Postgres MCP Pro by Crystal DBA

# Fix list: Postgres MCP Pro

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/postgres-mcp-pro, the October 2026 research run, assessed 1 October 2026. Grade F, 36.7 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 Postgres MCP Pro: 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, 26 out of 100, up to 13 more on the total

Why it scored 26: One database URI from `DATABASE_URI` or the command line, with whatever privileges its role has. The SSE and HTTP transports have no authentication and bind to localhost by default (10). Restricted mode parses every statement with pglast against an allowlist of statement types and functions, runs it in a read-only transaction and stops it after 30 seconds. But unrestricted is the default, every config example in the README uses it, and #178 (opened 6 June 2026) shows restricted mode reading server files through a function in the FROM clause. The fix (#200) is unmerged (10). The README discusses LLM-generated damage at length but says nothing about instructions hidden in table data, and rows reach the model unmarked (3). Restricted-mode queries are tagged `/* crystaldba */`, so they can be picked out in Postgres logs and pg_stat_statements (3). No SECURITY.md, and the #178 reporter says private advisories aren't enabled (0).

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. Reliability, 41 out of 100, up to 11.8 more on the total

Why it scored 41: Scored as a local stdio package. postgres-mcp on PyPI and the crystaldba/postgres-mcp Docker image, Python 3.12 or later stated. But PyPI's only current release, 0.3.0 from May 2025, asks for `mcp[cli]>=1.5.0` with no ceiling, and since MCP Python SDK 2.0.0 shipped on 28 July 2026 a fresh `uvx postgres-mcp` pulls it and fails with "No module named 'mcp.server.fastmcp'" (#187). The `<2.0` pin landed on main on 15 August and hasn't been released (10). CI runs ruff, pyright and pytest with a Postgres container on every push and pull request, with 25 unit test files. We couldn't see whether main passes (20). 37 open issues, among them the broken install, failures on custom schemas (#181) and two unmerged pull requests for connection-pool leaks (#177, #195) (6). Semver tags with no changelog (5). Pre-1.0 (0).

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. Maintenance & community, 8 out of 100, up to 8.1 more on the total

Why it scored 8: Last release 0.3.0 on 16 May 2025 (0). No release in the last 90 days. Main has a batch of 11 merges from 19 to 22 January 2026 and one commit on 15 August 2026 (0). 37 open issues and 35 open pull requests, a request for a release (#162) open since March 2026 and no maintainer reply on the security report. A commenter on #187 says the project is unmaintained since Crystal DBA's acquisition by Temporal, which we couldn't confirm (5). Not in the official MCP registry. A lookup for io.github.crystaldba/postgres-mcp returns 404 (0). Dependencies were refreshed on main in January, but the published package has the unbounded `mcp` dependency that now breaks it (3).

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.

## 4. Schema & documentation, 56 out of 100, up to 7.2 more on the total

Why it scored 56: FastMCP builds typed JSON Schema for all nine tools from Python type hints (25 less 5, since `hypothetical_indexes` is a list of free-form dicts) (20). No llms.txt. The README is Markdown on GitHub with a tool table (5). `explain_query` warns that `analyze` runs the query and `analyze_db_health` lists its checks, but most descriptions are one line ("List objects in a schema") and none says when not to use a tool (10). `method` is an enum. `object_type`, `health_type` and `sort_by` are free strings with valid values only in the prose, `limit` has no bounds and `execute_sql` gives `sql` a default of "all" (7). `explain_query` carries two worked examples. Errors come back as "Error: <Postgres message>" text (9). Semver tags, no changelog (5).

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

Why it scored 57: Nine tools with about 2,500 characters of descriptions, roughly 1,300 tokens of definitions by our estimate (23). `get_top_queries` takes a `limit`, but `execute_sql` returns every row and the index tools append a `_langfuse_trace` block by default unless `POSTGRES_MCP_INCLUDE_LANGFUSE_TRACE=false` (6). Restricted mode explains its refusals (`Only SELECT, ANALYZE, VACUUM, EXPLAIN, SHOW and other read-only statements are allowed`) and its 30-second timeout suggests simplifying the query. Errors arrive as ordinary text rather than flagged tool errors (13). Annotations exist on main since January 2026 but not in the released 0.3.0, and on main `explain_query` claims `readOnlyHint` though `analyze: true` executes the statement in unrestricted mode (5). Sensible defaults, at most one required parameter per tool. Python only, plus Docker (10).

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.

## 6. Payments & pricing, 60 out of 100, up to 5 more on the total

Why it scored 60: Free, self-hosted, nothing to buy, so 20 + 20 + 20. No payment protocol (0). The optional `llm` index method needs your own OpenAI key.

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.

## 7. Transparency & trust, 61 out of 100, up to 3.4 more on the total

Made of editorial 65, provenance 57.

Why it scored 61: MIT, copyright Crystal Corp. (30). Local software. The README says the experimental `llm` index method sends the schema and query plans to an LLM and needs an OpenAI key, but doesn't say what else leaves the machine or name the provider's terms (15). No deprecation policy, and SSE is still documented although the MCP specification replaced it (0). No telemetry in the code (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):

- Domain age: crystaldba.ai, registered 2024-11-25 (1 year) (3 of 15)
- Status page: not found (0 of 10)
- security.txt: not found (0 of 10)

## Deductions

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

- -8: 2026-06-06, a public issue showed restricted (read-only) mode can read arbitrary files on the database host with `SELECT * FROM pg_read_file('/etc/passwd')`, because the function allowlist checks only function calls outside the FROM clause. It needs a role with pg_read_server_files or superuser. Nearly four months later the issue has no maintainer reply and the fix (#200, opened 2026-08-16) is unmerged (https://github.com/crystaldba/postgres-mcp/issues/178; https://github.com/crystaldba/postgres-mcp/pull/200).

## 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: whether the crystaldba/postgres-mcp Docker image on Docker Hub was rebuilt after 0.3.0, and whether it still starts
- unchecked: whether CI passes on main
- Whether Crystal DBA was acquired by Temporal and whether anyone still maintains the project, as one commenter on #187 claims

## Weaknesses

- No release since 0.3.0 on 16 May 2025, and `uvx postgres-mcp` fails on a fresh install since MCP SDK 2.0
- Unrestricted is the default and every README example uses it
- Open restricted-mode bypass (#178) reads server files when the role has pg_read_server_files or superuser
- No security policy, no maintainer reply on the security report, 37 open issues and 35 open pull requests
- `execute_sql` has no row limit, and the released package carries no tool annotations

## 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.

- Launch with `uvx --with 'mcp<2' postgres-mcp`. Plain `uvx postgres-mcp` now fails with "No module named 'mcp.server.fastmcp'"
- Pass `--access-mode=restricted` explicitly. The default is unrestricted
- Connect with a role that lacks superuser and pg_read_server_files. Restricted mode alone doesn't stop server file reads
- Put `LIMIT` in every `execute_sql` query. The server returns every row
- Don't set `analyze: true` on `explain_query` for writes in unrestricted mode. It runs the statement

## What the review panel asked for

- enums for object types
- annotations in a release
- merge and release #200
- default to restricted mode

## 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.