Grist

# Fix list: Grist

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/grist, the October 2026 research run, assessed 8 October 2026. Grade D, 50.1 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 Grist: 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. Reliability, 39 out of 100, up to 12.2 more on the total

Why it scored 39: Graded on hosted Grist's REST API and MCP server at docs.getgrist.com, with the hosted lines. No status page was found. status.getgrist.com redirects to a signup form, and the site footer and help centre link to none (0). With no page there is no readable incident history (5). Limits are published with numbers, 3,000 calls a month on Free, 40,000 and 60,000 per document per day on Pro and Business, 5 requests a second per document on Free and 10 concurrent requests per document on every plan (15). The limits page says an eleventh concurrent request gets 429 and that a client sending one request at a time won't hit it. No Retry-After header or backoff guidance was found, and `PUT` on `/records` with `require` gives a retry-safe write (9 of 15). The pricing page names SLAs among Enterprise services and publishes no terms (0). The REST API is generally available and the MCP server, live since June 2026, carries no beta label (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.

## 2. Payments & pricing, 30 out of 100, up to 8.8 more on the total

Why it scored 30: Graded on hosted Grist. No x402, MPP or L402 found (0). Plan prices are public, Pro at $10 and Business at $30 a user a month ($8 and $24 billed yearly), with no per-call price (10). A Free plan with 5,000 rows a document and 3,000 API calls a month. The pricing page doesn't mention a card for it, and we didn't complete a signup (20). A person signs up in a browser and creates the key or approves the OAuth consent screen. A client can register itself by Client ID Metadata Document, but that doesn't remove the sign-in (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.

## 3. Agent ergonomics, 57 out of 100, up to 7 more on the total

Why it scored 57: The MCP docs list 34 tools, more than 30, which earns 5. We added 8 because OAuth scopes limit a connection to reading and the client can ask for fewer scopes (13 of 25). List records takes `filter`, `sort` and `limit` with no offset or cursor, and the read-only `/sql` endpoint takes a SELECT with parameters and a timeout (15 of 20). OAuth errors follow the standard `error` and `error_description` shape, and the MCP docs explain the missing scope and doc not found errors. The REST reference documents few error bodies (10 of 20). `PUT` on `/records` adds or updates by `require` keys, with `onmany`, `noadd` and `noupdate` to control it. No idempotency key was found, and we couldn't read whether the MCP tools carry readOnlyHint or destructiveHint (10 of 20). Few required parameters. Grist Labs maintains `grist-api` for JavaScript and Python, last published in February 2022 and October 2024 (9 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.

## 4. Security & auth, 62 out of 100, up to 6.7 more on the total

Why it scored 62: OAuth 2.0 authorisation code flow with PKCE, seven scopes, grants limited to chosen sites, workspaces or documents, access tokens of 1 hour, refresh tokens of 60 days and revocation per app. The alternative API key carries its owner's full access, with one key per account, and travels only in the `Authorization` header (28 of 30). A connection can hold `doc:read` alone, and schema changes need a separate scope. No server-side confirmation for `remove_records` or `remove_table` was found, and the docs leave approval to the client (15 of 20). Cells can hold text written by others. The MCP docs warn that anything read is visible to the client's AI provider, and no prompt-injection guidance was found (3 of 15). The Authorised apps page shows each app's scopes, resources and last use, and document history records changes. Audit log streaming is an Enterprise capability (7 of 15). SECURITY.md gives security@getgrist.com with acknowledgement in three working days, and 10 advisories are published on GitHub. No security.txt or bug bounty, and the docs say hosted Grist has no SOC 2 or ISO 27001 audit (9 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.

## 5. Schema & documentation, 63 out of 100, up to 6 more on the total

Why it scored 63: The OpenAPI 3.0.0 file in the public docs repository has 101 paths and 120 operations and is rendered at support.getgrist.com/api. The MCP server is in the proprietary full edition and its endpoint answers 401 without a token, so we couldn't read its tool schemas (22 of 25). No llms.txt on www.getgrist.com or support.getgrist.com. The docs source is Markdown in a public repository (3 of 10). Parameter descriptions are detailed, but the four `/records` operations carry a summary and no description. The MCP docs group 34 tools with one line each (13 of 20). Path and query parameters are typed with 16 enums. Record fields depend on the table, `filter` is a JSON string and `/apply` takes arrays in Grist's internal action format (8 of 15). The file has examples throughout, but only 36 of 120 operations document an error response (400, 403 or 404) and none documents 429 (8 of 15). The file is versioned 1.0.1 with no version in the path. GitHub release notes and monthly newsletters record API changes, and four `/data` operations are marked deprecated (9 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. Transparency & trust, 61 out of 100, up to 3.4 more on the total

Made of editorial 50, provenance 71.

Why it scored 61: The core in gristlabs/grist-core is Apache-2.0. The MCP server, the OAuth server and audit log streaming are in the proprietary full edition, which is what hosted Grist runs (24 of 30). The privacy policy is dated 1 April 2019 and names no retention periods. The data security page adds AWS in the United States, encrypted S3 storage, a 30-day trash and logs kept only as long as needed, and the two don't conflict. No DPA was found (12 of 30). No deprecation policy was found, and the terms allow changes and withdrawal without notice. The OpenAPI file marks the four `/data` operations deprecated, and release notes list API changes (5 of 20). AWS in the United States is stated, OpenAI is named for the AI Assistant, and the privacy policy names Amazon, Google and MailChimp. No sub-processor list was found (9 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 3 of the 7 things a reader expects, and has 1 clause that costs points (4.6 of 10)
- Privacy policy: read, states 3 of the 8 things a reader expects (6.3 of 10)
- Status page: not found (0 of 10)
- security.txt: not found (0 of 10)

## 7. Maintenance & community, 78 out of 100, up to 1.9 more on the total

Why it scored 78: Release v1.7.20 was tagged on 28 September 2026, 10 days before the check (30). Four releases in the last 90 days, v1.7.17 on 29 July, v1.7.18 on 28 August, v1.7.19 on 5 September and v1.7.20 (20). The repository shows 636 open issues, with 15 updated between 9 September and 5 October 2026, and the community forum had posts on the day of the check. We didn't read maintainers' reply times (15 of 25). The only Grist entry in the official MCP registry is a third party's. The official `grist-api` clients exist but date from February 2022 (npm 0.1.7) and October 2024 (PyPI 0.1.1) (5 of 15). The CI workflow runs lint, server, browser and Python test groups on every push to main, and v1.7.20 updated a dependency for a security fix. We didn't read the current CI result (8 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.

## Deductions

Each comes off the total. A fixed and documented problem counts for less at the next check.

- 2026-09-13. GHSA-9x4j-4rw5-3vmq, critical (CVSS 10.0), remote code execution through prototype pollution from an imported crafted document, affecting Docker images before 1.7.19 and fixed in 1.7.19. Four more advisories were published in the last 12 months, among them GHSA-7xvx-8pf2-pv5g (CVE-2026-24002, critical, 21 January 2026) on the pyodide sandbox option, fixed in 1.7.9. All five are fixed and published and none says whether hosted Grist was affected, so the deduction is reduced (https://github.com/gristlabs/grist-core/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 MCP server's tool definitions, input schemas and annotations. The server is in the proprietary full edition and `https://docs.getgrist.com/api/mcp` answers 401 without a token. The count of 34 tools is from the docs page, and the September 2026 newsletter names 6 layout tools the page doesn't list
- unchecked: whether signup on hosted Grist asks for a card, an email confirmation or a captcha. We didn't create an account, and the pricing page doesn't mention a card for the Free plan
- unchecked: maintainers' reply times on GitHub issues and the current CI result. The GitHub API refused us, and the counts of 11,900 stars and 636 open issues are read from the repository pages
- unchecked: whether hosted Grist was affected by the five advisories of the last 12 months. None of the advisories read says
- No status page, published SLA terms, DPA, sub-processor list, retention period, security.txt, bug bounty, llms.txt, Retry-After header, idempotency key, deprecation policy or prompt-injection guidance was found in the reviewed pages
- The terms page is an End-User Licence Agreement with no date. It covers Grist Labs' products, software, services and websites, so it is recorded as the governing terms. The privacy policy is dated 1 April 2019
- The pricing page names SLAs among Enterprise services. No points were given because no terms are published
- List records has `limit` with no offset or cursor in the OpenAPI file. Paging beyond that needs the `/sql` endpoint
- The lead was right on vendor, URL and the REST interface. It missed the MCP server at `https://docs.getgrist.com/api/mcp` and OAuth, both launched in June 2026, and the legal entity is Grist Labs Inc.

## Weaknesses

- No status page was found. status.getgrist.com redirects to a signup form, and the site footer and help centre link to none
- An API key carries its owner's full account access, and each account has one key, so it can't be limited to a document or revoked per integration
- The data security page says hosted Grist has no SOC 2, ISO 27001, HIPAA or GDPR certification, and no DPA, sub-processor list or security.txt was found
- The MCP server and OAuth server are in the proprietary full edition, so the Apache-2.0 community edition has neither and the tool definitions aren't public
- Five security advisories were published in the last 12 months, two rated critical, the latest on 13 September 2026 with CVSS 10.0
- The Free plan allows 3,000 API calls a month across a site, with REST and MCP calls sharing the pool

## 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.

- Connect through OAuth, not an API key. Ask for `doc:read` alone for reading, and have the user pick Selected resources on the consent screen
- Keep `doc.schema:write` off unless needed. The docs say formula editing lets its holder read any data in the document whatever the access rules
- Send requests for one document one at a time. The eleventh concurrent request gets 429, and no Retry-After header is documented
- Use `PUT /api/docs/{docId}/tables/{tableId}/records` with `require` to add or update by key, so a retried write doesn't create a duplicate row
- Use the `/records` endpoints, not the deprecated `/data` ones, and split large writes to stay under the 1 MB body limit

## 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.