SuperAPI AI Directory

ableton-mind

Control Ableton Live through a local Remote Script bridge: inspect sets, create tracks, scenes and clips, load devices, apply music recipes and verify changes.

Open website

About this tool

ableton-mind is an MCP (Model Context Protocol) server for Ableton Live, with the goal of building real music in Live using natural language through LLMs such as Claude, Cursor, or Codex. It communicates with Ableton Live through a local Remote Script bridge, exposes the complete Live Object Model to the LLM, and includes a native device knowledge base, declarative music recipes, an integrated validation loop, and reactive listeners.

Purpose This project is used to let large language models directly control and query Ableton Live: inspect project state, create tracks, scenes, and clips, load devices, apply music recipes, and validate the results of changes. It is suitable for users who want to operate projects in Live purely through language, automate repetitive setup, or experiment with musical ideas.

Main capabilities

  • Exposes the complete Live Object Model to the LLM.
  • Built-in native device knowledge base, including drum kits, basslines, racks, device modes, and other embedded JSON.
  • Supports declarative music recipes.
  • Integrated validation loop to confirm changes.
  • Provides reactive listeners.
  • Phase 8 slice 1 provides read-only Max for Live/plugin introspection, as well as Link/remote status discovery.
  • Supports three default tiers: read-only (safe), write (simple changes), and creative (recipe/browser loading).

Usage Installation is split into two parts: the Remote Script (Python bridge) and the MCP server. Remote Script installation:

  • Development mode (symlink):
    • macOS: copy live/AbletonMind/ to ~/Music/Ableton/User Library/Remote Scripts/AbletonMind/
    • Windows: copy to ~/Documents/Ableton/User Library/Remote Scripts/AbletonMind/
  • Then in Live → Preferences → Link/Tempo/MIDI → Control Surface → select AbletonMind.
  • Smoke testSee docs/smoke-test.md.

MCP server running and installation methods:

  • npm global install: npm install -g ableton-mind
  • Claude Code plugin marketplace: claude plugin marketplace add Pantani/ableton-mind, then claude plugin install ableton-mind@ableton-mind
  • Claude Desktop one-click: download ableton-mind-0.1.1.mcpb from the v0.1.1 GitHub Release
  • MCP Registry: io.github.Pantani/ableton-mind is active in the official registry
  • Source install: npm ci && npm run build && npm run install:remote-script
  • Docker: docker build -t ableton-mind . && docker run --rm -i --network host ableton-mind
  • Smithery: smithery.yaml is ready and can be used for listing/indexing
  • Glama: listed, but hosted publishing is separate from the GitHub Release, and the deployment and Make Release process still needs to be executed in the Glama admin panel

Architecture (three layers)

  • src/: TypeScript MCP server, including tools, resources, prompts, TCP client, recipe runner, and knowledge loader.
  • live/AbletonMind/: Python Remote Script, running a TCP server on port 9876 and dispatching JSON-RPC to LiveAPI.
  • recipes/, src/knowledge/: embedded JSON (drum kits, basslines, racks, device modes). The complete specification is in PLAN.md, and the frozen contracts are in _workspace/contracts/.

Applicable scenarios Suitable for Ableton Live 12 (preferred) and Live 11; macOS is primary, Windows is in PhasStage 1 final. For music producers, developers, or experimental users who want to operate Live projects in natural language via an LLM.

Limitations

  • Status is alpha, v0.1.1 has been released; core smoke tests have passed on Ableton Live 12.4.1 on macOS, and package validation has passed.
  • The API is unstable, and it is explicitly stated not to be used in production environments.
  • Deeper Max for Live control, the VST3 sidecar, remote DAW integration, and mobile companion features remain to be implemented.
  • Glama hosted publishing requires a separate admin-side deployment; Smithery metadata exists but indexing may lag.
  • Other features, charges, prices, platforms, or effects not described in the materials are not yet available.