# Fix list: Cohere Chat API (Command models) From Anchor Terminal's listing at https://www.anchorterminal.com/tools/cohere-chat, the October 2026 research run, assessed 9 October 2026. Grade B, 64.7 out of 100. This is everything the published grade says the listing lacks, the biggest possible gain to the total first. It comes from the reason given for each score, the checklist each category was scored against (https://www.anchorterminal.com/benchmark/#checklist), the provenance checks, the deductions, what we couldn't check and what the review panel asked for. A fix counts at the next check, once it's public. For a coding agent working on Cohere Chat API (Command models): work through the items below in the product, its docs and its public pages. Each category gives the reason for its score, with the points each checklist item earned, and the checklist itself, so the gap is the items that earned less than their points. Change the product, not the wording, and keep a note of what you changed and where it's published. ## 1. Payments & pricing, 30 out of 100, up to 8.8 more on the total Why it scored 30: No x402, MPP or L402 in the docs, the OpenAPI file or the pricing page (0). The docs say Command models are billed per input and output token, and the pricing page's Generative models tab is script-drawn, so current prices were unread. The page text shows only legacy prices, such as $2.50 and $10.00 per million tokens for Command R+ 08-2024. Half marks, and the gap is ours to state (10 of 20). A trial key is created at sign-up and is free with no card (20). Sign-up is in a browser, and a production key needs an Owner to complete a Go to Production form (0). The checklist (https://www.anchorterminal.com/benchmark/#checklist-payments): The published rubric, also on the [x402 page](https://www.anchorterminal.com/x402/). - 40, a machine payment protocol (x402, MPP or L402) on the tool's own endpoints. 10 to 30 when it covers only some endpoints or only goes through a third party, and the note says which. - 20, per-call or per-unit pricing published without a login. 10 for public plan-only pricing, 0 for "contact sales" or prices behind a login. - 20, a free tier or trial that doesn't need a card. - 20, autonomous onboarding, meaning an agent can get access without a person signing up in a browser (keyless use, x402, a programmatic key API). Payment platforms and agent wallets rarely charge for their own API over a machine protocol, so the first line has steps for them, and the highest one that applies counts. 40 when x402, MPP or L402 runs on all their own endpoints, 30 when it runs on part of their own API, 25 when their merchants can accept one, 20 for running a facilitator, 15 for paying as a buyer, and 0 when the only protocol is their own. Merchant acceptance sits above a facilitator because the platform's own customers can charge agents through it, while a facilitator settles for sellers who wire up the protocol themselves. The counter-argument (a facilitator does more for the protocol as a whole) has a point. Each note says which step applied. Open-source software you run yourself is scored on its hosted or paid option if it has one. A free, self-hosted package with nothing to buy gets 20, 20 and 20 for the last three lines, and 0 to 40 for the first only if it ships a payment protocol. ## 2. Security & auth, 57 out of 100, up to 7.5 more on the total Why it scored 57: Model reading. Bearer keys of two kinds, trial and production. Team roles are Owner and User, only Owners delete production keys, and a production key is shown in full once. No scopes, expiry or per-model restriction was found (18 of 30). Paying customers can switch training off under Data Controls, which implies it is on until they do, trial keys fall under the website terms, and the SaaS agreement grants an irrevocable right to use API data to improve the services and share it with third parties (8 of 20). Logged prompts and generations are deleted after 30 days unless flagged, and zero data retention needs approval (12 of 15). The dashboard shows usage history, and no customer audit log was found (2 of 15). The trust centre lists SOC 2 Type II, ISO 27001, ISO 42001, UK Cyber Essentials, penetration tests on request and a responsible disclosure policy with a bug bounty. cohere.com/.well-known/security.txt answers 404 (17 of 20). The checklist (https://www.anchorterminal.com/benchmark/#checklist-security): - 0 to 30, the credential model. 30 for OAuth 2.1 with scopes, or scoped and revocable keys with rotation. 20 for plain revocable API keys. 10 for one all-powerful key. 10 off when a secret can travel in a URL query string as a documented option. - 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions. - 0 to 15, prompt-injection posture where the tool returns untrusted content (documented mitigations or guidance). A tool that returns no untrusted content gets 10. - 0 to 15, audit logs or per-call visibility for the operator. - 0 to 20, a security programme. security.txt or a disclosure policy, a bug bounty, SOC 2 or ISO 27001, advisories handled in public. Models are read for retention, whether API data trains models (and whether that's off by default), zero-retention options and certifications. Frameworks for telemetry defaults, approval hooks, guardrails and sandboxing. ## 3. Reliability, 64 out of 100, up to 7.2 more on the total Why it scored 64: Hosted lines. status.cohere.com (incident.io) lists each Command model and the endpoints as components with uptime from July to October 2026 (20). Its feed shows three incidents in 90 days. 'Multiple Endpoint Disruptions' on 1 September 2026 touched embeddings, rerank and `command-a-reasoning-08-2025`, with no duration in the feed, and the other two were docs loading on 5 August and dashboard login on 23 July. Model and endpoint uptime still reads 100 per cent, so we count minor incidents only (20). Chat limits are published per model, 20 requests a minute and 1,000 calls a month on a trial key and 500 a minute on a production key for Command A, R, R+ and R7B, while the four newest variants say contact sales (12 of 15). The errors page documents 429 with sample messages. No Retry-After header, backoff guidance or idempotency key was found (5 of 15). cohere.com/slo is titled Service Level Objective and its text is script-drawn, so it was unread and scores as absent (0). The v2 chat endpoint is the default docs version, but the sample response on the chat guide still carries `is_experimental` and the newest models are not open to production keys (7 of 10). The checklist (https://www.anchorterminal.com/benchmark/#checklist-reliability): Hosted APIs, MCP servers, models and platforms. - 20, a public status page with component history (Statuspage, Instatus, BetterStack or the vendor's own). - 0 to 30, the incident record for the last 90 days on that page. 30 for a clean record or trivial incidents only, 20 for minor incidents only, 10 for one major outage (an hour or more of a core API down, or errors across the board), 0 for several. 5 when there's no history we could read, and the note says so. - 15, rate limits documented with numbers. - 15, documented 429 or overload handling (Retry-After, backoff guidance), and idempotency keys or safe-retry guidance where writes are involved. - 10, an SLA published for any paid tier. - 10, the surface agents use is generally available, not beta or preview. Local packages, SDKs, frameworks and stdio MCP servers. - 20, installs from an official package with supported runtimes stated. - 25, a public CI and test suite, passing on the default branch. - 0 to 25, open crash or regression issues relative to activity (25 for few and handled, 0 for many, old and unanswered). - 15, semver discipline and breaking changes called out in a changelog. - 15, version 1.0 or later, or declared stable. Protocols are read from their reference implementations, the public facilitators or servers, spec stability and test vectors. ## 4. Agent ergonomics, 72 out of 100, up to 4.6 more on the total Why it scored 72: Model reading. Function calling with `tool_choice` REQUIRED or NONE and `strict_tools`, which the reference marks beta (17 of 20). `response_format` takes `json_object` or a JSON Schema, with limits documented and no support in RAG mode (13 of 15). The usage object reports `cached_tokens`, and no guide or price for prompt caching was found (5 of 15). Context is 256,000 tokens on Command A and Command A Reasoning and 128,000 on Command A+ (10 of 15). `/v2/batches` runs a batch from an uploaded dataset, with no price or discount found (6 of 10). Official SDKs for Python, TypeScript, Java and Go (10). Error messages are specific and readable, with no stable error codes (11 of 15). The checklist (https://www.anchorterminal.com/benchmark/#checklist-ergonomics): - 0 to 25, context cost. For MCP, the number and size of the tool definitions (25 for ten or fewer compact tools, 15 for 11 to 30, 5 for more than 30, plus up to 10 back for toolsets, dynamic loading or read-only subsets). For APIs, whether responses can be sized (field selection, limits, summaries). - 20, pagination, filtering and output-size controls. - 20, actionable, documented error responses, codes and messages an agent can recover from. - 20, idempotency or safe retries, and for MCP the `readOnlyHint` and `destructiveHint` annotations. - 15, sensible defaults, few required parameters, and official SDKs in at least two languages. Models are read for tool use, structured output, prompt caching, context length, batch and SDKs. Frameworks for how much code and how many defaults a tool-calling agent with MCP needs. ## 5. Schema & documentation, 84 out of 100, up to 2.6 more on the total Why it scored 84: Model reading. Public OpenAPI 3.1 file `cohere-openapi.yaml` in cohere-ai/cohere-developer-experience, 43 operations, last changed 29 September 2026 (25). docs.cohere.com/llms.txt and a Markdown twin of every page by adding `.md`, though the llms.txt files list no pages (8 of 10). The v2 chat operation describes each of its 21 request fields with defaults and limits, and says less about when to prefer one option (16 of 20). Enums on `tool_choice`, `safety_mode` and `thinking.type`, numeric bounds on `temperature`, `k`, `p` and `priority`, with penalties bounded only in prose (12 of 15). Eight request examples on the operation and an errors page with sample messages for each status code, with error bodies carrying a message and no machine code (12 of 15). v1 and v2 live side by side in the path and the changelog is dated, but the OpenAPI file is version 1.0 throughout and its changes are committed as 'Add spec changes' with no per-change note (11 of 15). The checklist (https://www.anchorterminal.com/benchmark/#checklist-schema): APIs and MCP servers. - 25, a machine-readable contract (a public OpenAPI file or similar; for MCP, typed JSON Schema inputs on every tool). - 10, llms.txt or Markdown docs served for agents. - 0 to 20, descriptions that say what a tool is for, when to use it and when not to, read from the tool definitions in the source or the API reference. - 0 to 15, typed inputs with enums, constraints and required fields, and no free-form JSON blobs. - 0 to 15, examples and documented error responses. - 15, versioning and a public changelog. Models are read from the API reference, the OpenAPI file, llms.txt, the structured-output and tool-use docs and the model cards. Frameworks from docs a model can follow, typed interfaces, examples and the API reference. ## 6. Maintenance & community, 72 out of 100, up to 2.5 more on the total Why it scored 72: Model reading. The last dated release on the Chat endpoint is North Small Translate on 9 September 2026, 30 days before the check. The last Command model is Command A+ on 20 May 2026, and the SDKs last shipped on 26 August, so we score the 90-day band (20 of 30). The deprecations page defines Active, Legacy, Deprecated and Shutdown and promises email notice, with no minimum notice period (6 of 12). No chat model was retired in the last 90 days. The last retirement was two Aya models on 4 April 2026 (8 of 8). Dated changelog with three entries since 27 August, one of them for a chat model (12 of 15). SDK issue replies not sampled (5 of 10). Python SDK 7.0.7, 7.0.8, 7.0.9 and 7.1.0 between 21 July and 26 August 2026, with TypeScript 8.1.0, Java 1.11.0 and Go v2.19.0 on 26 August (15). The Python SDK is generated from the OpenAPI file and has a CI workflow, whose runs we did not read (6 of 10). The checklist (https://www.anchorterminal.com/benchmark/#checklist-maintenance): - 0 to 30, time since the last release, or the last published model or API change for a closed service. 30 within 30 days, 20 within 90, 10 within 180, 0 older. - 20, at least three releases or dated changelog entries in the last 90 days. - 0 to 25, responsiveness. Issues and pull requests answered on GitHub (the open issues and how recent the replies are). For closed services, a public changelog and a support or community channel that answers, 0 to 15. - 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models). - 10, package health, current dependencies and CI. Models are read for deprecation notice periods and model churn rather than release counts. ## 7. Transparency & trust, 74 out of 100, up to 2.3 more on the total Made of editorial 65, provenance 83. Why it scored 74: Closed service under a published SaaS agreement, MIT SDKs, and the docs say Command A+ weights are on Hugging Face under Apache 2.0, which we did not open (15 of 30). The Enterprise Data Commitments (5 December 2025) and the privacy policy (1 May 2026) agree on 30-day retention and a training opt-out. The SaaS agreement (8 April 2025) does not mention the opt-out and grants an irrevocable right to use and share API data, and the DPA is sent on request (16 of 30). Deprecations page with dated history back to 2024 and lifecycle terms, no minimum notice (16 of 20). The trust centre names nine sub-processors, all in the US, and Google Cloud US-Central as the data location (18 of 20). The checklist (https://www.anchorterminal.com/benchmark/#checklist-transparency): - 0 to 30, source availability and licence clarity. 30 for open source under an OSI licence, 15 for closed with clear terms, 0 for unclear terms. - 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors). - 0 to 20, a deprecation policy or notices with dates. - 0 to 20, telemetry disclosed with an opt-out (local software), or subprocessors and data locations disclosed (hosted). The other half of Transparency and trust is the provenance score, computed from checked facts (below). The category score is the mean of the two. Provenance checks not met in full (half of this category, computed from checked facts): - Terms of service: read, states 6 of the 7 things a reader expects, and has 3 clauses that cost points (3.1 of 10) - security.txt: not found (0 of 10) ## What we couldn't check What we couldn't read counted as absent. Publishing it on a page a plain HTTP fetch can read (not only in a browser) lets the next check count it. - unchecked: current per-token prices for Command A, Command A+ and the other live Command models. The Generative models tab of cohere.com/pricing is script-drawn - unchecked: the Service Level Objective at cohere.com/slo, whose text is script-drawn. Reliability scores an SLA as absent - unchecked: the Data Processing Addendum, sent on request, and the SOC 2 report, behind an NDA - unchecked: the Hugging Face page for Command A+ weights and its licence - unchecked: duration and scope of the 1 September 2026 'Multiple Endpoint Disruptions' incident, which the feed gives only as resolved - unchecked: GitHub stars, download counts and issue responsiveness for the SDK repositories - The chat guide's examples call api.cohere.ai while the OpenAPI file names api.cohere.com as the server. Both are Cohere domains and we sent nothing to either - Whether training on prompts is on by default for paying customers is implied by the opt-out toggle and not stated in the pages read - docs.cohere.com/llms.txt carries a section headed 'Instructions for AI Agents' (append `.md`, use the docs MCP server). Recorded as a fact and not acted on beyond reading Markdown twins - The SaaS agreement bars use of the services for benchmarking or competitive analysis. No deduction, and it matters before any probe is run - A slip on our side. One typed address, docs.cohere.com/docs/rate-limit.md, answered 404 and yielded nothing. The real page was then read from the docs source ## Weaknesses - Production keys work like trial keys on Command A+, Reasoning, Translate and Vision. The docs send production use of those models to sales or to Model Vault - Prices for current Command models sit in a script-drawn tab of the pricing page that we could not read. Only legacy prices appear in the page text - The SaaS agreement of 8 April 2025 grants Cohere an irrevocable right to use API data to improve its services and to share it with third parties - No Retry-After header, backoff guidance or idempotency key was found in the reviewed documentation - The SaaS agreement forbids use of the services for benchmarking or competitive analysis. Recorded as a fact, and it matters before any probe is run ## What costs an agent a turn today The notes we give agents before they call it. Each one is a workaround an agent shouldn't need. - Call `POST https://api.cohere.com/v2/chat` with `Authorization: bearer `. `model` and `messages` are the only required fields - Use `command-a-03-2025`, `command-r-08-2024`, `command-r-plus-08-2024` or `command-r7b-12-2024` for production traffic. Newer variants stay at trial limits on a production key - Keep a trial key under 20 chat requests a minute and 1,000 calls a month, and do not use it for commercial work - Set `response_format` with a `json_schema` for structured output, and tell the model to produce JSON when using `json_object` without a schema - Ask the account Owner to turn off training under Data Controls in the dashboard before sending confidential prompts ## When it's done Send what changed and where it's published as a dispute (https://www.anchorterminal.com/builders/#disputes, or `POST https://www.anchorterminal.com/api/v1/contact` with `"kind": "dispute"`). Disputes are answered in public, and the listing is checked again by the same checklist. Paying for an audit or a listing claim changes nothing here.