{
  "fixes": {
    "slug": "braintrust",
    "name": "Braintrust API + MCP",
    "listing": "https://www.anchorterminal.com/tools/braintrust",
    "markdown": "# Fix list: Braintrust API + MCP\n\nFrom Anchor Terminal's listing at https://www.anchorterminal.com/tools/braintrust, the October 2026 research run, assessed 1 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 Braintrust API + MCP: 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, 61 out of 100, up to 7.8 more on the total\n\nWhy it scored 61: Statuspage at status.braintrust.dev with component history and 25 incidents since February 2026 (20). The last 90 days hold several majors. A critical run of 504s on the EU data plane on 16 July (about 30 minutes), gateway 5xx incidents on 3 August and 27 to 28 August, control plane 5xxs on 4 September, a 40-minute data-loading failure in the US on 9 September and 78 minutes of elevated errors across the US data plane API on 30 September. Most were short and the gateway is a side product, so 5 rather than 0. The only published limit is about 20 BTQL queries a minute on Starter and Pro. Ingestion and REST limits have no numbers (8). The OpenAPI spec documents 429 with a `Retry-After` header on 154 operations and recommends exponential backoff, the SDKs retry 408, 429 and 5xx, and events written with the same `id` overwrite rather than duplicate (15). Enterprise lists SLAs and guaranteed response times with no published terms (3). API and MCP are GA (10).\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, 40 out of 100, up to 7.5 more on the total\n\nWhy it scored 40: No x402 or other machine payment (0). Usage prices published without login, $4 a GB and $2.50 per 1,000 scores on Starter, $3 and $1.50 on Pro (20). Starter is free with no card (20). A person signs up in a browser to create a key. The MCP OAuth flow also needs a browser sign-in (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, 58 out of 100, up to 7.4 more on the total\n\nWhy it scored 58: User API keys and service tokens as Bearer tokens. Service tokens attach to permission groups that can be scoped to single projects on any plan, keys can be deleted through `/v1/api_key` and `/v1/service_token`, and the MCP server also takes OAuth 2.0 with dynamic client registration (25). Viewer groups exist, but the MCP server gained write tools in August with no server-side read-only mode, and the docs leave confirmation to the client (12). Traces hold whatever the application logged, and we found no prompt-injection guidance for agents reading them (3). The October changelog mentions audit logging, and a knowledge-base page says org-wide API key auditing is in the UI only (8). SOC 2 Type II is claimed on the pricing page. The Vanta trust centre at trust.braintrust.dev rendered nothing readable to us, there's no security.txt and no SECURITY.md in either SDK repository, and the two credential-capture fixes below were explained in knowledge-base pages rather than advisories (10).\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. Agent ergonomics, 66 out of 100, up to 5.5 more on the total\n\nWhy it scored 66: 42 MCP tools load at once, with no toolsets or server-side allowlist found (5). REST lists take `limit`, `starting_after` and `ending_before`, and `sql_query` truncates field values to 1,024 characters by default and swaps results over 1 MB for a signed `overflow_url` (20). Error codes are declared per operation, the MCP server returns 413 above 100 MiB, and 429 carries `Retry-After`. Error bodies are typed as plain text (14). Event inserts upsert on `id` and ACL batch updates are documented as idempotent. Several MCP tools have a no-write `test_*` twin. We couldn't see whether the hosted tools carry `readOnlyHint` or `destructiveHint` (12). Few required parameters, and SDKs in TypeScript, Python, Go, Java, Ruby and C# (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## 5. Transparency \u0026 trust, 68 out of 100, up to 2.8 more on the total\n\nMade of editorial 50, provenance 86.\n\nWhy it scored 68: Platform closed source with clear terms. Both SDKs are Apache-2.0 (18). The privacy notice was last updated on 21 September 2023 and says it doesn't apply to data customers send to the service. Retention per plan is on the pricing page and a DPA is click-through on Pro. We found no retention statement for trace data beyond that (12). The changelog dates deprecations by month, such as `@braintrust/openai-agents` and the bt CLI `--api-key` flag in August 2026, and breaking SDK changes are called out (12). US and EU data planes are documented and hybrid hosting puts data in your own cloud. No subprocessor list we could read (8).\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: braintrust.dev, registered 2021-03-25 (5 years) (11 of 15)\n- security.txt: not found (0 of 10)\n\n## 6. Schema \u0026 documentation, 85 out of 100, up to 2.4 more on the total\n\nWhy it scored 85: OpenAPI 3.0.3 with 75 paths and 234 operations at /docs/openapi.json, pinned and refreshed weekly in the Python SDK repo (25). llms.txt with about 500 entries, every page served as Markdown (10). The MCP docs give a one-line purpose for each of the 42 tools and mark the `test_*` tools as dry runs, but say little about when not to use a tool. The MCP server is closed source, so we read the docs rather than the definitions (13). Typed OpenAPI schemas with pagination parameters and enums, though `sql_query` takes free-form SQL and event payloads are free-form by nature (11). 400, 401, 403, 429 and 500 responses are declared on most operations, but only 4 of 234 operations carry an inline example. The cookbook fills some of the gap (11). Versioned `/v1` API, a monthly product changelog, a data-plane changelog and per-SDK changelogs (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, 85 out of 100, up to 1.3 more on the total\n\nWhy it scored 85: TypeScript SDK `braintrust` 3.36.0 and Python SDK 0.44.0, both on 2026-10-01 (30). 15 TypeScript and 19 Python releases since 16 July (20). The product changelog lists dozens of September entries and both SDK repositories take outside contributions credited in the changelog. We didn't load issue reply times (12). Official SDKs in six languages, current (15). Public CI (`checks.yaml`), integration tests, Dependabot and a weekly dependency job. We didn't confirm the pass state (8).\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## Deductions\n\nEach comes off the total. A fixed and documented problem counts for less at the next check.\n\n- Fixed 2026-07-16. The Python SDK's LiteLLM instrumentation in versions 0.2.0 to 0.27.x wrote provider credentials into trace metadata, among them API keys, `Authorization` headers, Azure tokens and AWS access and secret keys. Fixed in 0.28.0 with a metadata allowlist, and a knowledge-base page tells users to rotate. It reads as an advisory but isn't filed as one and carries no date. -2 after decay (https://braintrust.dev/docs/kb/litellm-instrumentation-may-record-provider-credentials.md)\n- Fixed 2026-07-16. The TypeScript SDK from 0.4.0 to 3.23.0 recorded API keys passed to the Claude Agent SDK through `options.apiKey` into trace metadata. The 3.23.1 changelog line calls the fix \"Clean up span metadata\", and only the knowledge-base page explains the exposure and asks for rotation. -2 after decay (https://braintrust.dev/docs/kb/claude-agent-sdk-api-keys-recorded-in-trace-metadata.md)\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 trust centre at trust.braintrust.dev, which rendered nothing without JavaScript, so certifications beyond the pricing page's SOC 2 Type II claim and the subprocessor list are unverified\n- Whether the hosted MCP tools carry `readOnlyHint` and `destructiveHint` annotations, since the server source isn't public\n- When the two credential-capture knowledge-base pages were published and whether affected customers were contacted\n- Ingestion and REST API rate limits, which aren't published\n- GitHub issue reply times on the SDK repositories, which we didn't load\n\n## Weaknesses\n\n- Several major incidents on the status page between 16 July and 30 September 2026, the longest 78 minutes on the US data plane\n- Python SDK up to 0.27.x and TypeScript SDK up to 3.23.0 recorded provider credentials in trace metadata, fixed on 2026-07-16 without a formal advisory\n- All 42 MCP tools load by default and write tools act with the full permissions of the signed-in account\n- Only BTQL has a published rate limit, about 20 queries a minute on Starter and Pro\n- Privacy notice dates from September 2023 and excludes customer data. Starter keeps data for 14 days\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- Connect a project-scoped service token rather than a personal key, because MCP write tools act with the key's full permissions\n- Set the client to confirm MCP write tools such as `edit_dataset_rows` and `create_threshold_alert`. The server has no read-only mode\n- Cache `sql_query` results. Starter and Pro allow about 20 queries a minute and return 429\n- Pass `preview_length: -1` to `sql_query` only when full values are needed, and fetch `overflow_url` when a result passes 1 MB\n- Upgrade to Python `braintrust` 0.28.0 or TypeScript 3.23.1 or later and rotate any provider keys traced before July 2026\n\n## What the review panel asked for\n\n- advisories for security fixes\n- day-level deprecation dates\n- publish tool definitions\n- server-side toolsets\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": "2026-10-01",
    "run": "October 2026 research run",
    "categories": [
      {
        "key": "reliability",
        "name": "Reliability",
        "score": 61,
        "maxGain": 7.8,
        "reason": "Statuspage at status.braintrust.dev with component history and 25 incidents since February 2026 (20). The last 90 days hold several majors. A critical run of 504s on the EU data plane on 16 July (about 30 minutes), gateway 5xx incidents on 3 August and 27 to 28 August, control plane 5xxs on 4 September, a 40-minute data-loading failure in the US on 9 September and 78 minutes of elevated errors across the US data plane API on 30 September. Most were short and the gateway is a side product, so 5 rather than 0. The only published limit is about 20 BTQL queries a minute on Starter and Pro. Ingestion and REST limits have no numbers (8). The OpenAPI spec documents 429 with a `Retry-After` header on 154 operations and recommends exponential backoff, the SDKs retry 408, 429 and 5xx, and events written with the same `id` overwrite rather than duplicate (15). Enterprise lists SLAs and guaranteed response times with no published terms (3). API and MCP are GA (10).",
        "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": 40,
        "maxGain": 7.5,
        "reason": "No x402 or other machine payment (0). Usage prices published without login, $4 a GB and $2.50 per 1,000 scores on Starter, $3 and $1.50 on Pro (20). Starter is free with no card (20). A person signs up in a browser to create a key. The MCP OAuth flow also needs a browser sign-in (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": 58,
        "maxGain": 7.4,
        "reason": "User API keys and service tokens as Bearer tokens. Service tokens attach to permission groups that can be scoped to single projects on any plan, keys can be deleted through `/v1/api_key` and `/v1/service_token`, and the MCP server also takes OAuth 2.0 with dynamic client registration (25). Viewer groups exist, but the MCP server gained write tools in August with no server-side read-only mode, and the docs leave confirmation to the client (12). Traces hold whatever the application logged, and we found no prompt-injection guidance for agents reading them (3). The October changelog mentions audit logging, and a knowledge-base page says org-wide API key auditing is in the UI only (8). SOC 2 Type II is claimed on the pricing page. The Vanta trust centre at trust.braintrust.dev rendered nothing readable to us, there's no security.txt and no SECURITY.md in either SDK repository, and the two credential-capture fixes below were explained in knowledge-base pages rather than advisories (10).",
        "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": "ergonomics",
        "name": "Agent ergonomics",
        "score": 66,
        "maxGain": 5.5,
        "reason": "42 MCP tools load at once, with no toolsets or server-side allowlist found (5). REST lists take `limit`, `starting_after` and `ending_before`, and `sql_query` truncates field values to 1,024 characters by default and swaps results over 1 MB for a signed `overflow_url` (20). Error codes are declared per operation, the MCP server returns 413 above 100 MiB, and 429 carries `Retry-After`. Error bodies are typed as plain text (14). Event inserts upsert on `id` and ACL batch updates are documented as idempotent. Several MCP tools have a no-write `test_*` twin. We couldn't see whether the hosted tools carry `readOnlyHint` or `destructiveHint` (12). Few required parameters, and SDKs in TypeScript, Python, Go, Java, Ruby and C# (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": "transparency",
        "name": "Transparency \u0026 trust",
        "score": 68,
        "maxGain": 2.8,
        "reason": "Platform closed source with clear terms. Both SDKs are Apache-2.0 (18). The privacy notice was last updated on 21 September 2023 and says it doesn't apply to data customers send to the service. Retention per plan is on the pricing page and a DPA is click-through on Pro. We found no retention statement for trace data beyond that (12). The changelog dates deprecations by month, such as `@braintrust/openai-agents` and the bt CLI `--api-key` flag in August 2026, and breaking SDK changes are called out (12). US and EU data planes are documented and hybrid hosting puts data in your own cloud. No subprocessor list we could read (8).",
        "blend": "editorial 50, provenance 86",
        "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": "schema",
        "name": "Schema \u0026 documentation",
        "score": 85,
        "maxGain": 2.4,
        "reason": "OpenAPI 3.0.3 with 75 paths and 234 operations at /docs/openapi.json, pinned and refreshed weekly in the Python SDK repo (25). llms.txt with about 500 entries, every page served as Markdown (10). The MCP docs give a one-line purpose for each of the 42 tools and mark the `test_*` tools as dry runs, but say little about when not to use a tool. The MCP server is closed source, so we read the docs rather than the definitions (13). Typed OpenAPI schemas with pagination parameters and enums, though `sql_query` takes free-form SQL and event payloads are free-form by nature (11). 400, 401, 403, 429 and 500 responses are declared on most operations, but only 4 of 234 operations carry an inline example. The cookbook fills some of the gap (11). Versioned `/v1` API, a monthly product changelog, a data-plane changelog and per-SDK changelogs (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": 85,
        "maxGain": 1.3,
        "reason": "TypeScript SDK `braintrust` 3.36.0 and Python SDK 0.44.0, both on 2026-10-01 (30). 15 TypeScript and 19 Python releases since 16 July (20). The product changelog lists dozens of September entries and both SDK repositories take outside contributions credited in the changelog. We didn't load issue reply times (12). Official SDKs in six languages, current (15). Public CI (`checks.yaml`), integration tests, Dependabot and a weekly dependency job. We didn't confirm the pass state (8).",
        "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": "braintrust.dev, registered 2021-03-25 (5 years)",
        "points": 11,
        "max": 15
      },
      {
        "label": "security.txt",
        "value": "not found",
        "points": 0,
        "max": 10
      }
    ],
    "deductions": [
      "Fixed 2026-07-16. The Python SDK's LiteLLM instrumentation in versions 0.2.0 to 0.27.x wrote provider credentials into trace metadata, among them API keys, `Authorization` headers, Azure tokens and AWS access and secret keys. Fixed in 0.28.0 with a metadata allowlist, and a knowledge-base page tells users to rotate. It reads as an advisory but isn't filed as one and carries no date. -2 after decay (https://braintrust.dev/docs/kb/litellm-instrumentation-may-record-provider-credentials.md)",
      "Fixed 2026-07-16. The TypeScript SDK from 0.4.0 to 3.23.0 recorded API keys passed to the Claude Agent SDK through `options.apiKey` into trace metadata. The 3.23.1 changelog line calls the fix \"Clean up span metadata\", and only the knowledge-base page explains the exposure and asks for rotation. -2 after decay (https://braintrust.dev/docs/kb/claude-agent-sdk-api-keys-recorded-in-trace-metadata.md)"
    ],
    "unchecked": [
      "unchecked: the trust centre at trust.braintrust.dev, which rendered nothing without JavaScript, so certifications beyond the pricing page's SOC 2 Type II claim and the subprocessor list are unverified",
      "Whether the hosted MCP tools carry `readOnlyHint` and `destructiveHint` annotations, since the server source isn't public",
      "When the two credential-capture knowledge-base pages were published and whether affected customers were contacted",
      "Ingestion and REST API rate limits, which aren't published",
      "GitHub issue reply times on the SDK repositories, which we didn't load"
    ],
    "weaknesses": [
      "Several major incidents on the status page between 16 July and 30 September 2026, the longest 78 minutes on the US data plane",
      "Python SDK up to 0.27.x and TypeScript SDK up to 3.23.0 recorded provider credentials in trace metadata, fixed on 2026-07-16 without a formal advisory",
      "All 42 MCP tools load by default and write tools act with the full permissions of the signed-in account",
      "Only BTQL has a published rate limit, about 20 queries a minute on Starter and Pro",
      "Privacy notice dates from September 2023 and excludes customer data. Starter keeps data for 14 days"
    ],
    "agentNotes": [
      "Connect a project-scoped service token rather than a personal key, because MCP write tools act with the key's full permissions",
      "Set the client to confirm MCP write tools such as `edit_dataset_rows` and `create_threshold_alert`. The server has no read-only mode",
      "Cache `sql_query` results. Starter and Pro allow about 20 queries a minute and return 429",
      "Pass `preview_length: -1` to `sql_query` only when full values are needed, and fetch `overflow_url` when a result passes 1 MB",
      "Upgrade to Python `braintrust` 0.28.0 or TypeScript 3.23.1 or later and rotate any provider keys traced before July 2026"
    ],
    "requests": [
      {
        "text": "advisories for security fixes",
        "reviews": 1
      },
      {
        "text": "day-level deprecation dates",
        "reviews": 1
      },
      {
        "text": "publish tool definitions",
        "reviews": 1
      },
      {
        "text": "server-side toolsets",
        "reviews": 1
      }
    ],
    "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"
  }
}