# Fix list: Check From Anchor Terminal's listing at https://www.anchorterminal.com/tools/check-payroll, the October 2026 research run, assessed 8 October 2026. Grade B, 67.5 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 Check: 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, 5 out of 100, up to 11.9 more on the total Why it scored 5: Read with the hosted rubric. No x402, MPP or L402 (0). No public price. checkhq.com/pricing returns 404 and the site asks visitors to get in touch (0). A sandbox exists under a Sandbox User Agreement, but the key is requested from Check through its sales contact, so partial credit (5 of 20). A person must contact sales and sign a partner agreement before production (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, 74 out of 100, up to 4.6 more on the total Why it scored 74: A partner API key in the `Authorization` header, one per environment, carrying the whole partner account. Check says unused keys expire, last activity is shown and IP restrictions are available. Integration partners get OAuth tokens limited to one employer and permission level, and hosted MCP accepts a Console OAuth login bound to that user's role. No per-key scopes were found (22 of 30). The MCP server has read-only mode, toolset and tool filters and optional confirmation for destructive tools. The API requires a succeeded preview before approval and gates SSNs and account numbers behind the separately approved Sensitive Data API. No read-only API key was found (15 of 20). The MCP server's instructions tell the model to confirm before writes that affect payroll or money movement. No guidance on untrusted text in records was found (6 of 15). GET /logs with filters, `X-Request-Log-Id`, Console API logs for 14 days, last activity per key, and sensitive-data access logged (14 of 15). SOC 2 Type II, a bug bounty through Federacy and regular penetration tests per the security page. No security.txt (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, 80 out of 100, up to 4 more on the total Why it scored 80: Read with the hosted lines and scored on the REST API under an embedded-payroll partnership. status.checkhq.com on Atlassian Statuspage lists four components (API, Console, Onboard, Payments) with incident history (20). Three incidents between 10 July and 8 October 2026. The Usage API was unavailable for about 100 minutes on 9 September, marked minor, and notices on 31 July and 4 September concerned deposits delayed at receiving banks, with no failed processing found on Check's side. Payroll endpoints were not affected in the window, so this reads as minor incidents only (20). An earlier incident on 6 June 2026, marked major, returned errors on payroll preview and approval for about four and a half hours and falls outside the 90 days. Limits with numbers, 10 to 80 requests a second by tier and 100 concurrent requests (15). 429 carries `Retry-After`, every response carries `RateLimit-Limit` and `RateLimit-Remaining`, and `X-Idempotency-Key` covers writes for 24 hours (15). No SLA is published, and the terms of service make no representation about uptime unless agreed in writing. The startups page claims 99.999 per cent historical uptime, which is a vendor claim (0). The REST API carries no beta label. The MCP server is labelled beta in its usage terms (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. Schema & documentation, 78 out of 100, up to 3.6 more on the total Why it scored 78: Each reference page embeds an OpenAPI 3.1 fragment for its one operation, with typed parameters and request bodies. No single downloadable specification was found, and response bodies are given as examples, not schemas (15 of 25). llms.txt indexes 367 lines of guides and reference, and each page is served as Markdown (10). Guides say when to use synchronous or asynchronous preview and how sandbox differs from production. Reference descriptions are one line on most operations read (14 of 20). Request bodies list required fields and formats such as dates. No enums were seen on the operations read, and query filters are plain strings (10 of 15). Curl examples in the quickstart and guides, an error envelope with `input_errors` and `field_path`, and a table of error types. Most reference pages document only the success response (14 of 15). Dated versions through `Check-Version`, an API upgrades page listing six versions since 2021-01-15, and a product changelog (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. ## 5. Transparency & trust, 62 out of 100, up to 3.3 more on the total Made of editorial 38, provenance 86. Why it scored 62: Closed service. Terms of service (3 September 2021), a Sandbox User Agreement and employer terms are published. The partner agreement that governs API use is not. The MCP server and CLI are MIT (17 of 30). The privacy policy (effective 18 April 2024) says information is kept as long as needed, with no periods. The security page mentions a DPA, of which no public copy was found. The documentation states what request logs redact and that Console shows 14 days (12 of 30). Breaking changes ship as dated versions with a written definition of what counts as breaking, and a migration guide covers the legacy Tax Filings API. No notice period or removal dates were found (9 of 20). No sub-processor list or hosting locations were found on the security, privacy or terms pages (0). 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 1 clause that costs points (7.1 of 10) - Privacy policy: read, states 7 of the 8 things a reader expects (9.3 of 10) - security.txt: not found (0 of 10) ## 6. Agent ergonomics, 83 out of 100, up to 2.8 more on the total Why it scored 83: `limit` sizes a page, and since version 2025-01-01 payrolls omit items unless `include_items` is set. No field selection. The MCP server exposes three meta-tools over 270 tools, with toolset filters and a read-only mode (20 of 25). Cursor pagination with `next` and `previous`, page sizes up to 100 or 500 on 19 endpoints, filters by company, workplace, status, ids and metadata (18 of 20). Typed error codes, `field_path` pointing at the invalid input, `X-Request-Log-Id` on every response and 409 `preview_superseded` on a stale approval (18 of 20). `X-Idempotency-Key` on writes with a stated 24-hour window and retry guidance, and the MCP server's source sets readOnlyHint, destructiveHint and idempotentHint on every tool (20). Few required fields on the operations read. No official SDK in a general programming language was found, only a Postman collection, a Python CLI and React Components (7 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. ## 7. Maintenance & community, 72 out of 100, up to 2.5 more on the total Why it scored 72: The product changelog's latest entry is 10 September 2026 and the MCP server repository's latest commit 7 October 2026 (30). Four dated changelog entries since 10 July (13 and 20 August, 9 and 10 September), and 109 commits to the MCP repository on 27 days in the same period (20). Closed service with a dated changelog, a help centre and developer support. The MCP repository is public with 10 open issues and pull requests, whose replies we couldn't read (10 of 15). No official SDK in a general programming language was found, and the official MCP registry returned no entry for Check. The CLI and MCP server are current (5 of 15). The repository runs lint and 465 tests in CI with a lock file. It has no tags and isn't published to PyPI (7 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. ## 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: live API behaviour. We had no sandbox key, so limits, errors and the preview flow are as documented, apart from a 403 `not_authenticated` response and `X-Request-Log-Id` header seen on an unauthenticated request - unchecked: whether a single OpenAPI file can be downloaded. docs.checkhq.com/openapi and /openapi.json answered 429 to our reader, and the reference pages embed one operation each - unchecked: the partner agreement, the DPA and any SLA in it. None is published - unchecked: per-key scopes or read-only API keys in Console, which sits behind a login - unchecked: the Federacy bug bounty terms, which sit behind a sign-in at federacy.com/check - unchecked: replies on the MCP repository's 10 open issues and pull requests, which aren't in the git clone - Changelog entries older than the first page (81 pages) were not read, so the count of entries in the last 90 days is a minimum of four ## Weaknesses - No public price and no self-serve signup. The documentation says to request an API key through Check's sales contact - An API key carries the whole partner account. No per-key scopes or read-only keys were found in the reviewed documentation - No official SDK in a general programming language was found. The CLI and MCP server install from a git clone, not from PyPI - No SLA is published. The terms of service make no representation about uptime unless agreed in writing - No sub-processor list, hosting location or published DPA text was found, and the privacy policy states no retention periods ## 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. - Use https://sandbox.checkhq.com with the sandbox key and https://api.checkhq.com with the production key. A key is not valid across environments - Preview a payroll before approving it, then pass `preview_started_at` to POST /payrolls/{id}/approve so a stale preview fails with 409 - Send `X-Idempotency-Key` on every write. Keys expire after 24 hours, so list the resource before retrying later than that - On 429 wait for `Retry-After`. The limit is 10 requests a second in sandbox and for partners paying fewer than 1,000 payees a month, with 100 concurrent requests - Connect the MCP server with `?read_only=true` or `X-MCP-Readonly: true` unless writes are intended. In production, approving a payroll moves money ## 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.