{
  "fixes": {
    "slug": "muse-code",
    "name": "Muse Code",
    "listing": "https://www.anchorterminal.com/tools/muse-code",
    "markdown": "# Fix list: Muse Code\n\nFrom Anchor Terminal's listing at https://www.anchorterminal.com/tools/muse-code, the October 2026 research run, assessed 12 October 2026. Grade C, 61.3 out of 100.\n\nThis 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.\n\nFor a coding agent working on Muse Code: 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.\n\n## 1. Reliability, 45 out of 100, up to 11 more on the total\n\nWhy it scored 45: Local-package reading. Official installers for macOS, Linux and Windows that fetch a native binary from a release channel, with a SHA-256 check on the launcher and per-platform checksums in the release manifest (20). No public CI or test suite for the program, since the source isn't published. The SDK repository's workflows only publish to npm and PyPI (0). No public issue tracker for the CLI was found, so there's nothing to count either way. The SDK repository takes issues for the SDK and protocol only (5, a departure from the line, which assumes a tracker). Release notes per version under New, Improvements, Performance and Fixes, with security fixes labelled and deprecated aliases named inline, but no dates and no breaking-change heading (5). 1.4.6 (15). Same reading as Claude Code, which is also closed source.\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-reliability):\n\nHosted APIs, MCP servers, models and platforms.\n\n- 20, a public status page with component history (Statuspage, Instatus, BetterStack or the vendor's own).\n- 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.\n- 15, rate limits documented with numbers.\n- 15, documented 429 or overload handling (Retry-After, backoff guidance), and idempotency keys or safe-retry guidance where writes are involved.\n- 10, an SLA published for any paid tier.\n- 10, the surface agents use is generally available, not beta or preview.\n\nLocal packages, SDKs, frameworks and stdio MCP servers.\n\n- 20, installs from an official package with supported runtimes stated.\n- 25, a public CI and test suite, passing on the default branch.\n- 0 to 25, open crash or regression issues relative to activity (25 for few and handled, 0 for many, old and unanswered).\n- 15, semver discipline and breaking changes called out in a changelog.\n- 15, version 1.0 or later, or declared stable.\n\nProtocols are read from their reference implementations, the public facilitators or servers, spec stability and test vectors.\n\n## 2. Payments \u0026 pricing, 20 out of 100, up to 10 more on the total\n\nWhy it scored 20: Harness reading of the published rubric. No payment protocol (0). Plan prices ($5, $15, $50 and $200 a month) and Model API token prices are public without a login, with plan usage given as multiples of the Everyday plan's 10 to 50 prompts every 5 hours (20). No free plan or trial found, and pay-as-you-go needs a card on the Model API account (0). A person signs in through a browser or creates a key in the dashboard, and no local-model mode is documented (0).\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-payments):\n\nThe published rubric, also on the [x402 page](https://www.anchorterminal.com/x402/).\n\n- 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.\n- 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.\n- 20, a free tier or trial that doesn't need a card.\n- 20, autonomous onboarding, meaning an agent can get access without a person signing up in a browser (keyless use, x402, a programmatic key API).\n\nPayment 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.\n\nOpen-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.\n\n## 3. Security \u0026 auth, 67 out of 100, up to 5.8 more on the total\n\nWhy it scored 67: Framework reading (telemetry defaults, approvals, guardrails, sandboxing), five lines. A browser session or a Model API key, cleared by `muse logout`. Settings include telemetry options, and the release notes mention opt-outs for feedback and diagnostic upload and for remote feature settings, but no page says what is sent or what is on by default. Subscriptions default to Contributor mode, so Meta may train on code unless the user turns it off, with a notice before the first request (10 of 30). Ask me, Auto-review and Unrestricted profiles, stage-by-stage review of compound shell commands, and an OS sandbox on by default that keeps `.git`, `.muse` and `.agents` read-only and refuses to run when it can't confirm containment. The default is Auto-review, an automated judge rather than a person (17 of 20). Network is proxy-only by default with per-destination review, and the docs warn that MCP tools are not sandboxed. The 1.4.6 notes fix `find -exec rm` running without approval and say a `find -exec sh -c` script still can (11 of 15). An append-only event log with replay and a record of who authorised each action, plus `muse export` (14 of 15). Meta's bug bounty, named in the SDK's SECURITY.md, a valid security.txt on www.facebook.com, and security fixes labelled in the release notes, but no published advisories for Muse Code (15 of 20). Judgement call, written by an agent on Claude, on the checklist the other harnesses got.\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-security):\n\n- 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.\n- 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions.\n- 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.\n- 0 to 15, audit logs or per-call visibility for the operator.\n- 0 to 20, a security programme. security.txt or a disclosure policy, a bug bounty, SOC 2 or ISO 27001, advisories handled in public.\n\nModels 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.\n\n## 4. Transparency \u0026 trust, 57 out of 100, up to 3.8 more on the total\n\nMade of editorial 46, provenance 68.\n\nWhy it scored 57: Closed-source binary under the Meta Model API Terms of Service, with MIT SDKs (15). The subscription docs, the Contributor mode help page, the geographic use policy and the terms agree that a coding-harness subscription is in Contributor mode by default and that Standard mode content isn't used for training. Local sessions are kept forever unless `session_retention_days` is set, and no server-side retention period is stated (18 of 30). No deprecation policy. Deprecated aliases are named in undated release notes (5 of 20). Telemetry options exist in settings, with opt-outs mentioned in the release notes, but the data sent isn't documented (8 of 20).\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-transparency):\n\n- 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.\n- 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors).\n- 0 to 20, a deprecation policy or notices with dates.\n- 0 to 20, telemetry disclosed with an opt-out (local software), or subprocessors and data locations disclosed (hosted).\n\nThe other half of Transparency and trust is the provenance score, computed from checked facts (below). The category score is the mean of the two.\n\nProvenance checks not met in full (half of this category, computed from checked facts):\n\n- Domain age: meta.ai, registered 2017-12-16 (8 years) (11 of 15)\n- Privacy policy: published, but our reader couldn't read it (7 of 10)\n- Status page: not found (0 of 10)\n- security.txt: not found (0 of 10)\n\n## 5. Agent ergonomics, 79 out of 100, up to 3.4 more on the total\n\nWhy it scored 79: Framework reading, adapted to a harness driven by a pipeline. `muse exec` runs one prompt with the MCP servers in settings, and named toolsets select MCP tools by name or prefix. We found no deferred tool loading in Muse Code's own docs (18 of 25). `--max-model-steps` stops a stuck run, and `muse export` writes a session record, but no budget cap or output schema was found (15 of 20). Exit codes 0, 1, 2, 130 and 143 are documented, and retry notices carry the provider error and HTTP status (16 of 20). `--session-id` continues a headless run, `muse export` and the event log support replay, and rewind branches without changing the original (18 of 20). TypeScript and Python SDKs. The defaults (Auto-review, proxy-only network) can stop for a review nobody answers in CI, so a pipeline has to choose a posture (12 of 15).\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-ergonomics):\n\n- 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).\n- 20, pagination, filtering and output-size controls.\n- 20, actionable, documented error responses, codes and messages an agent can recover from.\n- 20, idempotency or safe retries, and for MCP the `readOnlyHint` and `destructiveHint` annotations.\n- 15, sensible defaults, few required parameters, and official SDKs in at least two languages.\n\nModels 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.\n\n## 6. Schema \u0026 documentation, 84 out of 100, up to 2.6 more on the total\n\nWhy it scored 84: Framework reading. `muse schema` prints the JSON schema of a versioned session protocol, the protocol declarations and a stable manifest are published in the SDK repository, and `muse exec --json` emits JSONL events. No JSON Schema for `settings.json` was found (22 of 25). llms.txt at dev.meta.ai/llms.txt with a Markdown copy of each page (10). The permissions page says what each profile and flag is for and when to use it, such as Unrestricted only in an isolated environment, and the extending page says when to use subagents and when not to (16 of 20). Profiles, approval modes, sandbox network modes and reasoning-effort levels are enumerated, with MCP servers typed by transport (12 of 15). Examples on every page, exit codes for `muse exec`, and named settings errors, though no full list of run errors (12 of 15). A changelog for every version, with no dates (12 of 15).\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-schema):\n\nAPIs and MCP servers.\n\n- 25, a machine-readable contract (a public OpenAPI file or similar; for MCP, typed JSON Schema inputs on every tool).\n- 10, llms.txt or Markdown docs served for agents.\n- 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.\n- 0 to 15, typed inputs with enums, constraints and required fields, and no free-form JSON blobs.\n- 0 to 15, examples and documented error responses.\n- 15, versioning and a public changelog.\n\nModels 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.\n\n## 7. Maintenance \u0026 community, 75 out of 100, up to 2.2 more on the total\n\nWhy it scored 75: 1.4.6, served by the muse-stable channel on 12 October 2026 and published as SDK 1.4.6 on 11 October (30). The SDK, released in step, published 1.3.0 on 18 September, 1.4.2 on 30 September, 1.4.4 on 8 October and 1.4.6 on 11 October (20). No public tracker for the CLI. The SDK repository takes issues with a written triage guide, and the help centre takes support tickets, neither tested (8 of 25). MIT TypeScript and Python SDKs at 1.4.6, labelled a developer preview (12 of 15). A checksummed launcher and release manifest, but no public CI for the program (5 of 10).\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-maintenance):\n\n- 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.\n- 20, at least three releases or dated changelog entries in the last 90 days.\n- 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.\n- 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models).\n- 10, package health, current dependencies and CI.\n\nModels are read for deprecation notice periods and model churn rather than release counts.\n\n## What we couldn't check\n\nWhat 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.\n\n- unchecked: the CLI's own release dates. The changelog has no dates, so recency rests on the release channel serving 1.4.6 on 12 October and the SDK's npm publish of 1.4.6 on 11 October\n- unchecked: what Muse Code's telemetry sends and what is on by default. The settings page lists telemetry options without detail\n- unchecked: the open issues and maintainer replies on the SDK repository. GitHub's API was not available to our reader, and the CLI has no public tracker that we found\n- The 1.4.6 release notes describe an approval gap that was fixed (`find -exec rm` ran without asking) and one that remains (`find -exec sh -c`). The sandbox still limits writes to the workspace. Recorded as a weakness with no deduction, a judgement call\n- The product page says Muse Code is available for macOS and Windows, while the docs and the 31 August post give macOS, Linux and Windows. Graded on the docs\n- No authenticated session was run\n\n## Weaknesses\n\n- Closed source, with no public CI or issue tracker for the CLI found and release notes that carry no dates\n- Subscriptions default to Contributor mode, which lets Meta train on prompts, code and responses until it's turned off\n- No free route. Plans run from $5 to $200 a month, and pay-as-you-go billing needs a card\n- New sessions start in Auto-review, where an automated judge clears eligible approvals rather than a person\n- MCP tools run outside the sandbox, and settings list telemetry options without documenting what is sent or the defaults\n\n## What costs an agent a turn today\n\nThe notes we give agents before they call it. Each one is a workaround an agent shouldn't need.\n\n- Set `META_API_KEY` for headless runs. Browser sign-in isn't available in CI or for Meta Managed Accounts\n- Gate on your own test command, not the exit code. `muse exec` returns 0 when the turn completes even if tests still fail\n- Pass `--max-model-steps` on unattended runs so a stuck task can't loop\n- Prefer `--disable-approval` to `--yolo` in CI so the sandbox stays on. On Linux the runner needs a working bubblewrap and a non-musl build\n- Turn Contributor mode off in `/settings` before a subscription session touches private code\n\n## When it's done\n\nSend 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.\n",
    "grade": "C",
    "score": 61.3,
    "assessed": "2026-10-12",
    "run": "October 2026 research run",
    "categories": [
      {
        "key": "reliability",
        "name": "Reliability",
        "score": 45,
        "maxGain": 11,
        "reason": "Local-package reading. Official installers for macOS, Linux and Windows that fetch a native binary from a release channel, with a SHA-256 check on the launcher and per-platform checksums in the release manifest (20). No public CI or test suite for the program, since the source isn't published. The SDK repository's workflows only publish to npm and PyPI (0). No public issue tracker for the CLI was found, so there's nothing to count either way. The SDK repository takes issues for the SDK and protocol only (5, a departure from the line, which assumes a tracker). Release notes per version under New, Improvements, Performance and Fixes, with security fixes labelled and deprecated aliases named inline, but no dates and no breaking-change heading (5). 1.4.6 (15). Same reading as Claude Code, which is also closed source.",
        "checklist": [
          "Hosted APIs, MCP servers, models and platforms.",
          "- 20, a public status page with component history (Statuspage, Instatus, BetterStack or the vendor's own).\n- 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.\n- 15, rate limits documented with numbers.\n- 15, documented 429 or overload handling (Retry-After, backoff guidance), and idempotency keys or safe-retry guidance where writes are involved.\n- 10, an SLA published for any paid tier.\n- 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.\n- 25, a public CI and test suite, passing on the default branch.\n- 0 to 25, open crash or regression issues relative to activity (25 for few and handled, 0 for many, old and unanswered).\n- 15, semver discipline and breaking changes called out in a changelog.\n- 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."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-reliability"
      },
      {
        "key": "payments",
        "name": "Payments \u0026 pricing",
        "score": 20,
        "maxGain": 10,
        "reason": "Harness reading of the published rubric. No payment protocol (0). Plan prices ($5, $15, $50 and $200 a month) and Model API token prices are public without a login, with plan usage given as multiples of the Everyday plan's 10 to 50 prompts every 5 hours (20). No free plan or trial found, and pay-as-you-go needs a card on the Model API account (0). A person signs in through a browser or creates a key in the dashboard, and no local-model mode is documented (0).",
        "checklist": [
          "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.\n- 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.\n- 20, a free tier or trial that doesn't need a card.\n- 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."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-payments"
      },
      {
        "key": "security",
        "name": "Security \u0026 auth",
        "score": 67,
        "maxGain": 5.8,
        "reason": "Framework reading (telemetry defaults, approvals, guardrails, sandboxing), five lines. A browser session or a Model API key, cleared by `muse logout`. Settings include telemetry options, and the release notes mention opt-outs for feedback and diagnostic upload and for remote feature settings, but no page says what is sent or what is on by default. Subscriptions default to Contributor mode, so Meta may train on code unless the user turns it off, with a notice before the first request (10 of 30). Ask me, Auto-review and Unrestricted profiles, stage-by-stage review of compound shell commands, and an OS sandbox on by default that keeps `.git`, `.muse` and `.agents` read-only and refuses to run when it can't confirm containment. The default is Auto-review, an automated judge rather than a person (17 of 20). Network is proxy-only by default with per-destination review, and the docs warn that MCP tools are not sandboxed. The 1.4.6 notes fix `find -exec rm` running without approval and say a `find -exec sh -c` script still can (11 of 15). An append-only event log with replay and a record of who authorised each action, plus `muse export` (14 of 15). Meta's bug bounty, named in the SDK's SECURITY.md, a valid security.txt on www.facebook.com, and security fixes labelled in the release notes, but no published advisories for Muse Code (15 of 20). Judgement call, written by an agent on Claude, on the checklist the other harnesses got.",
        "checklist": [
          "- 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.\n- 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions.\n- 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.\n- 0 to 15, audit logs or per-call visibility for the operator.\n- 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."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-security"
      },
      {
        "key": "transparency",
        "name": "Transparency \u0026 trust",
        "score": 57,
        "maxGain": 3.8,
        "reason": "Closed-source binary under the Meta Model API Terms of Service, with MIT SDKs (15). The subscription docs, the Contributor mode help page, the geographic use policy and the terms agree that a coding-harness subscription is in Contributor mode by default and that Standard mode content isn't used for training. Local sessions are kept forever unless `session_retention_days` is set, and no server-side retention period is stated (18 of 30). No deprecation policy. Deprecated aliases are named in undated release notes (5 of 20). Telemetry options exist in settings, with opt-outs mentioned in the release notes, but the data sent isn't documented (8 of 20).",
        "blend": "editorial 46, provenance 68",
        "checklist": [
          "- 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.\n- 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors).\n- 0 to 20, a deprecation policy or notices with dates.\n- 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."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-transparency"
      },
      {
        "key": "ergonomics",
        "name": "Agent ergonomics",
        "score": 79,
        "maxGain": 3.4,
        "reason": "Framework reading, adapted to a harness driven by a pipeline. `muse exec` runs one prompt with the MCP servers in settings, and named toolsets select MCP tools by name or prefix. We found no deferred tool loading in Muse Code's own docs (18 of 25). `--max-model-steps` stops a stuck run, and `muse export` writes a session record, but no budget cap or output schema was found (15 of 20). Exit codes 0, 1, 2, 130 and 143 are documented, and retry notices carry the provider error and HTTP status (16 of 20). `--session-id` continues a headless run, `muse export` and the event log support replay, and rewind branches without changing the original (18 of 20). TypeScript and Python SDKs. The defaults (Auto-review, proxy-only network) can stop for a review nobody answers in CI, so a pipeline has to choose a posture (12 of 15).",
        "checklist": [
          "- 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).\n- 20, pagination, filtering and output-size controls.\n- 20, actionable, documented error responses, codes and messages an agent can recover from.\n- 20, idempotency or safe retries, and for MCP the `readOnlyHint` and `destructiveHint` annotations.\n- 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."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-ergonomics"
      },
      {
        "key": "schema",
        "name": "Schema \u0026 documentation",
        "score": 84,
        "maxGain": 2.6,
        "reason": "Framework reading. `muse schema` prints the JSON schema of a versioned session protocol, the protocol declarations and a stable manifest are published in the SDK repository, and `muse exec --json` emits JSONL events. No JSON Schema for `settings.json` was found (22 of 25). llms.txt at dev.meta.ai/llms.txt with a Markdown copy of each page (10). The permissions page says what each profile and flag is for and when to use it, such as Unrestricted only in an isolated environment, and the extending page says when to use subagents and when not to (16 of 20). Profiles, approval modes, sandbox network modes and reasoning-effort levels are enumerated, with MCP servers typed by transport (12 of 15). Examples on every page, exit codes for `muse exec`, and named settings errors, though no full list of run errors (12 of 15). A changelog for every version, with no dates (12 of 15).",
        "checklist": [
          "APIs and MCP servers.",
          "- 25, a machine-readable contract (a public OpenAPI file or similar; for MCP, typed JSON Schema inputs on every tool).\n- 10, llms.txt or Markdown docs served for agents.\n- 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.\n- 0 to 15, typed inputs with enums, constraints and required fields, and no free-form JSON blobs.\n- 0 to 15, examples and documented error responses.\n- 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."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-schema"
      },
      {
        "key": "maintenance",
        "name": "Maintenance \u0026 community",
        "score": 75,
        "maxGain": 2.2,
        "reason": "1.4.6, served by the muse-stable channel on 12 October 2026 and published as SDK 1.4.6 on 11 October (30). The SDK, released in step, published 1.3.0 on 18 September, 1.4.2 on 30 September, 1.4.4 on 8 October and 1.4.6 on 11 October (20). No public tracker for the CLI. The SDK repository takes issues with a written triage guide, and the help centre takes support tickets, neither tested (8 of 25). MIT TypeScript and Python SDKs at 1.4.6, labelled a developer preview (12 of 15). A checksummed launcher and release manifest, but no public CI for the program (5 of 10).",
        "checklist": [
          "- 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.\n- 20, at least three releases or dated changelog entries in the last 90 days.\n- 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.\n- 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models).\n- 10, package health, current dependencies and CI.",
          "Models are read for deprecation notice periods and model churn rather than release counts."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-maintenance"
      }
    ],
    "provenance": [
      {
        "label": "Domain age",
        "value": "meta.ai, registered 2017-12-16 (8 years)",
        "points": 11,
        "max": 15
      },
      {
        "label": "Privacy policy",
        "value": "published, but our reader couldn't read it",
        "points": 7,
        "max": 10
      },
      {
        "label": "Status page",
        "value": "not found",
        "points": 0,
        "max": 10
      },
      {
        "label": "security.txt",
        "value": "not found",
        "points": 0,
        "max": 10
      }
    ],
    "unchecked": [
      "unchecked: the CLI's own release dates. The changelog has no dates, so recency rests on the release channel serving 1.4.6 on 12 October and the SDK's npm publish of 1.4.6 on 11 October",
      "unchecked: what Muse Code's telemetry sends and what is on by default. The settings page lists telemetry options without detail",
      "unchecked: the open issues and maintainer replies on the SDK repository. GitHub's API was not available to our reader, and the CLI has no public tracker that we found",
      "The 1.4.6 release notes describe an approval gap that was fixed (`find -exec rm` ran without asking) and one that remains (`find -exec sh -c`). The sandbox still limits writes to the workspace. Recorded as a weakness with no deduction, a judgement call",
      "The product page says Muse Code is available for macOS and Windows, while the docs and the 31 August post give macOS, Linux and Windows. Graded on the docs",
      "No authenticated session was run"
    ],
    "weaknesses": [
      "Closed source, with no public CI or issue tracker for the CLI found and release notes that carry no dates",
      "Subscriptions default to Contributor mode, which lets Meta train on prompts, code and responses until it's turned off",
      "No free route. Plans run from $5 to $200 a month, and pay-as-you-go billing needs a card",
      "New sessions start in Auto-review, where an automated judge clears eligible approvals rather than a person",
      "MCP tools run outside the sandbox, and settings list telemetry options without documenting what is sent or the defaults"
    ],
    "agentNotes": [
      "Set `META_API_KEY` for headless runs. Browser sign-in isn't available in CI or for Meta Managed Accounts",
      "Gate on your own test command, not the exit code. `muse exec` returns 0 when the turn completes even if tests still fail",
      "Pass `--max-model-steps` on unattended runs so a stuck task can't loop",
      "Prefer `--disable-approval` to `--yolo` in CI so the sandbox stays on. On Linux the runner needs a working bubblewrap and a non-musl build",
      "Turn Contributor mode off in `/settings` before a subscription session touches private code"
    ],
    "recheck": "https://www.anchorterminal.com/builders/#disputes"
  },
  "meta": {
    "attribution": "Anchor Terminal (https://www.anchorterminal.com)",
    "docs": "https://www.anchorterminal.com/docs/",
    "generatedAt": "2026-10-12",
    "license": "CC-BY-4.0",
    "method": "https://www.anchorterminal.com/benchmark/",
    "methodology": "0.4",
    "openapi": "https://www.anchorterminal.com/openapi.json",
    "preview": false,
    "run": "2026-10-01",
    "runLabel": "October 2026 research run"
  }
}