Give your AI agents a git repo as a brain. Local-first persistent memory for OpenCode, Claude Code, Cursor & Codex — MCP server + auto-capture plugin. Auditable, branchable, zero-dep.
View on GitHub ↗Homepage ↗Give your AI agents a git repo as a brain.
mnemo is persistent, local-first memory for AI agents (OpenCode, Claude Code, Cursor, Codex…). Agents remember by writing Markdown files into a git repo — so you get auditability, branching, snapshots, and rollback for free.
LLM agents are stateless: every session forgets everything. You repeat decisions, preferences, and context over and over — burning tokens and letting the agent re-open settled questions.
- Cross-session onboarding — a new session reads yesterday's memories and starts knowing. - Multi-agent shared reality — the planner writes decisions, the implementer reads them. - Auditability — git log shows exactly what the agent knew, and when. (The requirement the agent community keeps asking for in 2026.) - Token economy — recall the 3 memories that matter instead of re-reading the whole project. - 100% local — memory is a folder on your machine. No cloud, no lock-in.
Every memory is a human-readable Markdown file with metadata, and every mutation is a git commit. Search is classic BM25 plus a recency/importance rank — zero dependencies, ~90% recall@1 on synthetic corpora.
The OpenCode plugin makes memory automatic: at the end of a session it summarizes what was decided/learned and stores it; at the start of the next session it seeds the agent with the recent highlights. You don't maintain memory — it happens.
And drop prompts/AGENTS.md into your project so agents know to call recall before work and remember after decisions.
From the project README.
Add the radar badge to your README — it shows your project was picked up by MCP Radar and links to this page:
[](https://mcp.liqiwa.com/s/JoaquimLegal--mnemo.html)
Open plain-text file format for AI memory. Your assistant's long-term memory as .dai files on your disk: readable by Claude, GPT, Gemini, Cursor, local models and grep (all LLM models work). MCP serve
menot-you/n-memoryMemory your agent can trust — every answer carries its source; when it doesn't know, it says so. Hermetic, local, MCP.
2672243194/agentbrainLocal-first long-term memory for AI agents - plain Markdown vault + thin MCP server. CJK-aware BM25, token-efficient by design.
NiazMorshed2007/jev-reviewLocal-first MCP plugin for continuous software-quality review by AI coding agents, powered by Jev.
LilithGames/a2waveTurn the agent CLIs you already use into shared services your whole team can call. Natural-language-driven agent orchestration for Claude Code, Cursor Agent, and OpenAI Codex.
klarluft/gitwarren-appA cross-platform desktop app for doing local code reviews of your own git repositories. Single user, single machine, no server, no account.
The top new MCP servers of the week, every Monday. No spam, unsubscribe anytime.