LOCAL · PRIVATE · MCP-NATIVE

Teach your coding agent once. Let your repository remember.

Persistent, repository-scoped memory for AI coding agents—corrections, rules, conventions, and decisions kept close to the code.

Install in two minutes Choose your client

Abstract illustration of a repository connected to local memory cards and AI context

Repository-aware

Each workspace gets a stable identity and a local SQLite database. Unrelated repositories do not share memories.

Explicit by design

MCP hosts decide when tools are called. Startup context, recall, remember, update, and archive are deliberate operations.

Open source

MIT licensed, inspectable, forkable, and open to contributions from developers, client authors, and documentation writers.

Start here

Give your agent a memory it can actually use.

Install the public package, add the stdio server to your MCP client, and ask the agent to call get_startup_context at the beginning of each session.

pip install repo-memory-mcp-ai
repo-memory doctor

# zero-install option
uvx --from repo-memory-mcp-ai repo-memory-mcp

How it works

01 · Start

Call get_startup_context() to load high-value repository guidance.

02 · Teach

Call remember_correction() or remember_rule() when a lesson should survive.

03 · Recall

Call recall_memory(query) when the current task needs project history.

Built for the ecosystem

Claude. Cursor. VS Code. Any MCP host.

Use the core server directly, install the companion VS Code extension, or use the browser companion to prepare research notes for your agent.

See client setup

Conceptual illustration of an AI coding agent, local repository memory, and returning context

Questions

Does it automatically read every AI conversation?

No. MCP hosts decide when tools are called. The agent explicitly chooses when to load, search, store, update, or archive memory.

Where does the memory live?

In .repo-memory/memory.db inside the detected repository. The directory is ignored by Git by default.

Can I contribute?

Yes. The project is MIT licensed. Add client examples, tests, docs, bug fixes, or new integrations through GitHub pull requests.