Basic Memory

What is Basic Memory?

Basic Memory is an MCP server that gives AI assistants persistent memory across sessions. It stores knowledge as Markdown files with a knowledge graph and semantic search, and works with Claude, Codex, Cursor, ChatGPT, and any MCP client.

How to use Basic Memory?

You can use Basic Memory via the cloud (30-second sign-up at basicmemory.com) or install locally with uv tool install basic-memory and configure your AI client with the command uvx basic-memory mcp. The cloud version handles hosting and sync; the local version runs on your disk.

Key features of Basic Memory

• Local-first, plain text Markdown on your disk
• Two-way sync between AI and human edits
• Real knowledge graph with observations and wikilinks
• Semantic search by meaning, not just keywords
• MCP-native, works with every major AI client
• Progressive tool discovery with behavior hints

Use cases of Basic Memory

• Continue AI conversations across sessions without losing context
• Build a shared team knowledge base that both humans and AI read and write
• Integrate persistent memory into coding assistants like Claude Code and Cursor
• Sync notes across devices using optional cloud or manual sync (Git, Syncthing)

FAQ from Basic Memory

What’s the difference between cloud and local?

Cloud requires no install, offers cross-device sync and mobile access, and costs $15/mo. Local is free forever, runs on your disk, and works air-gapped.

What are the dependencies for local install?

Local install requires Python and the uv tool. The command uv tool install basic-memory installs Basic Memory.

Where is my data stored?

In the cloud version, data is hosted on Tigris S3 with snapshots and backups. In the local version, data is stored as plain Markdown files on your disk.

What transports and authentication does Basic Memory support?

It supports stdio (for local desktop clients) and HTTPS (for cloud clients). Cloud authentication is via the web app; local installs require no authentication.

Is Basic Memory open source?

Yes, the source code is licensed under AGPL-3.0. Both cloud and local versions use the same open-source engine.