Memo
Persistent Memory for AI Coding Agents
Your AI agent starts every session from zero. Memo fixes that — giving Claude Code, Cursor, and Codex searchable context that survives across sessions and projects.
The Problem
Every time you start a new session, your AI agent forgets everything — architecture decisions, coding patterns, bug fixes, project context. You repeat yourself constantly, and your agent never learns from past conversations.
The Solution
Memo gives your agent persistent, searchable memory stored locally in SQLite with vector embeddings. Decisions, patterns, and context carry over between sessions — your agent gets smarter with every conversation.
Key Features
100% Local
All data stays on your machine in SQLite. No cloud, no API keys, no data leaving your environment.
Semantic Search
Vector embeddings find relevant memories by meaning, not just keywords. Ask naturally, get accurate results.
MCP Compatible
Works with Claude Code, Cursor, Windsurf, Codex — any agent that supports the Model Context Protocol.
Cross-Project
Memories are shared across all your repositories. Patterns learned in one project carry over to the next.
Smart Dedup
Automatic semantic deduplication merges similar memories, keeping your knowledge base clean and relevant.
Dual Interface
Use as an MCP server for your AI agent, or query memories directly from the terminal with the CLI.
How It Works
Install
One command via Homebrew. Fast Go binary, minimal footprint.
Configure
Add Memo as an MCP server to your AI coding agent. One line of config.
Remember
Your agent stores decisions, patterns, and bugs automatically. Every session gets smarter.
Get Started in 2 Commands
Install via Homebrew:
brew install ybonda/tap/memoAdd to Claude Code:
claude mcp add --scope user memo -- memo serveThat's it. Your agent now has persistent memory across all sessions and projects.
Give Your AI Agent a Memory
Free, open source, and ready in under a minute.