OpenMemory

A lightweight, local, shared memory system for AI tools and agents via MCP.

Why OpenMemory?

Most AI tools either forget everything, or use expensive RAG pipelines. OpenMemory provides a middle ground:

• Fast - BM25 search via OpenSearch (< 10ms)
• Cheap - No embedding API calls needed
• Local - All data stays on your machine
• Simple - Just memory_save and memory_search

See docs/DESIGN.md for architecture details.

Quick Start

1. Start infrastructure

docker compose up -d

2. Build MCP server

cargo build --release --bin openmemory-mcp

3. Configure your AI tool

Example for Claude Code (~/.claude/settings.json):

{
  "mcpServers": {
    "openmemory": {
      "command": "/PATH-TO-PROJECT/openmemory/target/release/openmemory-mcp",
      "env": {
        "DATABASE_URL": "postgres://openmemory:openmemory@localhost:5432/openmemory",
        "OPENSEARCH_URL": "http://localhost:9200",
        "REDIS_URL": "redis://localhost:6379"
      }
    }
  }
}

4. Use it

The AI now has two tools:

memory_save - Save important information

{
  "content": "User prefers TypeScript",
  "importance": 0.8,
  "tags": ["preference"]
}

memory_search - Find relevant memories

{
  "query": "TypeScript",
  "limit": 5
}

Tip: The AI won't automatically save everything. Before ending a conversation, ask: "Please save anything important from our discussion" to ensure key information is remembered for next time.

Development

pnpm install
pnpm turbo run dev

Seed test data

cd scripts
source venv/bin/activate
python seed-data.py --count 1000

License

MIT