Install the Claude Code plugin, capture once — then come back to a populated resume card. No account, and capture → resume needs no key at all; the loopback backend keeps everything on your machine.
No venv, no path editing — the plugin's MCP server bootstraps its own environment on first run.
/plugin marketplace add anthonysuherli/br8n
/plugin install br8n@br8n
Then reload the session so the MCP server connects.
Capture and resume work with no key at all — snapshots are stored without embeddings.
/br8n:capture · /br8n:pickup
Semantic search needs an embedding key — AI_GATEWAY_API_KEY or OPENAI_API_KEY. The explore / gap-fill pipeline needs that plus TAVILY_API_KEY for web search. Without them, capture and resume still work.
Reports your Python version, which backend tier is configured, whether sqlite-vec loads, whether the DB path is writable, and whether the embedding and explore keys are present. Anything missing is named explicitly.
python -m br8n.api.main --check
The blessed launcher binds 127.0.0.1 and turns auth off — safe because it never leaves loopback. The MCP entry point serves any MCP-compatible client or agent — exposing br8n_capture, br8n_resume, br8n_explore.
BR8N_BACKEND=local python -m br8n.api.main
BR8N_BACKEND=local python -m br8n.interfaces.mcp.server
~/.br8n/brain.db, bound to loopback, no account. Capture and resume need no key; search and explore need one. Your code context never leaves your laptop.
cd backend and pip install -e ".[dev]" into a Python 3.11 venv. That's the contributor path, not the try-it path.
Under the hood, the cloud path swaps the local SQLite store for a hosted Supabase backend behind the same get_store() seam, with a login for a real per-request token. When it's ready, the same capture→resume loop and the same surfaces will simply follow you across machines — no change to how you work.
Want it when it lands? Start on the free tier today — switch to Free / Local ›