pgmcp CLAUDE.md

pgmcp CLAUDE.md is an instructions file for coding agents from sbamin/pgmcp. It costs 409 tokens per session, scanned A, original, MIT.

A set of project instructions for `pgmcp`, a read-only MCP server that provides guarded SQL access to a PostgreSQL database. It defines security rules, required documentation, configuration, and setup practices.

In plain words
What is it for?
Use it when working on `pgmcp`, especially its database grants, SQL safeguards, configuration, network binding, documentation, or setup and testing.
Why use it?
It helps keep database access read-only and prevents changes to security, query handling, network exposure, and sensitive-data protections.

Instructions file

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add instructions/sbamin/pgmcp/claude-md
Clone the repo
git clone --depth 1 https://github.com/sbamin/pgmcp

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for pgmcp CLAUDE.md

README.md
[![agentmods](https://agentmods.dev/badge/instructions/sbamin/pgmcp/claude-md.svg)](https://agentmods.dev/instructions/sbamin/pgmcp/claude-md)
Your own site
<a href="https://agentmods.dev/instructions/sbamin/pgmcp/claude-md"><img src="https://agentmods.dev/badge/instructions/sbamin/pgmcp/claude-md.svg" alt="Measured on agentmods" height="20"></a>
Per session 409 This file is loaded in full into every session.
When invoked 409 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00409 $0.00409
Opus 5 $0.00204 $0.00204
Sonnet 5 $0.00082 $0.00082
Haiku 4.5 $0.00041 $0.00041

Measured 4d ago against content hash d28f71591f00, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

pgmcp CLAUDE.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

CLAUDE.md · 38 lines

What it actually says

pgmcp

A read-only MCP server exposing a PostgreSQL database as a guarded SQL query gateway.

Read first

  • CONTEXT.md — glossary. Key point: "read-only" is layered; the server connects as pgmcp_ro, which is not pg_read_all_data.
  • SETUP.md — setup runbook, database-first.
  • docs/adr/0001..0006 — the decisions and why. Supersede an ADR explicitly rather than silently reversing it.

Invariants — never regress these

  • Grants are the security boundary. pgmcp_ro has no write grants and is never a member of pg_read_all_data. The AST guard and READ ONLY transaction are defense-in-depth, not the boundary.
  • Reject multi-statement SQL. The COMMIT; DROP … payload that escaped a READ ONLY envelope in a reference Postgres MCP server is an explicit test case.
  • Sensitive columns: expose them only through curated views that omit them; never grant the sensitive base tables.
  • Network: the server binds loopback; the reverse proxy is the only listener, with Origin/Host validated.
  • Prefer uv pip install under a local .venv/.

Config is the single source of truth

PGMCP_ALLOWED_SCHEMAS must mirror the grants in db/03_grants.sql. The guard, cost gate, and introspection all read the allowlist from SETTINGS.

Build & test

  • Role/view/grant DDL: db/01_role.sql, db/02_curated_views.sql, db/03_grants.sql.
  • Tests: pytest tests/ (no DB needed); PGMCP_DSN=… pytest tests/test_live.py for the end-to-end corpus.
Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 4d ago First seen · 38 lines · 409 tokens per session scan A d28f71591f00

Subscribe to this mod's changes

pgmcp CLAUDE.md is an instructions file published in the GitHub repository sbamin/pgmcp (1 stars, last pushed 1mo ago), licensed MIT. It adds 409 tokens to every session, about $0.0020 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.