Memento

Persistent semantic memory for AI coding agents. Save, recall, and search context across sessions — even after autocompact.

Works with Claude Code, OpenCode, Cursor, and Windsurf. Runs entirely on your machine — no cloud, no API keys, no data leaving your computer.

Quick Start

npx memento-memory setup

That's it. Your next session will automatically capture and recall context.

Requires Node.js >= 20. See Installation Guide for all methods and IDE-specific setup.

Features

• Auto-capture — hooks silently capture every meaningful tool call in the background
• Survives compaction — memories persist even when the context window is compressed
• Semantic recall — find relevant context by meaning, not exact keywords
• Hybrid search — vector (cosine similarity) + keyword (BM25), weighted scoring
• Smart memory — contradiction detection, importance scoring, entity extraction, relationship tracking
• 17 MCP tools — save, recall, search, forget, list, health, export, import, migrate, session management, project indexing, profiling, compaction, and more
• Cross-project search — find knowledge across all your projects
• Auto-tagging — heuristic classification (code, decision, error, architecture, config, dependency, todo)
• Sensitive data redaction — API keys, tokens, passwords stripped before storage
• Local-first — works offline with local embeddings (all-MiniLM-L6-v2), zero config
• Browser-compatible — runs in browsers and extensions via IndexedDB + fetch-based embeddings
• HTTP API + Graph UI — REST endpoints and D3 visualization of your memory graph

How It Works

Session active → PostToolUse hook captures context → queue
Session ends   → Stop hook triggers pipeline:
                 redact → tag → chunk → dedup → embed → store
Next session   → memory_recall restores context automatically

See How It Works for the full architecture walkthrough.

Documentation

sanathshetty444.github.io/memento

• Installation — setup for all IDEs, troubleshooting
• Quickstart — your first memory in 5 minutes
• Search Modes — vector, keyword, and hybrid explained
• Smart Memory — contradiction detection, importance scoring, entities
• Configuration — full config reference
• MCP Tools Reference — all 17 tools with parameters and examples
• CLI Reference — setup, status, teardown, serve
• Architecture — system overview, memory pipeline, search internals

Development

npm install && npm run build   # Build
npm test                       # Run tests
npm run lint                   # Lint
npx tsc --noEmit               # Type-check

Contributing

Contributions welcome! See the Contributing Guide.

License

AGPL-3.0