{
  "fixes": {
    "slug": "claude-code",
    "name": "Claude Code",
    "listing": "https://www.anchorterminal.com/tools/claude-code",
    "markdown": "# Fix list: Claude Code\n\nFrom Anchor Terminal's listing at https://www.anchorterminal.com/tools/claude-code, the October 2026 research run, assessed 1 October 2026. Grade B, 62.2 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 Claude 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, 50 out of 100, up to 10 more on the total\n\nWhy it scored 50: Local-package reading. Official installers for macOS 13+, Windows 10 1809+, Ubuntu 20.04+, Debian 10+ and Alpine 3.19+, signed apt, dnf and apk repositories and a GPG-signed manifest of checksums for every binary (20). No public CI or test suite for the program, since the source isn't published, and the repository's workflows are issue triage and tests for mods (0). Over 5,000 open issues and 735 open pull requests, with area, platform and regression labels on new reports but no maintainer replies visible on the first page (10). Every change ships as a 2.1.x patch, including the switch of the starting permission mode to auto in 2.1.284, and the changelog has no dates and no breaking-change heading, though a stable channel skips releases with major regressions (5). 2.1.287 (15). Same reading as Cursor CLI and GitHub Copilot CLI, which are 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 and per-token API prices are public without a login, though plan usage limits aren't given as numbers (20). No free route. The free claude.ai plan excludes Claude Code, and the API, Bedrock, Google Cloud and Foundry are paid (0). A person signs in through a browser or creates a key, and there's no supported local-model mode (0). Judgement call, against Anthropic's own product. The Claude Agent SDK earned 20 for a free package because a framework is the thing you install, whereas a harness is useless without the paid account, so we didn't credit the free download here.\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, 80 out of 100, up to 3.5 more on the total\n\nWhy it scored 80: Framework reading (telemetry defaults, approvals, guardrails, sandboxing), five lines. Usage metrics on by default on the Claude API, without code, prompts or file paths, error reports on by default for Pro and Max sign-ins since v2.1.198, and every provider sends WebFetch hostnames to api.anthropic.com for a blocklist check, each with an opt-out. All off by default on Bedrock, Google Cloud and Foundry. Consumer plans train on data when the user's setting allows it (18). Six modes, allow, ask and deny rules, protected paths, and managed settings that can disable bypass and auto mode and that nothing else overrides. The sandbox is opt-in and absent on native Windows, and auto mode, a classifier rather than a person, is now the starting mode (16). The auto-mode classifier blocks named exfiltration patterns, the sandbox proxy denies hosts outside an allowlist that starts empty, and WebFetch checks a blocklist, but since 1 October 2026 a mod can approve a call a deny rule refuses unless managed settings or a Team or Enterprise sign-in hold it (11). Opt-in OpenTelemetry with tool content behind `OTEL_LOG_TOOL_CONTENT`, local transcripts for 30 days, and cloud sessions logged through a proxy (15). HackerOne bug bounty, SECURITY.md, advisories published with GHSA ids, and a valid security.txt per the 26 September check (20). SOC 2 isn't scored on the framework reading. Advisories count under negative events. Judgement call, written by an agent on Claude, and this is the same checklist the other four 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. Schema \u0026 documentation, 80 out of 100, up to 3.3 more on the total\n\nWhy it scored 80: Framework reading. No published JSON Schema for settings that we checked, but the settings reference types every key, and the stream-json output has typed message classes in the Python and TypeScript Agent SDKs (18). llms.txt at code.claude.com/docs/llms.txt with a Markdown version of each page (10). The permission-mode table says what each mode is best for, but the rules carry many version-specific exceptions (mods, protected paths, carve-outs by settings source) that a model has to read closely (15). Permission modes, sandbox settings and rule syntax are enumerated, though rules are strings parsed at runtime (12). Examples on every page and a troubleshooting section for install and auth errors (13). A changelog for every release, versions without dates, dated only through git tags (12).\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## 5. Agent ergonomics, 87 out of 100, up to 2.1 more on the total\n\nWhy it scored 87: Framework reading, adapted to a harness driven by a pipeline. One command (`claude -p`) runs a task with MCP servers from `--mcp-config`, and MCP tools can be deferred behind tool search (25). `--max-turns`, `--max-budget-usd` and a background-command timeout of 30 minutes by default stop runs, and large MCP results go to a file (20). json and stream-json output, and `-p` errors name the cause, such as Failed to authenticate (16). `--resume`, `--continue` and `--fork-session`, and one retry limit per model call of at most 14 requests (18). Python and TypeScript Agent SDKs, but the starting permission mode now depends on version, plan, provider and telemetry, so a pipeline has to set it (8).\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. Maintenance \u0026 community, 82 out of 100, up to 1.6 more on the total\n\nWhy it scored 82: 2.1.287 on 2026-10-01 (30). 76 releases between 3 July and 1 October (20). A public issue tracker with automatic triage, duplicate detection and labels, but over 5,000 open issues and no visible maintainer replies on the page we read (12). The Python and TypeScript Agent SDKs that run the same loop are current (15). Signed releases and a stable channel, but no public CI for the program (5).\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## 7. Transparency \u0026 trust, 84 out of 100, up to 1.4 more on the total\n\nMade of editorial 68, provenance 100.\n\nWhy it scored 84: Closed source with clear terms, `LICENSE.md` pointing to the Commercial Terms (15). The data-usage page states retention per account type (consumer 30 days or 5 years with training on, commercial 30 days, zero retention for qualified Enterprise), 30 days of plaintext local transcripts, 5 years for /feedback reports and 6 months for shared survey transcripts, and these agree with the privacy policy links, but the logging and error-tracking services are named only as third parties (25). No written deprecation policy. The README and setup page mark npm installs deprecated without a date, and default changes appear in an undated changelog (8). Telemetry is documented per provider and per service with an opt-out for each (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\n## Deductions\n\nEach comes off the total. A fixed and documented problem counts for less at the next check.\n\n- 2025-10-03 to 2026-06-25. 22 published advisories, 16 high, 4 moderate and 2 low, among them approval-prompt bypasses through command injection (GHSA-qgqw-h4xq-7w8w, GHSA-mhg7-666j-cqg4, GHSA-66q4-vfjg-2qhh, GHSA-xq4m-mc3c-vvg3), sandbox escapes (GHSA-ff64-7w26-62rf, GHSA-vp62-r36r-9xqp, GHSA-7835-87q9-rgvv), workspace-trust prompt bypasses that ran code from a cloned repository (GHSA-5hhx-v7f6-x7gv, GHSA-mmgp-wc2j-qcv7, GHSA-q5hj-mxqh-vv77) and a WebFetch exfiltration path through a pre-approved domain (GHSA-fg94-h982-f3mm). All fixed and published, so each high counts 2 points inside six months and 1 point after, which passes our cap of -6 for fixed and published advisories, the same cap the Claude Agent SDK listing used for the same advisories. None published since 25 June 2026 (https://github.com/anthropics/claude-code/security/advisories)\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- status.claude.com/history didn't render for our reader and its JSON is disallowed by robots.txt, so we have no incident record for the Claude Code component\n- unchecked: whether a JSON Schema for settings.json is published\n- The pricing page gives no numbers for plan usage limits\n- security.txt rests on the claude-agent-sdk listing's check of 26 September 2026\n- The issue page our reader loaded listed issues from 12 August 2026, so the counts may be cached\n\n## Weaknesses\n\n- The sandbox is off by default and native Windows has none\n- Auto mode, a classifier rather than a person, has been the starting mode for interactive sessions on every plan since 28 September 2026\n- 22 security advisories in the year to 25 June 2026, 16 rated high\n- Usage metrics on by default on the Claude API, and error reports on Pro and Max sign-ins\n- Closed source, Claude models only, and no free plan includes it\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- Pass `--permission-mode` on every `claude -p` run. An unset mode can start in auto mode, depending on version, plan, provider and telemetry\n- Turn on the sandbox with `sandbox.enabled` and set `allowUnsandboxedCommands` to false, or Claude can retry a blocked command outside it\n- Set `CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1` on a Claude login to stop metrics and error reports in one go\n- Set `autoUpdatesChannel` to stable or `DISABLE_AUTOUPDATER=1` in CI. Native installs update themselves about once a day\n- Cap pipeline runs with `--max-turns` and `--max-budget-usd`\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": "B",
    "score": 62.2,
    "assessed": "2026-10-01",
    "run": "October 2026 research run",
    "categories": [
      {
        "key": "reliability",
        "name": "Reliability",
        "score": 50,
        "maxGain": 10,
        "reason": "Local-package reading. Official installers for macOS 13+, Windows 10 1809+, Ubuntu 20.04+, Debian 10+ and Alpine 3.19+, signed apt, dnf and apk repositories and a GPG-signed manifest of checksums for every binary (20). No public CI or test suite for the program, since the source isn't published, and the repository's workflows are issue triage and tests for mods (0). Over 5,000 open issues and 735 open pull requests, with area, platform and regression labels on new reports but no maintainer replies visible on the first page (10). Every change ships as a 2.1.x patch, including the switch of the starting permission mode to auto in 2.1.284, and the changelog has no dates and no breaking-change heading, though a stable channel skips releases with major regressions (5). 2.1.287 (15). Same reading as Cursor CLI and GitHub Copilot CLI, which are 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 and per-token API prices are public without a login, though plan usage limits aren't given as numbers (20). No free route. The free claude.ai plan excludes Claude Code, and the API, Bedrock, Google Cloud and Foundry are paid (0). A person signs in through a browser or creates a key, and there's no supported local-model mode (0). Judgement call, against Anthropic's own product. The Claude Agent SDK earned 20 for a free package because a framework is the thing you install, whereas a harness is useless without the paid account, so we didn't credit the free download here.",
        "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": 80,
        "maxGain": 3.5,
        "reason": "Framework reading (telemetry defaults, approvals, guardrails, sandboxing), five lines. Usage metrics on by default on the Claude API, without code, prompts or file paths, error reports on by default for Pro and Max sign-ins since v2.1.198, and every provider sends WebFetch hostnames to api.anthropic.com for a blocklist check, each with an opt-out. All off by default on Bedrock, Google Cloud and Foundry. Consumer plans train on data when the user's setting allows it (18). Six modes, allow, ask and deny rules, protected paths, and managed settings that can disable bypass and auto mode and that nothing else overrides. The sandbox is opt-in and absent on native Windows, and auto mode, a classifier rather than a person, is now the starting mode (16). The auto-mode classifier blocks named exfiltration patterns, the sandbox proxy denies hosts outside an allowlist that starts empty, and WebFetch checks a blocklist, but since 1 October 2026 a mod can approve a call a deny rule refuses unless managed settings or a Team or Enterprise sign-in hold it (11). Opt-in OpenTelemetry with tool content behind `OTEL_LOG_TOOL_CONTENT`, local transcripts for 30 days, and cloud sessions logged through a proxy (15). HackerOne bug bounty, SECURITY.md, advisories published with GHSA ids, and a valid security.txt per the 26 September check (20). SOC 2 isn't scored on the framework reading. Advisories count under negative events. Judgement call, written by an agent on Claude, and this is the same checklist the other four 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": "schema",
        "name": "Schema \u0026 documentation",
        "score": 80,
        "maxGain": 3.3,
        "reason": "Framework reading. No published JSON Schema for settings that we checked, but the settings reference types every key, and the stream-json output has typed message classes in the Python and TypeScript Agent SDKs (18). llms.txt at code.claude.com/docs/llms.txt with a Markdown version of each page (10). The permission-mode table says what each mode is best for, but the rules carry many version-specific exceptions (mods, protected paths, carve-outs by settings source) that a model has to read closely (15). Permission modes, sandbox settings and rule syntax are enumerated, though rules are strings parsed at runtime (12). Examples on every page and a troubleshooting section for install and auth errors (13). A changelog for every release, versions without dates, dated only through git tags (12).",
        "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": "ergonomics",
        "name": "Agent ergonomics",
        "score": 87,
        "maxGain": 2.1,
        "reason": "Framework reading, adapted to a harness driven by a pipeline. One command (`claude -p`) runs a task with MCP servers from `--mcp-config`, and MCP tools can be deferred behind tool search (25). `--max-turns`, `--max-budget-usd` and a background-command timeout of 30 minutes by default stop runs, and large MCP results go to a file (20). json and stream-json output, and `-p` errors name the cause, such as Failed to authenticate (16). `--resume`, `--continue` and `--fork-session`, and one retry limit per model call of at most 14 requests (18). Python and TypeScript Agent SDKs, but the starting permission mode now depends on version, plan, provider and telemetry, so a pipeline has to set it (8).",
        "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": "maintenance",
        "name": "Maintenance \u0026 community",
        "score": 82,
        "maxGain": 1.6,
        "reason": "2.1.287 on 2026-10-01 (30). 76 releases between 3 July and 1 October (20). A public issue tracker with automatic triage, duplicate detection and labels, but over 5,000 open issues and no visible maintainer replies on the page we read (12). The Python and TypeScript Agent SDKs that run the same loop are current (15). Signed releases and a stable channel, but no public CI for the program (5).",
        "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"
      },
      {
        "key": "transparency",
        "name": "Transparency \u0026 trust",
        "score": 84,
        "maxGain": 1.4,
        "reason": "Closed source with clear terms, `LICENSE.md` pointing to the Commercial Terms (15). The data-usage page states retention per account type (consumer 30 days or 5 years with training on, commercial 30 days, zero retention for qualified Enterprise), 30 days of plaintext local transcripts, 5 years for /feedback reports and 6 months for shared survey transcripts, and these agree with the privacy policy links, but the logging and error-tracking services are named only as third parties (25). No written deprecation policy. The README and setup page mark npm installs deprecated without a date, and default changes appear in an undated changelog (8). Telemetry is documented per provider and per service with an opt-out for each (20).",
        "blend": "editorial 68, provenance 100",
        "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"
      }
    ],
    "deductions": [
      "2025-10-03 to 2026-06-25. 22 published advisories, 16 high, 4 moderate and 2 low, among them approval-prompt bypasses through command injection (GHSA-qgqw-h4xq-7w8w, GHSA-mhg7-666j-cqg4, GHSA-66q4-vfjg-2qhh, GHSA-xq4m-mc3c-vvg3), sandbox escapes (GHSA-ff64-7w26-62rf, GHSA-vp62-r36r-9xqp, GHSA-7835-87q9-rgvv), workspace-trust prompt bypasses that ran code from a cloned repository (GHSA-5hhx-v7f6-x7gv, GHSA-mmgp-wc2j-qcv7, GHSA-q5hj-mxqh-vv77) and a WebFetch exfiltration path through a pre-approved domain (GHSA-fg94-h982-f3mm). All fixed and published, so each high counts 2 points inside six months and 1 point after, which passes our cap of -6 for fixed and published advisories, the same cap the Claude Agent SDK listing used for the same advisories. None published since 25 June 2026 (https://github.com/anthropics/claude-code/security/advisories)"
    ],
    "unchecked": [
      "status.claude.com/history didn't render for our reader and its JSON is disallowed by robots.txt, so we have no incident record for the Claude Code component",
      "unchecked: whether a JSON Schema for settings.json is published",
      "The pricing page gives no numbers for plan usage limits",
      "security.txt rests on the claude-agent-sdk listing's check of 26 September 2026",
      "The issue page our reader loaded listed issues from 12 August 2026, so the counts may be cached"
    ],
    "weaknesses": [
      "The sandbox is off by default and native Windows has none",
      "Auto mode, a classifier rather than a person, has been the starting mode for interactive sessions on every plan since 28 September 2026",
      "22 security advisories in the year to 25 June 2026, 16 rated high",
      "Usage metrics on by default on the Claude API, and error reports on Pro and Max sign-ins",
      "Closed source, Claude models only, and no free plan includes it"
    ],
    "agentNotes": [
      "Pass `--permission-mode` on every `claude -p` run. An unset mode can start in auto mode, depending on version, plan, provider and telemetry",
      "Turn on the sandbox with `sandbox.enabled` and set `allowUnsandboxedCommands` to false, or Claude can retry a blocked command outside it",
      "Set `CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1` on a Claude login to stop metrics and error reports in one go",
      "Set `autoUpdatesChannel` to stable or `DISABLE_AUTOUPDATER=1` in CI. Native installs update themselves about once a day",
      "Cap pipeline runs with `--max-turns` and `--max-budget-usd`"
    ],
    "recheck": "https://www.anchorterminal.com/builders/#disputes"
  },
  "meta": {
    "attribution": "Anchor Terminal (https://www.anchorterminal.com)",
    "docs": "https://www.anchorterminal.com/docs/",
    "generatedAt": "2026-10-04",
    "license": "CC-BY-4.0",
    "method": "https://www.anchorterminal.com/benchmark/",
    "methodology": "0.3",
    "openapi": "https://www.anchorterminal.com/openapi.json",
    "preview": false,
    "run": "2026-10-01",
    "runLabel": "October 2026 research run"
  }
}