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.
npx agentmods add instructions/devopam/mcpg/claude-mdgit clone --depth 1 https://github.com/devopam/MCPgWhat 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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.01295 | $0.01295 |
| Opus 5 | $0.00647 | $0.00647 |
| Sonnet 5 | $0.00259 | $0.00259 |
| Haiku 4.5 | $0.00129 | $0.00129 |
Grade A, and why
MCPg 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 2d 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.
How it starts
The opening of the file, as written. The whole thing — 94 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md — MCPg project memory
Durable facts and conventions for working in this repo. Kept short so it stays true; verify anything here against the code before relying on it.
What MCPg is
A production-grade PostgreSQL MCP server. Single-process, async
(asyncio), Python 3.12–3.14. 254 MCP tools (read-only mode exposes
a subset). Ships to PyPI (mcpg), GHCR, the MCP registry, Smithery, and
a read-only HF Spaces demo.
Golden rules
- Verify before you write — especially docs. Every claim in a doc must be checked against the code, the CLI, or a live endpoint, never from memory. This is a hard project rule.
- Generated beats hand-maintained. If a catalogue can drift, it's generated and guarded by a test. Don't hand-edit generated blocks.
- Commit per logical slice, push often. Big multi-batch changes have been lost to context resets — small commits survive.
- Never hand-edit
src/mcpg/_vendor/(see below).
Source-of-truth map (don't duplicate these — point at them)
- Version:
src/mcpg/__init__.py__version__.pyproject.tomlreads it dynamically via hatchling (dynamic = ["version"]). The MCPB bundle pins are synced bypackaging/mcpb/sync_version.py;tests/unit/test_mcpb_bundle.pyenforces the match. A release bumps exactly one line. - Tool surface:
tests/contract/tool_surface.snapshot.json(256 tools) is the contract.about.py'stool_count = len(tool_names). Two contract snapshots (surface + return shapes) — regenerate withMCPG_REGENERATE_TOOL_SNAPSHOT=1/MCPG_REGENERATE_TOOL_RETURN_SHAPES=1and commit the diffs. - Doc reference tables: the tool index (
docs/tools.md) and module map (docs/architecture.md) are generated bytools/generate_doc_tables.pyand guarded bytests/contract/test_doc_tables.py. Regenerate and paste between the<!-- BEGIN/END generated -->markers; don't hand-edit. - Access modes:
src/mcpg/policy.py.read-only(reads) /restricted(reads + DML writes — the "safe read-write" tier) /unrestricted(everything).DDL/SHELL/LISTEN(andMIGRATE, which piggybacks onDDL) each additionally needMCPG_ALLOW_*inunrestrictedmode. - NL→SQL providers: 19 built-in via the
_PROVIDERSregistry insrc/mcpg/nl2sql.py(single source; deviating key env vars:HF_TOKEN,GITHUB_TOKEN,DEEPINFRA_TOKEN). The provider table indocs/user-guide.mdmirrors it.
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.
- 2d ago First seen · 94 lines · 1,295 tokens per session scan A c951a42f7cc4
MCPg CLAUDE.md is an instructions file published in the GitHub repository devopam/MCPg (9 stars, last pushed 7d ago), licensed MIT. It adds 1,295 tokens to every session, about $0.0065 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.
Other instructions, from other repositories
sqlmate-mcp CLAUDE.md
Claude Code instructions for adamrpostjr/sqlmate-mcp, covering claude.md, commands, start the mcp server, build the frontend (svelte → /public) and frontend dev server (proxies /api to localhost:4737).
querybridge-mcp CLAUDE.md
Claude Code instructions for MahmoudHassanMustafa/querybridge-mcp, covering claude.md, quick orientation, commands, release flow and hard rules.
Connect_MSSQL_MCP AGENTS.md
Instructions for XingCheng3/Connect_MSSQL_MCP, covering repository guidelines（仓库贡献指南）, 项目结构与模块, 构建、运行与开发, 代码风格与命名 and 测试指南.
db-tools-mcp AGENTS.md
AGENTS.md instructions for urjeetpatel/db-tools-mcp, covering agents.md — db-tools-mcp, what this server does, setup checklist (one-time), tool reference and discovery.
db-tools-mcp copilot-instructions.md
Copilot instructions for urjeetpatel/db-tools-mcp, covering project guidelines, code style, architecture, build and test and conventions.
codex AGENTS.md
AGENTS.md instructions for openai/codex, covering rust/codex-rs, the codex-core crate, code review rules, crate api surface and model visible context.