OpenHands by All Hands AI

# Fix list: OpenHands

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/openhands, the October 2026 research run, assessed 1 October 2026. Grade BB, 70.9 out of 100.

This is everything the published grade says the listing lacks, the biggest possible gain to the total first. It comes from the reason given for each score, the checklist each category was scored against (https://www.anchorterminal.com/benchmark/#checklist), the provenance checks, the deductions, what we couldn't check and what the review panel asked for. A fix counts at the next check, once it's public.

For a coding agent working on OpenHands: work through the items below in the product, its docs and its public pages. Each category gives the reason for its score, with the points each checklist item earned, and the checklist itself, so the gap is the items that earned less than their points. Change the product, not the wording, and keep a note of what you changed and where it's published.

## 1. Security & auth, 66 out of 100, up to 6 more on the total

Why it scored 66: Harness reading of the framework checklist, used for all five harnesses in this batch. 30 for what leaves the machine by default, 20 for approvals and sandboxing, 15 for prompt-injection posture, 15 for audit and 20 for the security programme. Agent Canvas sends a `canvas_install` event (platform, user agent, referrer, origin) to PostHog through z.openhands.dev on first use, before asking, then asks with the opt-in box already ticked. The Agent Server's exporter follows that consent, and `AGENT_CANVAS_DISABLE_TELEMETRY=1` or `DO_NOT_TRACK=1` stops all of it (15). Confirmation policies (always, never, at or above a risk level) and a Docker sandbox per conversation exist, but Canvas starts with confirmation off and the npm install runs on the host with full filesystem access. We found no network egress controls. Local listeners bind to 127.0.0.1 with an injected session key (12). LLM, Invariant and GraySwan risk analysers and a policy-rail module, documented, all advisory (11). Conversation event logs, completion logs and OpenTelemetry tracing through Laminar (13). A valid security.txt with a responsible-disclosure policy, a Vanta-hosted trust page linked from pricing, and the git diff injection published as a GitHub advisory with a CVE. No bug bounty found (15).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-security):

- 0 to 30, the credential model. 30 for OAuth 2.1 with scopes, or scoped and revocable keys with rotation. 20 for plain revocable API keys. 10 for one all-powerful key. 10 off when a secret can travel in a URL query string as a documented option.
- 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions.
- 0 to 15, prompt-injection posture where the tool returns untrusted content (documented mitigations or guidance). A tool that returns no untrusted content gets 10.
- 0 to 15, audit logs or per-call visibility for the operator.
- 0 to 20, a security programme. security.txt or a disclosure policy, a bug bounty, SOC 2 or ISO 27001, advisories handled in public.

Models are read for retention, whether API data trains models (and whether that's off by default), zero-retention options and certifications. Frameworks for telemetry defaults, approval hooks, guardrails and sandboxing.

## 2. Payments & pricing, 60 out of 100, up to 5 more on the total

Why it scored 60: No payment protocol (0). Scored on OpenHands Cloud and its LLM provider. The docs publish per-token prices for the OpenHands LLM provider at provider rates with no markup, Cloud Individual is free, and Enterprise is priced by sales (20). The MIT stack is free and the pricing page mentions no card for the free plan (20). It runs any model through LiteLLM, local ones included, so an agent can install and run it with no signup (20). The price table lists 2025 models, so it may be stale.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-payments):

The published rubric, also on the [x402 page](https://www.anchorterminal.com/x402/).

- 40, a machine payment protocol (x402, MPP or L402) on the tool's own endpoints. 10 to 30 when it covers only some endpoints or only goes through a third party, and the note says which.
- 20, per-call or per-unit pricing published without a login. 10 for public plan-only pricing, 0 for "contact sales" or prices behind a login.
- 20, a free tier or trial that doesn't need a card.
- 20, autonomous onboarding, meaning an agent can get access without a person signing up in a browser (keyless use, x402, a programmatic key API).

Payment platforms and agent wallets rarely charge for their own API over a machine protocol, so the first line has steps for them, and the highest one that applies counts. 40 when x402, MPP or L402 runs on all their own endpoints, 30 when it runs on part of their own API, 25 when their merchants can accept one, 20 for running a facilitator, 15 for paying as a buyer, and 0 when the only protocol is their own. Merchant acceptance sits above a facilitator because the platform's own customers can charge agents through it, while a facilitator settles for sellers who wire up the protocol themselves. The counter-argument (a facilitator does more for the protocol as a whole) has a point. Each note says which step applied.

Open-source software you run yourself is scored on its hosted or paid option if it has one. A free, self-hosted package with nothing to buy gets 20, 20 and 20 for the last three lines, and 0 to 40 for the first only if it ships a payment protocol.

## 3. Agent ergonomics, 78 out of 100, up to 3.6 more on the total

Why it scored 78: Harness reading of the framework line, scored on what an agent or pipeline driving it has to supply. The default agent carries three tools (terminal, file editor, task tracker), and `filter_tools_regex` trims MCP tools before they reach the model (23). Stuck detection is on by default and a condenser compacts long histories, but Canvas leaves `max_iterations` unset and we found no cost cap (15). Headless runs go through the SDK, the Agent Server REST API or the Cloud API with typed events, and the old CLI's `--headless` mode is unmaintained (14). Conversations persist and resume, and a Docker conversation keeps its workspace and history when the container is replaced (16). Python SDK and TypeScript client, but the quickstart needs Node 24 and uv, and the defaults skip both confirmation and the sandbox (10).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-ergonomics):

- 0 to 25, context cost. For MCP, the number and size of the tool definitions (25 for ten or fewer compact tools, 15 for 11 to 30, 5 for more than 30, plus up to 10 back for toolsets, dynamic loading or read-only subsets). For APIs, whether responses can be sized (field selection, limits, summaries).
- 20, pagination, filtering and output-size controls.
- 20, actionable, documented error responses, codes and messages an agent can recover from.
- 20, idempotency or safe retries, and for MCP the `readOnlyHint` and `destructiveHint` annotations.
- 15, sensible defaults, few required parameters, and official SDKs in at least two languages.

Models are read for tool use, structured output, prompt caching, context length, batch and SDKs. Frameworks for how much code and how many defaults a tool-calling agent with MCP needs.

## 4. Reliability, 83 out of 100, up to 3.4 more on the total

Why it scored 83: Read as a local package. Official npm package @openhands/agent-canvas (Node 24 or later, plus uv), a Docker image on GHCR and openhands-sdk on PyPI (Python 3.12 or later) (20). The CI workflow passed on every run on main we loaded (25). 456 open issues and 395 open pull requests, with recent reports labelled by priority and ready-for-dev within a day or two, and a high-priority Cloud resume bug (#17664) open (18). Dated release notes for every Canvas version and a written SDK rule of at least five minor releases before a public API or REST contract is removed, but the Canvas CHANGELOG.md stops at 1.0.0-alpha.2 and the product was reshaped twice in a year, with the local GUI deprecated and the CLI left unmaintained (12). Canvas is 1.24 with a beta badge on its README, and the SDK is 1.50 (8). Judgement call on the last line, half credit for a 1.x release its maker still calls beta.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-reliability):

Hosted APIs, MCP servers, models and platforms.

- 20, a public status page with component history (Statuspage, Instatus, BetterStack or the vendor's own).
- 0 to 30, the incident record for the last 90 days on that page. 30 for a clean record or trivial incidents only, 20 for minor incidents only, 10 for one major outage (an hour or more of a core API down, or errors across the board), 0 for several. 5 when there's no history we could read, and the note says so.
- 15, rate limits documented with numbers.
- 15, documented 429 or overload handling (Retry-After, backoff guidance), and idempotency keys or safe-retry guidance where writes are involved.
- 10, an SLA published for any paid tier.
- 10, the surface agents use is generally available, not beta or preview.

Local packages, SDKs, frameworks and stdio MCP servers.

- 20, installs from an official package with supported runtimes stated.
- 25, a public CI and test suite, passing on the default branch.
- 0 to 25, open crash or regression issues relative to activity (25 for few and handled, 0 for many, old and unanswered).
- 15, semver discipline and breaking changes called out in a changelog.
- 15, version 1.0 or later, or declared stable.

Protocols are read from their reference implementations, the public facilitators or servers, spec stability and test vectors.

## 5. Transparency & trust, 69 out of 100, up to 2.7 more on the total

Made of editorial 67, provenance 71.

Why it scored 69: MIT across Canvas, SDK, tools and Agent Server (30). The privacy policy (effective 3 September 2025) doesn't separate the open-source stack from the Cloud service, keeps data as long as reasonably necessary, allows training on content from the Services and names PostHog and Google Analytics only in passing, with no subprocessor list (12). A written SDK deprecation runway of five minor releases, deprecated projects grouped in the docs, and the CLI's unmaintained notice in its README without a date on the page (15). The consent prompt calls the data anonymous and the Docker page documents `AGENT_CANVAS_DISABLE_TELEMETRY`, but no user-facing page mentions the install event sent before consent (10).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-transparency):

- 0 to 30, source availability and licence clarity. 30 for open source under an OSI licence, 15 for closed with clear terms, 0 for unclear terms.
- 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors).
- 0 to 20, a deprecation policy or notices with dates.
- 0 to 20, telemetry disclosed with an opt-out (local software), or subprocessors and data locations disclosed (hosted).

The other half of Transparency and trust is the provenance score, computed from checked facts (below). The category score is the mean of the two.

Provenance checks not met in full (half of this category, computed from checked facts):

- Domain age: openhands.dev, no registry record we could read (0 of 15)
- Status page: not found (0 of 10)

## 6. Schema & documentation, 87 out of 100, up to 2.1 more on the total

Why it scored 87: The Agent Server REST API has an OpenAPI 3.1 spec in the docs repository, the Cloud API has another, SDK models are Pydantic-typed and there's a TypeScript client (25). llms.txt at docs.openhands.dev, which leaves out the V0 pages on purpose (10). The docs have a when-to-use page and examples of good and bad task instructions, but the navigation still carries the deprecated local GUI, the unmaintained CLI and V0 reference beside Agent Canvas, which is a lot for a model to sort through (12). Confirmation policies, security analysers and MCP configs are discriminated Pydantic unions with validators (13). SDK examples are checked by a CI workflow and there are troubleshooting pages, but we found no error reference for the Agent Server (12). Dated release notes for each Canvas version in the docs, and GitHub releases per tag (15).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-schema):

APIs and MCP servers.

- 25, a machine-readable contract (a public OpenAPI file or similar; for MCP, typed JSON Schema inputs on every tool).
- 10, llms.txt or Markdown docs served for agents.
- 0 to 20, descriptions that say what a tool is for, when to use it and when not to, read from the tool definitions in the source or the API reference.
- 0 to 15, typed inputs with enums, constraints and required fields, and no free-form JSON blobs.
- 0 to 15, examples and documented error responses.
- 15, versioning and a public changelog.

Models are read from the API reference, the OpenAPI file, llms.txt, the structured-output and tool-use docs and the model cards. Frameworks from docs a model can follow, typed interfaces, examples and the API reference.

## 7. Maintenance & community, 85 out of 100, up to 1.3 more on the total

Why it scored 85: SDK 1.50.1 on 2026-09-30 and Agent Canvas 1.24.0 on 2026-09-25 (30). 21 Canvas releases since 2026-07-24 and 13 SDK tags in September (20). Issues are labelled by priority within a day or two, but 456 open issues and 395 open pull requests, and we couldn't see reply times (16). npm, PyPI and GHCR packages are current, but the `openhands` CLI on PyPI has been unmaintained since 2026-08-11 and still appears in the docs and on the pricing page (10). CI passes on main, and Dependabot, a security scan and an API-breakage check run on the SDK (9).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-maintenance):

- 0 to 30, time since the last release, or the last published model or API change for a closed service. 30 within 30 days, 20 within 90, 10 within 180, 0 older.
- 20, at least three releases or dated changelog entries in the last 90 days.
- 0 to 25, responsiveness. Issues and pull requests answered on GitHub (the open issues and how recent the replies are). For closed services, a public changelog and a support or community channel that answers, 0 to 15.
- 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models).
- 10, package health, current dependencies and CI.

Models are read for deprecation notice periods and model churn rather than release counts.

## Deductions

Each comes off the total. A fixed and documented problem counts for less at the next check.

- Current behaviour, checked 2026-10-02. Agent Canvas sends a `canvas_install` event with platform, user agent, referrer and origin to PostHog through OpenHands' proxy at z.openhands.dev on first use, before the consent prompt, opting the client in for that one event. The source comment says the proxy is there to get past ad blockers, the consent prompt's box is ticked by default, and no user-facing doc mentions the early event. Undisclosed telemetry, -3. https://github.com/OpenHands/OpenHands/blob/main/src/services/telemetry.ts
- 2026-03-23. GHSA-7h8w-hj9j-8rjw (CVE-2026-33718, 7.6 in the advisory, 9.9 at NVD), command injection through the `path` parameter of the git diff endpoint let an authenticated user run commands in the agent sandbox. Fixed in 1.5.0 and published, a little over six months old, -1. https://github.com/OpenHands/OpenHands/security/advisories/GHSA-7h8w-hj9j-8rjw
- 2026-08-06. CVE-2026-19022 (6.3), command injection in `initialize_repo` in the pull request resolver of OpenHands 0.62.0 and earlier, the V0 line the V1 rewrite replaced. No GitHub advisory found, -1. https://nvd.nist.gov/vuln/detail/CVE-2026-19022

## What we couldn't check

What we couldn't read counted as absent. Publishing it on a page a plain HTTP fetch can read (not only in a browser) lets the next check count it.

- We found no terms of service URL for OpenHands Cloud on the pricing or privacy pages
- Unchecked: a status page for OpenHands Cloud and the domain's registration date
- Whether the Cloud Individual plan needs a card isn't stated on the pricing page
- The Agent Server OpenAPI file in the docs repository says 0.1.0 while the server is 1.50.1, so it may lag the code
- The issue pages we loaded didn't show reply times
- The OpenHands LLM price table lists 2025 models and may be out of date

## Weaknesses

- Confirmation mode is off by default in Agent Canvas, and the npm install gives the agent the host's whole filesystem
- One anonymous install event goes to PostHog before the consent prompt, whose opt-in box is pre-ticked
- The terminal CLI has been unmaintained since 11 August 2026 and the Docker-based local GUI is deprecated, yet both fill much of the docs
- Agent Canvas carries a beta badge, and its CHANGELOG.md stops at 1.0.0-alpha.2
- The privacy policy (3 September 2025) allows training on Cloud content and gives no retention period

## What costs an agent a turn today

The notes we give agents before they call it. Each one is a workaround an agent shouldn't need.

- Set `AGENT_CANVAS_DISABLE_TELEMETRY=1` and `DO_NOT_TRACK=1` before the first start
- Start Canvas with `OH_CONVERSATION_RUNTIME=docker` or use the Docker image. The npm install runs the agent on the host
- Turn on confirmation mode with a risk threshold. Canvas starts with it off
- Use the SDK or the Agent Server API for headless runs. The `openhands --headless` CLI is no longer maintained
- Set `filter_tools_regex` on the agent to keep unneeded MCP tool definitions out of the context

## What the review panel asked for

- dated end-of-life notices
- same policy for Canvas
- no events before consent
- Docker runtime by default

## When it's done

Send what changed and where it's published as a dispute (https://www.anchorterminal.com/builders/#disputes, or `POST https://www.anchorterminal.com/api/v1/contact` with `"kind": "dispute"`). Disputes are answered in public, and the listing is checked again by the same checklist. Paying for an audit or a listing claim changes nothing here.