repo-memory-mcp

mcp-name: io.github.Umarjaum/repo-memory-mcp

Persistent memory for AI coding agents

AI coding agents are powerful, but they forget repository-specific corrections, conventions, and architectural decisions. repo-memory-mcp gives an MCP-compatible coding agent a persistent, local, repository-aware memory layer.

Teach your coding agent once. Let your repository remember.

Why it exists

MCP hosts decide when tools are called; MCP does not automatically intercept every conversation. This project therefore exposes explicit tools for startup context, searching, storing corrections and decisions, updating stale guidance, and archiving obsolete memories.

Features

• Repository-scoped SQLite storage in .repo-memory/
• FastMCP tools for corrections, rules, decisions, search, and startup context
• Deterministic local relevance scoring; no pretend semantic search
• Local-first, API-key-free, cloud-free, telemetry-free operation
• Conservative secret detection for explicitly submitted content
• Import/export and audit-friendly archiving
• Python 3.10+, Pydantic v2, pytest, standard-library-first design

Install

For local development:

pip install -e .[dev]
repo-memory doctor

After publishing the package to PyPI:

pip install repo-memory-mcp-ai
uvx repo-memory-mcp-ai

The published-package commands should only be used once the package is publicly available.

Configure an MCP client

The examples in examples/ use the common stdio configuration shape. Clients can differ; follow the client’s documented MCP configuration format.

Tool workflow

1. New session: call get_startup_context().
2. Developer correction: call remember_correction(topic, lesson).
3. Relevant work: call recall_memory(query).
4. Obsolete guidance: call forget_memory(memory_id).

Your memories stay on your machine. The implementation does not require API keys, cloud accounts, remote databases, external embeddings, or telemetry. The claims apply to this codebase as shipped; your operating system, package manager, backups, or MCP host may have separate behavior.

Development

pytest
repo-memory init
repo-memory list
repo-memory search "pydantic models"

Roadmap

Future work may improve ranking, add confidence feedback, provide optional local embeddings, generate project context, and add a memory inspection UI. Team/shared memory remains explicitly opt-in.

License

MIT. See LICENSE.