enterprise separately advertise SSO/SAML.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Cloud organization roles; Teams/Enterprise self-hosted SSO/SAML
Surface: OpenHands Cloud; OpenHands Agent Server backends
Version: Not specified

Only the named atomic facts are established by documentation. Remaining rubric scenarios and unknown answers were not demonstrated.

Source: https://docs.openhands.dev/openhands/usage/cloud/organizations/roles-permissions
Locator: Roles table; Private settings
Finding: Cloud/Enterprise organization roles are Member, Admin and Owner. Members create conversations and own private MCP/secrets/settings; Admins invite/promote users and manage LLM settings/credits; Owners delete organizations and claim Git organizations. Private settings remain private from other members including admins.
Retrieved: 2026-11-07

Source: https://www.openhands.dev/pricing
Locator: Plan cards; Comparison table
Finding: Canvas is free, MIT, single-user/shared-identity and self-hosted. Teams is $500 monthly for up to ten users with $50 monthly per additional seat, 5000 sessions per month, 30-day trial, RBAC and SSO/SAML. Enterprise has custom pricing. LLM keys are user-provided.
Retrieved: 2026-11-07

Source: https://docs.openhands.dev/enterprise/integrations/saml-sso
Locator: Prerequisites, Step 1 and Step 2 sections
Finding: The guide explains how to let users of an OpenHands Enterprise installation sign in through a corporate SAML identity provider such as Okta, Entra ID, Google Workspace or ADFS.
Retrieved: 2026-11-10

Source: https://www.openhands.dev/pricing
Locator: Plan cards and comparison table (Agent Canvas, Teams, Enterprise)
Finding: Agent Canvas is free and MIT licensed. Teams costs 500 USD per month for up to ten users plus 50 USD per extra seat and lists SSO/SAML and RBAC; Enterprise has custom pricing. The page returns HTTP 403 to some automated fetchers but loads with a browser user agent.
Retrieved: 2026-11-10

## openhands:TEAM-03

Finding: Self-hosted public mode requires the backend API key.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Not specified
Surface: Agent Canvas; OpenHands Agent Server backends
Version: First-party sources checked 2026-11-07; pinned implementations retain their cited revision.

The remaining atomic questions and the rubric scenario are unverified. Documentation is not a successful product test.

Source: https://github.com/OpenHands/OpenHands/blob/b811a4c33979ca4b5d241730b995b56362f4ab34/docs/SELF_HOSTING.md
Locator: Quickstart; public mode; security posture
Finding: User provisions and hardens a VM; public mode requires an API key, with TLS/firewall setup and shared-origin editor caveats.
Retrieved: 2026-11-07

## openhands:TEAM-04

Finding: Canvas exports timestamped conversation/tool history and retains archived messages.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Agent Canvas unless answer explicitly scopes paid Teams/Enterprise
Surface: Agent Canvas; OpenHands Agent Server backends
Version: First-party documentation/source checked 2026-11-07; exact pinned source revision is recorded in each citation. No installed-product fixture.

Only the named atomic facts are established by documentation. Remaining rubric scenarios and unknown answers were not demonstrated.

Source: https://docs.openhands.dev/openhands/usage/agent-canvas/conversations
Locator: Child conversations; Commits drawer; Goals; Export; Archive
Finding: Canvas child conversations stay linked and return results; local children may use isolated worktrees or shared workspace. Commit drawer shows diffs and agent Git/PR actions. Goals use a judge with round caps and user-message interruption. Transcripts export locally as Markdown/HTML with tool details; archive keeps backend history.
Retrieved: 2026-11-07

## openhands:TEAM-05

Finding: Canvas exposes combined commit/uncommitted diffs and experimental critic feedback.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Agent Canvas unless answer explicitly scopes paid Teams/Enterprise
Surface: Agent Canvas; OpenHands Agent Server backends
Version: First-party documentation/source checked 2026-11-07; exact pinned source revision is recorded in each citation. No installed-product fixture.

Only the named atomic facts are established by documentation. Remaining rubric scenarios and unknown answers were not demonstrated.

Source: https://docs.openhands.dev/openhands/usage/agent-canvas/conversations
Locator: Child conversations; Commits drawer; Goals; Export; Archive
Finding: Canvas child conversations stay linked and return results; local children may use isolated worktrees or shared workspace. Commit drawer shows diffs and agent Git/PR actions. Goals use a judge with round caps and user-message interruption. Transcripts export locally as Markdown/HTML with tool details; archive keeps backend history.
Retrieved: 2026-11-07

Source: https://docs.openhands.dev/openhands/usage/agent-canvas/critic
Locator: Experimental critic; Iterative refinement
Finding: Experimental critic applies to OpenHands-agent conversations; ACP agents may not expose the path. Default threshold 0.6 and maximum three refinement attempts. Results expose score and detected issue categories; this is additional evaluation, not a successful acceptance-test result.
Retrieved: 2026-11-07

## openhands:SHIP-01

Finding: Canvas renders file artifacts, images and unified code-change views.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Not specified
Surface: Agent Canvas; OpenHands Agent Server backends; OpenHands Cloud
Version: First-party sources checked 2026-11-07; pinned implementations retain their cited revision.

The remaining atomic questions and the rubric scenario are unverified. Documentation is not a successful product test.
Only the named atomic facts are established by documentation. Remaining rubric scenarios and unknown answers were not demonstrated.

Source: https://github.com/OpenHands/OpenHands/blob/b811a4c33979ca4b5d241730b995b56362f4ab34/docs/architecture.md
Locator: System boundaries; Runtime modes; Packaging
Finding: Canvas renders conversations and controls backend APIs; execution, isolation and automation belong to separate runtime services.
Retrieved: 2026-11-07

Source: https://docs.openhands.dev/openhands/usage/agent-canvas/conversations
Locator: Child conversations; Commits drawer; Goals; Export; Archive
Finding: Canvas child conversations stay linked and return results; local children may use isolated worktrees or shared workspace. Commit drawer shows diffs and agent Git/PR actions. Goals use a judge with round caps and user-message interruption. Transcripts export locally as Markdown/HTML with tool details; archive keeps backend history.
Retrieved: 2026-11-07

## openhands:SHIP-02

Finding: Experimental critic evaluates OpenHands-agent results and can request bounded refinement.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Agent Canvas unless answer explicitly scopes paid Teams/Enterprise
Surface: Agent Canvas; OpenHands Agent Server backends
Version: First-party documentation/source checked 2026-11-07; exact pinned source revision is recorded in each citation. No installed-product fixture.

Only the named atomic facts are established by documentation. Remaining rubric scenarios and unknown answers were not demonstrated.

Source: https://docs.openhands.dev/openhands/usage/agent-canvas/critic
Locator: Experimental critic; Iterative refinement
Finding: Experimental critic applies to OpenHands-agent conversations; ACP agents may not expose the path. Default threshold 0.6 and maximum three refinement attempts. Results expose score and detected issue categories; this is additional evaluation, not a successful acceptance-test result.
Retrieved: 2026-11-07

## openhands:SHIP-03

Finding: Canvas Git controls prompt the agent to commit, pull, push and work with PRs.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Agent Canvas unless answer explicitly scopes paid Teams/Enterprise
Surface: Agent Canvas; OpenHands Agent Server backends
Version: First-party documentation/source checked 2026-11-07; exact pinned source revision is recorded in each citation. No installed-product fixture.

Only the named atomic facts are established by documentation. Remaining rubric scenarios and unknown answers were not demonstrated.

Source: https://docs.openhands.dev/openhands/usage/agent-canvas/conversations
Locator: Child conversations; Commits drawer; Goals; Export; Archive
Finding: Canvas child conversations stay linked and return results; local children may use isolated worktrees or shared workspace. Commit drawer shows diffs and agent Git/PR actions. Goals use a judge with round caps and user-message interruption. Transcripts export locally as Markdown/HTML with tool details; archive keeps backend history.
Retrieved: 2026-11-07

## openhands:OWN-01

Finding: The Canvas source distribution is MIT.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2027-02-05
Plan: Not specified
Surface: Agent Canvas
Version: First-party sources checked 2026-11-07; pinned implementations retain their cited revision.

The remaining atomic questions and the rubric scenario are unverified. Documentation is not a successful product test.

Source: https://github.com/OpenHands/OpenHands/blob/b811a4c33979ca4b5d241730b995b56362f4ab34/LICENSE
Locator: Root license grant
Finding: The cited root license governs this repository distribution, subject to component notices.
Retrieved: 2026-11-07

## openhands:OWN-02

Finding: Canvas self-hosts and provides portable conversation and automation exports.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Not specified
Surface: Agent Canvas; OpenHands Agent Server backends
Version: First-party sources checked 2026-11-07; pinned implementations retain their cited revision.

The remaining atomic questions and the rubric scenario are unverified. Documentation is not a successful product test.
Only the named atomic facts are established by documentation. Remaining rubric scenarios and unknown answers were not demonstrated.

Source: https://github.com/OpenHands/OpenHands/blob/b811a4c33979ca4b5d241730b995b56362f4ab34/README.md
Locator: Overview, setup, features and component boundaries
Finding: Canvas launches local Agent Server/automation services with Node 24 and uv, or optional Docker runtimes. Backends can live on laptop, remote VM or managed cloud. Same host workspace mounts still share files.
Retrieved: 2026-11-07

Source: https://github.com/OpenHands/OpenHands/blob/b811a4c33979ca4b5d241730b995b56362f4ab34/docs/SELF_HOSTING.md
Locator: Quickstart; public mode; security posture
Finding: User provisions and hardens a VM; public mode requires an API key, with TLS/firewall setup and shared-origin editor caveats.
Retrieved: 2026-11-07

Source: https://docs.openhands.dev/openhands/usage/agent-canvas/conversations
Locator: Child conversations; Commits drawer; Goals; Export; Archive
Finding: Canvas child conversations stay linked and return results; local children may use isolated worktrees or shared workspace. Commit drawer shows diffs and agent Git/PR actions. Goals use a judge with round caps and user-message interruption. Transcripts export locally as Markdown/HTML with tool details; archive keeps backend history.
Retrieved: 2026-11-07

Source: https://docs.openhands.dev/openhands/usage/agent-canvas/managing-automations
Locator: Run statuses; Activity costs; Shared automation conversations; Export/import
Finding: Automations have pending/running/completed/failed/cancelled/skipped status; concurrency limits may skip runs. Activity exports CSV/JSON with numeric USD cost or null. Automations export/import versioned JSON; disabling preserves configuration. Other Cloud organization users can view run conversations read-only but cannot send messages.
Retrieved: 2026-11-07

## openhands:OWN-03

Finding: Backend settings/secrets and browser-generated exports have explicit boundaries. At v1.38.0 an anonymous usage-data switch defaults on with a consent prompt and can be turned off on local and self-hosted backends; it is forced on for OpenHands Cloud.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-12-07
Plan: Agent Canvas unless answer explicitly scopes paid Teams/Enterprise
Surface: Agent Canvas; OpenHands Agent Server backends
Version: Telemetry from OpenHands v1.38.0 source (e14b76c); other parts from first-party documentation checked 2026-11-07. No installed-product fixture.

Only the named atomic facts are established by documentation. Remaining rubric scenarios and unknown answers were not demonstrated.

Source: https://docs.openhands.dev/openhands/usage/agent-canvas/conversations
Locator: Child conversations; Commits drawer; Goals; Export; Archive
Finding: Canvas child conversations stay linked and return results; local children may use isolated worktrees or shared workspace. Commit drawer shows diffs and agent Git/PR actions. Goals use a judge with round caps and user-message interruption. Transcripts export locally as Markdown/HTML with tool details; archive keeps backend history.
Retrieved: 2026-11-07

Source: https://github.com/OpenHands/OpenHands/blob/b811a4c33979ca4b5d241730b995b56362f4ab34/docs/ACP_AGENTS.md
Locator: Authentication; Per-conversation isolation; Model switching
Finding: Local ACP agents reuse provider-native logins; clean containers/cloud need credentials. Pi/OpenCode are offered only on local/self-hosted backends. Same-provider conversations sharing container HOME can race on auth/config/lock files; Canvas cannot yet send SDK per-conversation data-dir isolation. Agent choice is per backend and running conversations retain their original agent.
Retrieved: 2026-11-07

Source: https://github.com/OpenHands/OpenHands/blob/e14b76cca01a71597cee242c202eed730e745ce3/src/routes/app-settings.tsx#L217-L231
Locator: src/routes/app-settings.tsx lines 112-136 and 217-231 (enable-analytics-switch), release v1.38.0 commit e14b76c
Finding: App settings render a 'send anonymous usage data' switch that defaults on and saves user_consents_to_analytics and the telemetry consent as granted or denied. On a Cloud backend the switch is disabled and stays on.
Retrieved: 2026-11-10

Source: https://github.com/OpenHands/OpenHands/blob/e14b76cca01a71597cee242c202eed730e745ce3/src/components/features/analytics/telemetry-consent-banner.tsx#L60-L112
Locator: src/components/features/analytics/telemetry-consent-banner.tsx lines 60-112, release v1.38.0 commit e14b76c
Finding: A telemetry consent modal asks the user whether to send anonymous usage data and saves the choice as granted or denied.
Retrieved: 2026-11-10

## openhands:OWN-04

Finding: The current public self-hosted offer is free Canvas, Teams at $500/month for ten users plus $50/month additional seats with a 5000-session monthly cap, and custom Enterprise. No one-time licence is listed. The optional Cloud backend is not priced in this table; its documented OpenHands LLM credits buy model access, so product-wide metered ADE and compute pricing remain unestablished.

Verification: sourced
Checked: 2026-11-07
Scheduled recheck: 2026-11-21
Plan: Agent Canvas unless answer explicitly scopes paid Teams/Enterprise
Surface: Agent Canvas; OpenHands Agent Server backends
Version: First-party documentation/source checked 2026-11-07; exact pinned source revision is recorded in each citation. No installed-product fixture.

Only the named atomic answers are established by this source pass. Remaining atomic answers retain their documented gaps.

Source: https://www.openhands.dev/pricing
Locator: Plan cards; Comparison table
Finding: Canvas is free, MIT, single-user/shared-identity and self-hosted. Teams is $500 monthly for up to ten users with $50 monthly per additional seat, 5000 sessions per month, 30-day trial, RBAC and SSO/SAML. Enterprise has custom pricing. LLM keys are user-provided.
Retrieved: 2026-11-07

Source: https://www.openhands.dev/pricing
Locator: Agent Canvas, Teams and Enterprise; Comparison table
Finding: The current public price table lists free self-hosted Canvas, self-hosted Teams at $500/month for ten users plus $50/month additional seats, capped at 5000 sessions monthly, and custom Enterprise. It lists no one-time licence. The separately linked Cloud backend is not priced in this table.
Retrieved: 2026-11-13

Source: https://docs.openhands.dev/openhands/usage/settings/api-keys-settings
Locator: OpenHands LLM Key
Finding: Cloud users must buy at least $10 in credits before generating an OpenHands LLM Key. That key can be used in the CLI, self-hosted OpenHands or other coding agents. These are model-provider credits, not proof of a metered ADE or sandbox price.
Retrieved: 2026-11-13

## openhands:OWN-06

Finding: Teams and Enterprise plans carry paid identity governance; support ranges from community support for Agent Canvas to best-effort email on Teams and SLA-based response on Enterprise.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-11-21
Plan: Agent Canvas unless answer explicitly scopes paid Teams/Enterprise
Surface: Agent Canvas; OpenHands Cloud; OpenHands Agent Server backends
Version: First-party documentation/source checked 2026-11-07; exact pinned source revision is recorded in each citation. No installed-product fixture.

Only the named atomic facts are established by documentation. Remaining rubric scenarios and unknown answers were not demonstrated.

Source: https://www.openhands.dev/pricing
Locator: Plan cards; Comparison table
Finding: Canvas is free, MIT, single-user/shared-identity and self-hosted. Teams is $500 monthly for up to ten users with $50 monthly per additional seat, 5000 sessions per month, 30-day trial, RBAC and SSO/SAML. Enterprise has custom pricing. LLM keys are user-provided.
Retrieved: 2026-11-07

## openhands:OWN-08

Finding: The vendor advertises free local and self-managed Canvas use.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-11-21
Plan: Not specified
Surface: Agent Canvas; OpenHands Agent Server backends
Version: First-party sources checked 2026-11-07; pinned implementations retain their cited revision.

The remaining atomic questions and the rubric scenario are unverified. Documentation is not a successful product test.

Source: https://www.openhands.dev/product/canvas
Locator: ACP; Where your agents run; Get started
Finding: Vendor describes desktop installs, ACP agents with existing subscriptions, extensions and local, self-hosted or managed backends.
Retrieved: 2026-11-07

## openhands:OWN-09

Finding: Pricing page publishes monthly Teams rate and session caps, with Enterprise by quote.

Verification: agent-cross-checked
Checked: 2026-11-07
Scheduled recheck: 2026-11-21
Plan: Agent Canvas unless answer explicitly scopes paid Teams/Enterprise
Surface: OpenHands Agent Server backends
Version: First-party documentation/source checked 2026-11-07; exact pinned source revision is recorded in each citation. No installed-product fixture.

Only the named atomic facts are established by documentation. Remaining rubric scenarios and unknown answers were not demonstrated.

Source: https://www.openhands.dev/pricing
Locator: Plan cards; Comparison table
Finding: Canvas is free, MIT, single-user/shared-identity and self-hosted. Teams is $500 monthly for up to ten users with $50 monthly per additional seat, 5000 sessions per month, 30-day trial, RBAC and SSO/SAML. Enterprise has custom pricing. LLM keys are user-provided.
Retrieved: 2026-11-07