# Fix list: Centrifugo From Anchor Terminal's listing at https://www.anchorterminal.com/tools/centrifugo, the October 2026 research run, assessed 9 October 2026. Grade B, 63.1 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 Centrifugo: work through the items below in the product, its docs and its public pages. Each category gives the reason for its score, with the points each checklist item earned, and the checklist itself, so the gap is the items that earned less than their points. Change the product, not the wording, and keep a note of what you changed and where it's published. ## 1. Security & auth, 37 out of 100, up to 11 more on the total Why it scored 37: The server API has one key for every method, set in the configuration file and changed by editing it. Client connections use JWTs with expiry and per-channel subscription tokens (15). The docs give the `api_key` URL query parameter as an option (minus 10). Client permissions are denied by default and granted per namespace, and the admin UI, Swagger UI and keyless mode are off by default. The open-source edition has no read-only API key and no confirmation on `history_remove` or `disconnect` (9). Channel messages are whatever other publishers sent. The docs mark emulated headers as untrusted input and give no guidance for agents reading messages (5). Logs at debug level show API requests and Prometheus metrics count API commands. No audit log in the open-source edition (6). `SECURITY.md`, GitHub private reporting, eleven advisories published with fixes in 2026, Dependabot and Renovate. No security.txt (HTTP 404), bug bounty or certification found (12). The checklist (https://www.anchorterminal.com/benchmark/#checklist-security): - 0 to 30, the credential model. 30 for OAuth 2.1 with scopes, or scoped and revocable keys with rotation. 20 for plain revocable API keys. 10 for one all-powerful key. 10 off when a secret can travel in a URL query string as a documented option. - 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions. - 0 to 15, prompt-injection posture where the tool returns untrusted content (documented mitigations or guidance). A tool that returns no untrusted content gets 10. - 0 to 15, audit logs or per-call visibility for the operator. - 0 to 20, a security programme. security.txt or a disclosure policy, a bug bounty, SOC 2 or ISO 27001, advisories handled in public. Models are read for retention, whether API data trains models (and whether that's off by default), zero-retention options and certifications. Frameworks for telemetry defaults, approval hooks, guardrails and sandboxing. ## 2. Payments & pricing, 60 out of 100, up to 5 more on the total Why it scored 60: Read with the self-hosted rule, on the open-source server an agent would use. No x402, MPP or L402 (0). The server is free with nothing to buy (20). No card, trial or account (20). An agent can fetch the binary or Docker image and generate its own configuration and key with `centrifugo genconfig`, with no signup (20). Centrifugo PRO is priced by quote, which this score does not count. The checklist (https://www.anchorterminal.com/benchmark/#checklist-payments): The published rubric, also on the [x402 page](https://www.anchorterminal.com/x402/). - 40, a machine payment protocol (x402, MPP or L402) on the tool's own endpoints. 10 to 30 when it covers only some endpoints or only goes through a third party, and the note says which. - 20, per-call or per-unit pricing published without a login. 10 for public plan-only pricing, 0 for "contact sales" or prices behind a login. - 20, a free tier or trial that doesn't need a card. - 20, autonomous onboarding, meaning an agent can get access without a person signing up in a browser (keyless use, x402, a programmatic key API). Payment platforms and agent wallets rarely charge for their own API over a machine protocol, so the first line has steps for them, and the highest one that applies counts. 40 when x402, MPP or L402 runs on all their own endpoints, 30 when it runs on part of their own API, 25 when their merchants can accept one, 20 for running a facilitator, 15 for paying as a buyer, and 0 when the only protocol is their own. Merchant acceptance sits above a facilitator because the platform's own customers can charge agents through it, while a facilitator settles for sellers who wire up the protocol themselves. The counter-argument (a facilitator does more for the protocol as a whole) has a point. Each note says which step applied. Open-source software you run yourself is scored on its hosted or paid option if it has one. A free, self-hosted package with nothing to buy gets 20, 20 and 20 for the last three lines, and 0 to 40 for the first only if it ships a payment protocol. ## 3. Agent ergonomics, 76 out of 100, up to 3.9 more on the total Why it scored 76: `history` returns only the stream position unless `limit` is set, and `channels` takes a glob `pattern`. No field selection (17). `history` pages with `limit`, `since` and `reverse`. `channels` has no pagination, which the docs state, and `presence` returns every client (13). Errors are numeric codes with a message, and a `temporary` flag that says a retry may succeed. By default the HTTP API answers 200 with the error in the body, and the `transport` error mode maps codes to HTTP statuses (15). `idempotency_key` on publish and broadcast, cached for five minutes per channel on the Memory and Redis engines only, plus `version` to drop stale updates (17). A publish needs `channel` and `data`. Official HTTP API libraries for Python, PHP, Go and Ruby (14). The checklist (https://www.anchorterminal.com/benchmark/#checklist-ergonomics): - 0 to 25, context cost. For MCP, the number and size of the tool definitions (25 for ten or fewer compact tools, 15 for 11 to 30, 5 for more than 30, plus up to 10 back for toolsets, dynamic loading or read-only subsets). For APIs, whether responses can be sized (field selection, limits, summaries). - 20, pagination, filtering and output-size controls. - 20, actionable, documented error responses, codes and messages an agent can recover from. - 20, idempotency or safe retries, and for MCP the `readOnlyHint` and `destructiveHint` annotations. - 15, sensible defaults, few required parameters, and official SDKs in at least two languages. Models are read for tool use, structured output, prompt caching, context length, batch and SDKs. Frameworks for how much code and how many defaults a tool-calling agent with MCP needs. ## 4. Transparency & trust, 56 out of 100, up to 3.9 more on the total Made of editorial 71, provenance 41. Why it scored 56: The server is open source under Apache-2.0. Centrifugo PRO is a separate closed build (30). No privacy policy or terms were found for the site or the open-source server. The server runs on the owner's machines, and the PRO Licence Agreement says the software sends the vendor no message content or personal data (15). Prometheus Summary metrics are marked deprecated since v6.8.1 with removal in v7, and one Redis option is marked deprecated, both without dates. No written deprecation policy (10). Anonymous usage statistics are sent once in 24 hours by default. The docs disclose this, describe the content as counters only and give the `usage_stats.disabled` opt-out, without a field list (16). The checklist (https://www.anchorterminal.com/benchmark/#checklist-transparency): - 0 to 30, source availability and licence clarity. 30 for open source under an OSI licence, 15 for closed with clear terms, 0 for unclear terms. - 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors). - 0 to 20, a deprecation policy or notices with dates. - 0 to 20, telemetry disclosed with an opt-out (local software), or subprocessors and data locations disclosed (hosted). The other half of Transparency and trust is the provenance score, computed from checked facts (below). The category score is the mean of the two. Provenance checks not met in full (half of this category, computed from checked facts): - Domain age: centrifugal.dev, registered 2021-05-27 (5 years) (11 of 15) - Endpoint on the vendor's domain: is not on centrifugal.dev (0 of 15) - Terms of service: not found (0 of 10) - Privacy policy: not found (0 of 10) - Status page: not found (0 of 10) - security.txt: not found (0 of 10) ## 5. Schema & documentation, 81 out of 100, up to 3.1 more on the total Why it scored 81: A Swagger 2.0 file with 40 operations and 162 definitions and an `api.proto` are in the repository, and the server can show the file at `/swagger` (25). llms.txt and a full Markdown corpus (10). Every operation has a summary and fields carry 874 descriptions, some with limits, such as the five-minute window of `idempotency_key`. Few say when not to call a method (14). Fields are typed, but the file has no enums, required fields are named only in description text, and `data` is any JSON by design (7). The docs give a curl example and a response for each method, the error codes and their HTTP and gRPC mapping. The Swagger file has no examples and lists only 200, 400, 401 and 500 (12). Docs are versioned from v3 to v6 with a migration guide for each major version and release notes on GitHub. The API itself has no version in its path (13). 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. Reliability, 86 out of 100, up to 2.8 more on the total Why it scored 86: Read with the local-software lines, because Centrifugo is a server the owner runs and has no hosted service. Prebuilt binaries for seven named platforms, a Docker image, deb and rpm packages, a Helm chart and a Homebrew tap (20). `test.yml` runs lint, unit and integration tests on Go 1.26.8 for every push, across 120 test files. We did not read the current pass state on `master` (20). The issues page shows 20 open issues, the newest opened on 7 October 2026 and eleven opened before 2025. Replies were not read (20). Release notes on GitHub separate improvements, fixes and breaking changes, and v6.9.0 flagged its breaking configuration change with migration steps. That change came in a minor release, and `CHANGELOG.md` stops at v5.1.1 (11). Version 6.9.7 (15). 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. ## 7. Maintenance & community, 86 out of 100, up to 1.2 more on the total Why it scored 86: v6.9.7 was tagged on 28 September 2026, 11 days before this check (30). Seven releases from 12 July to 28 September 2026 (20). 20 open issues with the newest two days old, and community rooms on Telegram and Discord. Reply times were not read (18). Official HTTP API libraries in four languages and client SDKs in six. Their release dates were not checked (10). Dependabot and Renovate are configured, releases build on Go 1.26.8 and CI runs on every push. The pass state was not read (8). The checklist (https://www.anchorterminal.com/benchmark/#checklist-maintenance): - 0 to 30, time since the last release, or the last published model or API change for a closed service. 30 within 30 days, 20 within 90, 10 within 180, 0 older. - 20, at least three releases or dated changelog entries in the last 90 days. - 0 to 25, responsiveness. Issues and pull requests answered on GitHub (the open issues and how recent the replies are). For closed services, a public changelog and a support or community channel that answers, 0 to 15. - 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models). - 10, package health, current dependencies and CI. Models are read for deprecation notice periods and model churn rather than release counts. ## Deductions Each comes off the total. A fixed and documented problem counts for less at the next check. - 2026-02-19 to 2026-09-29. Eleven security advisories were published on the repository in just over seven months, two rated Critical, three High, three Moderate and three Low. GHSA-j77h-rr39-c552 (12 March 2026, Critical) is an unauthenticated SSRF through dynamic JWKS URLs, patched in 6.7.0. GHSA-4r3x-2rwr-6w65 (18 August 2026, Critical) is an unauthenticated crash in the protobuf stream decoder. GHSA-9468-v6mj-fppw (7 July 2026, High) let clients forge headers forwarded to proxy backends, fixed in 6.9.0. All were published by the maintainer with fixes, so the deduction is reduced to 6 (https://github.com/centrifugal/centrifugo/security/advisories). ## 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: the pass state of the `test.yml` workflow on `master`. The Actions page was not read - unchecked: replies and reply times on the 20 open issues. Only the list page was read - unchecked: release dates and package names of the official HTTP API libraries and client SDKs, and Docker Hub pull counts. npm and PyPI download figures are left null - unchecked: which Centrifugo release first shipped the fix for GHSA-4r3x-2rwr-6w65. The advisory gives versions of the `centrifugal/protocol` package - No terms of service or privacy policy were found for centrifugal.dev or the open-source server, so `provenance.terms` and `provenance.privacy` are left out - The fields sent in anonymous usage statistics are not listed in the docs. The source in `internal/usage` was not read line by line - First release date not established from a 200-commit clone - Centrifugo PRO's price is by quote, and the PRO build was not graded ## Weaknesses - Two advisories in 2026 are rated Critical. One is an unauthenticated SSRF fixed in 6.7.0, the other an unauthenticated crash in the protobuf decoder - The HTTP server API has one key for every method, set in the config file, with no scopes or read-only key in the open-source edition - The docs accept the API key as an `api_key` URL query parameter, while recommending the `X-API-Key` header - By default the HTTP API answers 200 with an `error` object. HTTP status codes need the `transport` error mode - Anonymous usage statistics are sent every 24 hours unless `usage_stats.disabled` is set. Centrifugo PRO is priced by quote only ## 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. - Check every response for an `error` field, or send `X-Centrifugo-Error-Mode: transport` to get HTTP status codes. Batch and broadcast replies still need per-item checks - Send `idempotency_key` on publish and broadcast before retrying. It works per channel for five minutes, on the Memory and Redis engines only - Call `history` with `limit`. Without it the result holds only `offset` and `epoch`, and history is off until the namespace enables it - Send the key in the `X-API-Key` header, never as `?api_key=`. For gRPC set metadata `authorization` to `apikey ` - Avoid `channels` on a large deployment. It returns every active channel with no pagination, so pass `pattern` ## 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.