bci-mcp

Ask Claude what your brain is doing.

Focus. Calm. Attention. Band powers. Over MCP — the same pipe GitHub, Notion, and the rest use to talk to Claude.

No EEG headset? There’s a built-in demo brain. You can wire this up and have a real conversation about “focus level” before you buy hardware.

What this package actually is

This is not the EEG stack. It’s the doorbell.

npx bci-mcp downloads and runs the Python server from PyPI. You get Node for install; Python does the signal processing. If uv or pip isn’t on your machine yet, the launcher will tell you.

Claude Code

claude mcp add bci-mcp -- npx -y bci-mcp

Open Claude, run /mcp, make sure bci-mcp is green. Then:

Connect to the demo brain. What’s my focus right now?

Claude Desktop

Settings → Developer → Edit Config:

{
  "mcpServers": {
    "bci-mcp": {
      "command": "npx",
      "args": ["-y", "bci-mcp"]
    }
  }
}

Quit Claude fully. Open it again. Same question as above.

Cursor

In ~/.cursor/mcp.json, under mcpServers:

"bci-mcp": {
  "command": "npx",
  "args": ["-y", "bci-mcp"]
}

Real hardware

Muse, OpenBCI, NeuroFocus, LSL streams — same server, different device URI. Start with synthetic:// until the headset is on your desk.

Repo + docs: github.com/enkhbold470/bci-mcp

MIT