# Fix list: PlanetScale MCP Server From Anchor Terminal's listing at https://www.anchorterminal.com/tools/planetscale-mcp, the October 2026 research run, assessed 9 October 2026. Grade B, 66.4 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 PlanetScale MCP Server: 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, 15 out of 100, up to 10.6 more on the total Why it scored 15: No x402, MPP or L402 in the MCP docs, the pricing page or the source (0). Monthly prices for every cluster size are public without a login, as a Markdown page and as a read-only SQL catalogue, from $5 a month for a single-node Postgres PS-5. Storage, backup and egress rates were not on the page read, and there is no per-call price (15). No free tier or trial was found, and the billing docs say all plans need a valid payment method on file (0). A person signs up in a browser and authorises OAuth or creates a service token in organisation settings (0). Total 15. 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. Reliability, 63 out of 100, up to 7.4 more on the total Why it scored 63: Graded as a hosted service. Status page at planetscalestatus.com with per-region component uptime for July to October 2026 (20). The page shows no open incident and its lowest figures are 99.71 per cent for Query Insights, 99.97 per cent for Postgres in us-east-1 and 99.98 per cent for Dashboard and API Requests. The incident entries are script-drawn and went unread, so the record is scored from the uptime figures as minor only (20). No rate limit with numbers was found for the MCP server or the API (0). The OpenAPI file lists a 429 response that says to reduce request frequency, with no Retry-After or backoff guidance, and no idempotency key for writes. The 50-second query limit is documented (3). The SLA commits to 99.99 per cent for single-region and 99.999 per cent for multi-region clusters with service credits, and excludes single-node clusters and beta functions (10). The docs give the MCP server no beta or preview label and the registry entry is 1.0.0 (10). Total 63. 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. ## 3. Agent ergonomics, 74 out of 100, up to 4.2 more on the total Why it scored 74: 25 tools (15). The insights-only endpoint drops the two SQL tools, and OAuth scopes limit which tool groups a client is granted (5). Insights, error, tag and log tools take `limit` with documented maxima and report `truncated` or `has_next`, and the logs tool pages. `planetscale_execute_read_query` returns every row with no cap (15). Error texts say what to do next, for example the safe-migrations message with two ways forward, the `postgres_rls_active` warning on empty results and the full-access message on a payment 403 (15). The 13 source tools set `readOnlyHint` and `destructiveHint`, with a test that fails on a missing annotation. Annotations on the 12 generated tools were not read. No idempotency for writes was found, though the payment tool tells the model not to start a second setup while one is pending (15). Sensible defaults (replica routing, 24-hour windows, curated metrics), but three identifiers are required on nearly every call. Official SDKs do not apply to a hosted MCP server and none were checked (9). Total 74. 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, 77 out of 100, up to 4 more on the total Why it scored 77: OAuth with each client registered as an OAuth application, scopes that set no, read-only or full access per organisation or per database, a separate payment-method scope, and service tokens with per-permission grants for headless use. Each query runs on a database credential created for the call and deleted after it (30). Read queries use a reader password on Vitess and a `pg_read_all_data` role on Postgres that does not bypass row-level security. An insights-only endpoint has no SQL tools, and since 8 October 2026 administrators can cap members' grants at none, read-only or read and write. TRUNCATE and UPDATE or DELETE without WHERE are blocked. DELETE and DDL need `confirm_destructive`, which the model sets itself, and the check is a regular expression on the start of the statement. Write queries run as the `postgres` role or a Vitess admin password. Read and write is the default (16). Query results return database content to the model. The docs warn about write access to production, and no prompt-injection guidance or result wrapping was found (4). Organisation audit log and security log on all plans, every MCP query tagged `source=planetscale-mcp` in Insights, and a list of installed OAuth apps with revoke (13). Vulnerability disclosure policy with security@planetscale.com, SOC 1 Type 2 and SOC 2 Type 2 with HIPAA, PCI DSS for PlanetScale Managed. security.txt returns 404 and no bug bounty was found (14). Total 77. 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, 82 out of 100, up to 2.9 more on the total Why it scored 82: The 13 tools in planetscale/mcp-server have typed zod input schemas, and the other 12 are generated from the public OpenAPI file at planetscale.com/docs/openapi.yaml, 304 operations (25). llms.txt and a Markdown twin of every docs page (10). The custom descriptions state purpose, limits and when to ask the person, for example the 50-second limit, row-level security on Postgres reads and the Checkout sequence for the payment tools. The generated tools are one-line summaries (17). Enums for log levels and summary fields, clamped limits and required fields. `organization`, `database` and `branch` are free strings and SQL is free text by nature (12). The docs give example prompts and a permissions table per tool. Tool errors are plain text beginning `Error:` with the API status code (10). The product changelog is dated and carries MCP entries, but the server has no version history of its own. The package is 0.0.0, the repository has no tags or changelog, and the registry entry has stayed at 1.0.0 since 15 April 2026 (8). Total 82. The tool definitions were read from the source and the docs, not from the live server, which answers 401 without a token. 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, 70 out of 100, up to 2.6 more on the total Made of editorial 66, provenance 73. Why it scored 70: The tools PlanetScale maintains by hand are public under Apache-2.0. The hosted server, its OAuth layer and the 12 generated tools are closed, under the Terms of Service of 25 August 2026 (22). Privacy Policy of 23 July 2026, a DPA covering all plans and a data-locality statement in the security docs. Retention is described without periods, and the DPA was not read (20). The Terms of Service promise at least six months' notice before material functionality is discontinued, with exceptions. The docs say the local `pscale mcp` server has been removed, with no date (14). Customers choose the cloud region and the security docs refer to sub-processors, but the list was not read. The MCP endpoint's response headers name Gram, Speakeasy's hosting platform, and no page read says how tool calls are handled there (10). Total 66. 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): - Endpoint on the vendor's domain: mcp.pscale.dev is not on planetscale.com (0 of 15) - Terms of service: read, states 6 of the 7 things a reader expects (9.1 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) ## 7. Maintenance & community, 80 out of 100, up to 1.8 more on the total Why it scored 80: The changelog's newest MCP entry is 8 October 2026 (organisation MCP query access), and the repository's last commit is 17 September 2026 (30). Commits on 13 days since 11 July added Postgres logs, query errors, query tags, payment tools, annotations and Neki support, and the changelog has MCP or OAuth entries on 28 September and 8 October (20). Closed hosted service with a dated public changelog, a support desk and a Discord community. The repository's issues and pull requests were not read (10). Listed in the official MCP registry as `io.github.planetscale/mcp-server`, a GitHub-verified namespace, version 1.0.0 published 15 April 2026 (15). Dependencies are current (`@modelcontextprotocol/sdk` ^1.30, zod 4) and the repository has unit tests, but no CI workflow is published (5). Total 80. 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: the tool definitions as the live server serves them. `initialize` answered 401 without a token, so `tools/list` was not sent. The 13 source tools were read from the repository and the 12 generated ones from the docs list and the OpenAPI file - unchecked: annotations and input schemas of the 12 generated tools as served - unchecked: rate limits. None in the MCP docs, the service-token page or the OpenAPI file. The API reference pages were not read, to keep to the page budget for planetscale.com - unchecked: incident entries on planetscalestatus.com, which are script-drawn. The reliability record rests on the component uptime figures for July to October 2026 - unchecked: the DPA, the acceptable use policy, the sub-processor list and trust.planetscale.com - unchecked: the audit log docs page. The audit and security logs are taken from the plan table on the security page - unchecked: open issues and pull requests on planetscale/mcp-server, and its star count - unchecked: storage, backup and egress rates, which were not on the pricing page read - Whether Speakeasy, whose Gram platform serves the endpoint per the response headers, is a named sub-processor, and whether tool inputs and results are logged there - When the local `pscale mcp` server was removed and whether users were given notice - Whether the hosted server shows the DDL and DELETE confirmation to the person through the client. In the source it is a flag the model sets - The weakness about the write-query check comes from reading the regular expressions in the source. It was not tested against the service - PlanetScale's Website Terms of Service (27 December 2021) forbid any robot, spider or automatic process retrieving the site. robots.txt allows every path and says ai-input=yes. No deduction, and it matters before any probe is run - 17 requests went to planetscale.com, two over the brief's fifteen, to read the Terms of Service and the OpenAPI file - The lead was right about the endpoint, the 25 tools, OAuth and service tokens, and the removal of the local server ## Weaknesses - DELETE and DDL confirmation is a `confirm_destructive` flag the model sets, not a prompt the client shows the person - The write-query check matches the start of the statement by regular expression, so a statement behind a leading comment or a type it does not list is not matched, per the source - No rate limit for the MCP server or the API was found. The OpenAPI file lists 429 with no figures - No free tier. The billing docs say every plan needs a payment method on file - The hosted server and 12 of its 25 tools are closed. The repository has no tags, changelog or CI workflow - PlanetScale's website terms forbid robots and other automatic retrieval of the site. 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. - Use `https://mcp.pscale.dev/mcp/planetscale-insights-only` when the task needs no SQL. It omits both query tools - Pass `organization`, `database` and `branch` on every query call. Record them in `AGENTS.md` to skip the listing calls - Show the person the exact DELETE or DDL statement and get approval before setting `confirm_destructive: true` - Keep queries under 50 seconds and add your own `LIMIT`. The query tools cancel at 50 seconds and return every row - For a service token send `Authorization: Bearer pscale_tkn_...` with the secret only. The `id:secret` form is rejected ## 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.