README.md

Basic Memory

Ready-to-use Basic Memory environment — a local-first knowledge graph backed by Markdown notes, exposed as an MCP server for Claude Code, Cursor, ChatGPT, and other MCP clients.

Quick start

Include in your manifest:

[include]
environments = ["flox/basic-memory"]

What it provides

• basic-memory — the CLI
• bm — short alias for basic-memory

Both commands resolve the same Python entry point.

Configuration

Variable	Default	Purpose
BASIC_MEMORY_CONFIG_DIR	~/.basic-memory	Per-user state dir
FASTEMBED_CACHE_PATH	<data>/fastembed_cache	ONNX model cache

Semantic search

Bundled — fastembed and sqlite-vec are included in the Nix closure. The first time you enable semantic search, fastembed downloads the embedding model into FASTEMBED_CACHE_PATH.

Using with Claude Code

The minimal env exposes the CLI. To wire it as an MCP server for Claude Code, either:

1. Use the demo env (flox/basic-memory-demo), which composes this env, installs the flox-ai launcher, and writes a project-scope .mcp.json. First install Claude Code (e.g. flox install flox/claude-code) or use your own install, then run flox-ai launch claude to start it with this environment's skills and rules injected.
2. Or wire manually: claude mcp add basic-memory -- basic-memory mcp.

See also

• basic-memory-demo — interactive demo with Claude Code MCP wiring
• basic-memory upstream