Skip to main content
Glama

Ableton Live MCP Server

CI Release License: MIT Python Glama tool quality

Website and installation guide: abletonmcp.com

Control Ableton Live from an AI assistant. This is a Model Context Protocol (MCP) server that gives Claude, Cursor, Codex, or any MCP client 156 tools for building tracks, editing MIDI, loading instruments and effects, mixing, and mastering inside a running Ableton Live set, plus offline tools that read and diff saved .als project files with Live closed.

A Remote Script runs inside Live and exposes the Live API over a local socket. The MCP server turns that into typed, validated tools that an AI can call correctly, with per-tool documentation, read-only and destructive hints, and workflow prompts.

Quickstart

You need Ableton Live 11 or 12, Python 3.10+, and uv (brew install uv).

1. Register the MCP server.

Codex (local desktop, CLI or IDE host):

codex mcp add ableton -- uvx mcp-server-ableton-live@1.8.1
codex mcp list

For manual configuration and troubleshooting, see the Codex guide.

Claude Code:

claude mcp add AbletonMCP -s user -- uvx mcp-server-ableton-live@1.8.1

Claude Desktop or Cursor, in the mcpServers block of the config file:

{
  "mcpServers": {
    "AbletonMCP": {
      "command": "uvx",
      "args": ["mcp-server-ableton-live@1.8.1"]
    }
  }
}

(To run an unreleased revision instead, use uvx --from git+https://github.com/wstierhout/ableton-live-mcp ableton-live-mcp.)

2. Install the Remote Script. This copies the Live-side script into your User Library:

uvx mcp-server-ableton-live@1.8.1 install

Then activate the bridge in Live:

  1. Fully restart Ableton Live.

  2. Open Settings > Tempo & MIDI, then find the MIDI section and its Control Surface list.

  3. In an unused Control Surface row, select AbletonMCP. Leave existing hardware controller rows unchanged.

  4. In that same row, set Input: None and Output: None.

Link is a separate Settings page in current Live versions. In older versions, look under Link, Tempo & MIDI (or Link/MIDI) in Settings/Preferences, then find the MIDI section.

This selection is required for every MCP client. Installation copies the Remote Script; selecting the Control Surface starts it inside Live. The bridge uses a local socket, so its row does not need hardware MIDI input/output ports.

Run the installer again after upgrading the package so the Live-side script stays version-matched. See Ableton's guide to third-party Remote Scripts if the entry does not appear.

3. Check the setup.

uvx mcp-server-ableton-live@1.8.1 doctor --json

Use the same --user-library "/path/to/User Library" for install and doctor if your library has moved. If an older/different script exists, installation stops: review it before adding --replace. A unique backup is retained. Restart Live before diagnosing; doctor checks both installed bytes and the running bridge version.

4. Start read-only. Ask for the tempo and track names without changes. Then, in a disposable set, try: "Make a lofi beat at 80 BPM with a dusty drum kit, an upright bass, and Rhodes chords, then put a limiter on the master at -1 dB."

Related MCP server: remix-mcp

What it can do

Area

Tools

Session and transport

session info, one-call session snapshot, play/stop, tempo, tap tempo, time signature, loop, locators and cue navigation, scenes, undo/redo, capture MIDI, song scale, global groove and swing, Ableton Link

Tracks and mixer

create and delete MIDI, audio, and return tracks; delete devices; take lanes; group-track fold; volume, pan, mute, solo, arm, sends; input/output routing; meters

Clips and notes

create clips, write and edit MIDI notes (with probability), quantize with strength, Groove Pool swing, loop, gain, pitch, warp mode

Devices

browse and search by name, load instruments and effects onto any track including Master and Returns, read and set any parameter, sidechain routing, sample preview, rack macro variations, Simpler sample slicing, per-pad drum control, curated device knowledge base

Arrangement

place clips on the timeline, read and delete arrangement clips, write clip automation

Generators

drum patterns in 7 styles, euclidean rhythms, chord progressions, voice-led jazz voicings, 50-plus genre-aware progressions, voice-leading melodies, walking basslines, motif transforms (invert, retrograde, augment), minimalist processes, humanize, one-call session setup

Batch

run many edits in one round trip and one undo step

Audio and analysis

record a section to a WAV without the Export dialog, detect key and scale, scan the mix for problems, diff the session since the last check

Offline (no Live)

summarize, list tracks, extract MIDI, detect key, read mixer and locator detail, diff two versions, lint a saved .als, and parse .adg/.adv racks with edition detection

Recipes

scaffold a genre starter (lofi, house) in one call

There are also five workflow prompts (make_a_beat, mix_and_master, start_a_track, sound_design, analyze_and_improve) and a set of server instructions that teach the model the conventions before its first call: 0-based indices, beats for time, 0.85 volume equals 0 dB, and native ranges for device parameters.

Notes

Every action is a specific tool with validated arguments; there is no arbitrary-code path. Some tools still make destructive edits (delete a track, replace a clip's notes, overwrite an arrangement region), so save your work before a big session.

Audio analysis and session changes

analyze_audio_file reads WAV (integer and floating point), AIFF/AIFF-C, FLAC, Ogg Vorbis, MP3, AAC and M4A/MP4 (AAC or Apple Lossless) locally. The package installs the PyAV decoder dependency. It preserves the file, sample rate, channel layout and floating-point headroom; compressed files are measured after decoding. Analysis covers the first 30 seconds by default (up to 120 seconds, with a 32-million-sample processing budget), and reports when the result is partial. Measurements include sample peak and RMS, not LUFS or true peak.

session_diff distinguishes renames, reordering and duplicate track names using identifiers from the running bridge. Call it before and after an edit; the first call establishes the comparison baseline. Installing an updated Remote Script and restarting Live keeps the bridge's snapshot format matched to the server.

Focusing the toolset

The server registers 156 tools. That is a lot for a model to choose from on a small task. Set ABLETON_TOOLSETS to load only the groups you need, for example ABLETON_TOOLSETS=session,tracks,clips,generators. Groups (each may span several modules): session, tracks, clips, devices, browser, arrangement, generators, audio, analysis, offline, recipes. Unset loads everything.

Conventions

  • Indices are 0-based. Times and lengths are in beats. Volume uses Live's 0.0 to 1.0 fader range where 0.85 is 0 dB.

  • Device parameters use each parameter's own range. Read get_device_parameters first and check min, max, and the display value before setting.

  • add_notes_to_clip replaces the clip's notes. Use edit_notes to add or remove a few without touching the rest.

  • Placing a session clip into the arrangement overwrites whatever is under it, which is also how you replace a section.

  • Transport replies report the state before the command ran. Confirm with get_session_info.

Troubleshooting

  • AbletonMCP is missing from the Control Surface list: the folder has to be named exactly AbletonMCP, and Live only scans scripts at startup, so restart it.

  • Every command times out: a dialog is open in Live and it blocks the script. On the trial, that is the startup nag. Dismiss it and retry.

  • Port 9877 is not listening: Live is not running, or the Control Surface is not set. If another program already uses 9877, set ABLETON_MCP_PORT in Live's launch environment and ABLETON_PORT for the server so both sides match.

  • You cannot export or freeze from a tool: Live's API has no render or freeze function, so those stay manual. For rendered audio, export with the shortcut and analyse the file separately.

Security

The Remote Script listens on 127.0.0.1:9877 with no authentication. It is bound to loopback so only local processes can reach it. Do not forward that port or change the bind address to a public interface. See SECURITY.md.

Development

ableton_live_mcp/
  server.py        entrypoint: runs the MCP server or a setup subcommand
  app.py           FastMCP app, lifecycle, and server instructions
  connection.py    socket client to the Remote Script
  cli.py           install / uninstall / doctor subcommands
  tools/           session, tracks, clips, devices, browser, arrangement,
                   generators, prompts
  remote_script/
    __init__.py    the Live-side script and its command dispatch tables
tests/             protocol, dispatch-contract, and registration tests
uv run --extra dev pytest      # tests
uvx ruff check .               # lint

CI runs lint, a byte-compile of the Remote Script, and tests on Python 3.10, 3.12, 3.13, and 3.14.

Frequently asked questions

What is the Ableton Live MCP Server?

It is a local Model Context Protocol server that lets AI clients such as Claude, Cursor, and Codex inspect and control Ableton Live through typed tools. It can build tracks, edit MIDI, load and configure devices, arrange clips, mix a set, and analyze saved .als projects.

How is this different from other Ableton MCP servers?

This server exposes 156 specific, validated tools rather than an arbitrary-code execution tool. It includes destructive/read-only hints, workflow prompts, built-in music generators, mixing and analysis tools, and offline .als/.adg inspection. The Remote Script and MCP package are versioned and tested together.

Can it work while Ableton Live is closed?

The real-time control tools require Ableton Live and the bundled Remote Script to be running. The offline tools can summarize, inspect, lint, extract MIDI from, and diff saved .als, .adg, and .adv files without launching Live.

Which MCP clients are supported?

Any client that can launch a local stdio MCP server can use it, including Claude Desktop, Claude Code, Cursor, Codex, and compatible IDEs. The canonical command is uvx mcp-server-ableton-live.

Is this an official Ableton product?

No. It is an independent, open-source integration and is not affiliated with or endorsed by Ableton.

Credits and license

MIT licensed, maintained by Wouter Stierhout. See LICENSE. Ableton and Live are trademarks of Ableton AG. This independent open-source project is not affiliated with or endorsed by Ableton AG.

Available Tools

156 tools
add_notes_to_clipA
Destructive

REPLACES the clip's entire note content with the given notes. Use edit_notes to add or remove a subset without touching the rest.

notes: list of {"pitch": 0-127 (60 = C3 in Live's naming), "start_time": beats from clip start (float), "duration": beats, "velocity": 1-127, "mute": bool}. Precondition: the clip must exist (create_clip first).

Inputs: notes (required), clip_index (required), track_index (required). Returns: result

adg_analyzeA
Read-only

Full offline analysis of a .adg rack or .adv preset: named macros (name/value), the chain + device tree (recursive, incl. racks-inside-racks with each device's kind/name/on-state and any sample), legacy macro-> parameter mappings, every referenced sample (name + relative path), the required Live edition, and — for drum racks — the pad map (pad -> MIDI note, label, and sample). No Live required.

Inputs: path (required). Returns: result

adg_editionA
Read-only

Classify the Live edition a saved .adg/.adv needs, offline: 'suite' if any Suite-only device is present (Operator, Wavetable, Echo, Drift, Meld, Sampler, Collision, ...), else 'standard' if any Standard-only device is present (EQ Eight, Compressor, Simpler, Impulse, Drum Rack, ...), else 'intro'. Returns the verdict plus which devices drove it, so an agent can tell whether a preset will load on the user's edition before loading it.

Inputs: path (required). Returns: result

adg_summaryA
Read-only

Summarize a saved .adg rack or .adv device preset WITHOUT Live running: name, rack type, total device count, top-level chain count, named-macro count, required Live edition (intro/standard/suite), and whether it is a drum kit (is_drum_rack is true when the preset is or wraps a Drum Rack — factory kits ship wrapped in an Instrument Rack). path is a filesystem path to a .adg or .adv file.

Inputs: path (required). Returns: result

als_detailsA
Read-only

Read mixing and arrangement detail from a saved .als, offline: the set's locators (name + beat time) and each track's fader volume, pan, and number of automation lanes. Complements als_summary for reviewing a mix without Live.

Inputs: path (required). Returns: result

als_detect_keyA
Read-only

Detect the key/scale of a saved .als file WITHOUT Live running (Krumhansl-Kessler). Pools notes from every MIDI clip in the set, or only one track when track_index is given. weight selects the histogram weighting ("duration", "velocity", "product", "count"). path is a filesystem path to a .als file.

Inputs: path (required), weight, track_index. Returns: result

als_diffA
Read-only

Diff two saved .als files (e.g. song_v4.als vs song_v5.als) offline: tempo/time-signature changes, tracks added/removed by name, and per-track deltas in device chain, clip count, note count, and mute state. Great for logging what changed between versions.

Inputs: path_a (required), path_b (required). Returns: result

als_extract_midiA
Read-only

Extract MIDI notes from a track in a saved .als file, as {pitch,start,duration,velocity} dicts per clip. track_index is the track's position in the set (see als_summary). No Live required.

Inputs: path (required), track_index (required). Returns: result

als_find_unfinishedA
Read-only

Lint a saved .als for likely-unfinished work, offline: no master limiter/compressor, MIDI track with no instrument, MIDI clip with no notes, empty audio track, or muted tracks. Returns machine-readable issues so an agent can decide what to fix before rendering.

Inputs: path (required). Returns: result

als_list_tracksA
Read-only

List every track in a saved .als file with its clips (name, MIDI/audio, start, length, looping, sample path, note count). No Live required.

Inputs: path (required). Returns: result

als_summaryA
Read-only

Summarize a saved .als file WITHOUT Live running: version/creator, tempo, time signature, and each track's kind, name, mute/solo, device names, clip count, and note count. path is a filesystem path to a .als file.

Inputs: path (required). Returns: result

analyze_audio_fileA
Read-only

Measure local WAV (integer/float), AIFF/AIFF-C, FLAC, Ogg Vorbis, MP3, AAC or M4A/MP4 (AAC/ALAC) without Live: per-channel sample peak/RMS dBFS, DC offset, crest factor and full-scale samples. Analyzes the first max_seconds (default 30, maximum 120); marks partial results. Not LUFS or true peak. No upload. Compressed files are measured after decoding; this does not recover lost audio. Files are read locally without modification, normalization or upload.

Inputs: path (required), max_seconds. Returns: result

analyze_mixA
Read-only

Heuristic session-state checks, NOT audio, peak, LUFS or headroom measurements. Scan the current live set for inspection leads: several high faders, a muted or empty Session track, or a MIDI track without devices. External routing and Arrangement audio may explain these states. Returns machine-readable findings so you can decide what to fix. Reads the session; changes nothing.

Inputs: no named parameters. Returns: result

apply_recipeA
Destructive

Scaffold a genre starter in one call: set the tempo, add Drums, Bass, and Chords MIDI tracks, try to load fitting instruments, and write generated parts. name is one from list_recipes (e.g. "lofi_beat", "house_groove"). tempo/key/scale/bars override the recipe defaults. Adds new tracks; it does not overwrite existing ones.

Inputs: name (required), key, bars, scale, +1 more. Returns: result

back_to_arrangerA

Return playback control to the Arrangement (after Session clips have overridden arrangement content - the orange 'Back to Arrangement' state).

Inputs: no named parameters. Returns: result

batch_commandsA
Destructive

Run several commands in ONE round-trip and ONE undo step (atomic-ish: stops at the first error; completed steps are a single undo away).

commands = [{"type": "", "params": {...}}, ...]. Command names match the MCP tool names (the few that differ are auto-translated), e.g. [{"type": "set_tempo", "params": {"tempo": 80}}, {"type": "create_midi_track", "params": {"index": -1}}]. Composite tools (record_section, load_drum_kit, apply_recipe, save_set, the generate_* tools) are NOT batchable - they orchestrate multiple commands server-side. Prefer this for multi-step edits: fewer round-trips, one undo reverts all.

Inputs: commands (required). Returns: result

capture_and_insert_sceneA

Snapshot the currently playing Session clips into a new scene - the 'keep that combination' workflow.

Inputs: no named parameters. Returns: result

capture_midiA

Grab recently played MIDI into a new clip (Live's Capture MIDI). Requires that MIDI was recently played into an armed/monitored track.

Inputs: no named parameters. Returns: result

clear_automationA
DestructiveIdempotent

Delete one device parameter's entire automation envelope from a clip.

Resolve the device and parameter with get_device_parameters first. This removes every automation point for that parameter, not a time range; manual parameter state remains. Use write_automation to replace the envelope with new points, or undo immediately to recover an accidental clear.

Inputs: parameter (required), clip_index (required), track_index (required), device_index (required). Returns: result

clip_operationA
Destructive

Structural clip edits. op: "duplicate_loop" (double the loop, duplicating notes - extend 2 bars to 4 before adding variation), "crop" (discard material outside the loop), "duplicate_region" (copy notes from region_start/region_length to destination_time, optionally transposing by transposition_amount semitones; pitch=-1 means all pitches).

Inputs: op (required), clip_index (required), track_index (required), pitch, +4 more. Returns: result

continue_playingA

Resume playback from the current position (start_playback restarts from the playhead's last start point instead).

Inputs: no named parameters. Returns: result

create_audio_clipA

Create a new audio clip in an audio track's clip slot by importing a file.

Requires Ableton Live 12.0.5 or newer - the underlying ClipSlot.create_audio_clip Live API was introduced in 12.0.5 and is not available in earlier 12.0.x releases.

Parameters:

  • track_index: The index of the audio track to create the clip in

  • clip_index: The index of the clip slot to create the clip in

  • path: Absolute path to a supported audio file (e.g. a .wav). The target track must be an audio track and the clip slot must be empty.

Inputs: path (required), clip_index (required), track_index (required). Returns: result

create_audio_trackA

Insert an empty audio track at a chosen position, or append with -1.

Non-negative indices insert before the current track at that position and shift it and later track indices up. Use create_midi_track for instruments and MIDI clips, or create_return_track for a shared send-effect bus.

Inputs: index. Returns: result

create_clipA

Create an empty MIDI clip in a Session slot. Fails if the slot already has a clip (pick another clip_index or delete_clip first) or the track is an audio track. length is in beats (4.0 = one 4/4 bar). Typical chain: create_clip -> add_notes_to_clip -> fire_clip or duplicate_to_arrangement.

Inputs: clip_index (required), track_index (required), length. Returns: result

create_locatorA

Create or rename an Arrangement locator at an absolute beat position.

SIDE EFFECT: first moves the Arrangement playhead to time, because Live creates cues at the playhead. If a locator already exists there, it is reused and renamed when name is provided. Use set_arrangement_time when only the playhead should move, and get_locators to inspect existing markers.

Inputs: time (required), name. Returns: result

create_midi_trackA

Create a new MIDI track in the Ableton session.

Parameters:

  • index: The index to insert the track at (-1 = end of list)

Inputs: index. Returns: result

create_return_trackA

Create a new return track (a send bus, e.g. for shared reverb or delay).

Inputs: no named parameters. Returns: result

create_sceneA

Create a new scene (row of clip slots). index -1 appends at the end.

Inputs: index. Returns: result

create_take_laneA

Add a take lane to a track (for comping multiple recorded takes on one track). Returns the new take-lane count.

Inputs: track_index (required). Returns: result

delete_arrangement_clipA
Destructive

Delete a clip from the Arrangement timeline by its position in get_arrangement_clips' list. NOTE: indices of later clips on the same track shift down by one after each delete - re-read get_arrangement_clips between deletes.

Inputs: track_index (required), arrangement_clip_index (required). Returns: result

delete_clipA
Destructive

Delete the existing Session-view clip in a slot, including its content.

This leaves the slot empty and does not shift other slot indices. It fails for an empty slot. Use stop_clip if the intent is only to stop playback; use undo immediately if the deletion was accidental.

Inputs: clip_index (required), track_index (required). Returns: result

delete_deviceA
Destructive

Delete a device from a track's chain by its position (0-based). Shifts the indices of later devices, so re-read the chain afterwards.

Inputs: track_index (required), device_index (required). Returns: result

delete_return_trackA
Destructive

Delete a return track by index (0 = Return A). Later returns shift down.

Inputs: return_index (required). Returns: result

delete_sceneA
Destructive

Delete a scene (row). Later scene indices shift down by one.

Inputs: scene_index (required). Returns: result

delete_trackA
Destructive

Delete a track. NOTE: indices of all later tracks shift down by one.

Inputs: track_index (required). Returns: result

describe_capabilitiesA
Read-only

A high-level map of this server: the tool groups and what each covers, plus the conventions to follow, so you can orient before your first call. Set ABLETON_TOOLSETS to load only some groups. For the current live state, call get_session_snapshot; for mix problems, analyze_mix.

Inputs: no named parameters. Returns: result

describe_deviceA
Read-only

Return curated guidance for a common Ableton native device: which parameters matter, their units (note curve = perceptual, not proportional), sensible values, and mixing tips. device_name is case-insensitive (e.g. "EQ Eight", "Compressor", "Glue Compressor", "Reverb", "Auto Filter", "Saturator", "Utility", "Limiter", "Delay"). Pair with get_device_parameters for the live indices and ranges.

Inputs: device_name (required). Returns: result

detect_clip_keyA
Read-only

Detect the key/scale of one Session MIDI clip (Krumhansl-Kessler). Reads the clip's notes and returns the best-fit key with a confidence (-1..1), margin over the runner-up, and the runner-up key. weight is how notes are weighted into the pitch-class histogram: "duration" (default), "velocity", "product", or "count". Reads only; changes nothing.

Inputs: clip_index (required), track_index (required), weight. Returns: result

detect_session_keyA
Read-only

Detect the overall key/scale of the whole set by pooling the notes of every MIDI track's Session clips (Krumhansl-Kessler). Gives the harmonic center of the arrangement as a whole. weight selects the histogram weighting ("duration", "velocity", "product", "count"). Reads only.

Inputs: weight. Returns: result

detect_track_keyA
Read-only

Detect the key/scale of a whole track by pooling the notes of every Session clip it holds (Krumhansl-Kessler). Useful when a part is spread across several clips. weight selects the histogram weighting ("duration", "velocity", "product", "count"). Reads only; changes nothing.

Inputs: track_index (required), weight. Returns: result

duplicate_clip_toA
Destructive

Copy a Session clip to any other track/scene slot (overwrites the destination slot). The cross-track way to build variations.

Inputs: dst_scene (required), dst_track (required), src_scene (required), src_track (required). Returns: result

duplicate_sceneA

Insert a copy of one Session scene, including every clip in its row.

This creates new content and increases the scene count; later scene indices shift to make room. Use duplicate_track for a whole track or duplicate_clip_to for one clip. Use the new scene as an editable section variation.

Inputs: scene_index (required). Returns: result

duplicate_to_arrangementA
Destructive

Copy a Session-view clip into the Arrangement timeline.

OVERWRITES whatever already occupies the destination range on that track (like recording over tape) - this is also the supported way to REPLACE a section. Uses Live's track.duplicate_clip_to_arrangement() API (Live 11/12). The clip is placed at destination_time beats from the start of the arrangement on the same track it lives in.

Typical workflow:

  1. create_clip / add_notes_to_clip to build a Session clip

  2. Call duplicate_to_arrangement once per bar/section you need

  3. Call switch_to_arrangement_view to confirm the result in Live

Parameters:

  • track_index: Index of the track that owns the Session clip

  • clip_index: Index of the clip slot in that track (Session view)

  • destination_time: Beat position in the arrangement to place the clip (e.g. 0.0 = start, 8.0 = bar 3 in 4/4)

Inputs: clip_index (required), track_index (required), destination_time (required). Returns: result

duplicate_trackA

Insert a copy of one regular track, including its devices and clips.

This creates new content and increases the track count; later track indices shift to make room. Use duplicate_scene for a Session row or duplicate_clip_to for one clip. Rename or edit the copy to create a variation.

Inputs: track_index (required). Returns: result

edit_notesA

Edit a subset of notes in a MIDI clip without rewriting all of them. add = [{pitch,start_time,duration,velocity,mute}], remove = [{pitch,start_time}] (matched by pitch+start).

Inputs: clip_index (required), track_index (required), add, remove. Returns: result

fire_clipA

Launch one existing Session-view clip using Live's launch quantization.

The slot must contain a clip. Launching it takes Session control of that track and replaces any other playing Session clip on the same track; firing an already-playing clip may retrigger it according to its launch settings. Use fire_scene to launch a whole row, or continue_playing for transport only.

Inputs: clip_index (required), track_index (required). Returns: result

fire_sceneA

Launch a Session-view scene using Live's launch quantization.

This fires the row's clips together and can replace or stop clips currently playing on the affected tracks according to each slot's launch behavior. Use fire_clip to launch only one track's clip, or start_playback when the intent is Arrangement transport rather than Session launching.

Inputs: scene_index (required). Returns: result

generate_additiveA
Destructive

Write a Philip-Glass additive process into a Session clip. REPLACES the clip's notes if it already exists.

Builds a short one-voice motif (each pitch a note_length-beat note, back to back) from pitches -- a space-/comma-separated list of MIDI numbers. Then plays the first 1 note, then the first 2, then the first 3 ... concatenated end to end, so the phrase audibly grows one note per stage. steps is how many stages to build (0 = use every note; capped at the number of pitches).

If source_clip_index is given, the motif is read from that clip on the same track instead of built from pitches. Returns a short summary.

Inputs: clip_index (required), track_index (required), steps, pitches, +3 more. Returns: result

generate_basslineA
Destructive

Write a bassline following a chord progression. REPLACES existing notes.

progression: same syntax as generate_chord_progression. style: "roots" (long root notes), "root_fifth" (root + fifth bounce), "walking" (root, approach tones), "eighth_pump" (driving 8ths on the root). octave: 1 = very low (C1=24), 2 = typical bass (C2=36).

Inputs: clip_index (required), progression (required), track_index (required), style, +3 more. Returns: result

generate_chord_progressionA
Destructive

Generate voiced chords and write them into one Session-view MIDI clip.

DESTRUCTIVE: creates a clip when the slot is empty; otherwise resizes its loop and replaces every existing note. The track must accept MIDI. Progressions accept dash/comma/space-separated symbols such as Am9-Fmaj7-Cmaj9-G13; supported qualities include maj7, m7, m9, 9, 13, 7#9, sus4, dim, and add9. Use add_notes_to_clip for exact notes or edit_notes to preserve existing ones.

Inputs: clip_index (required), progression (required), track_index (required), strum, +4 more. Returns: result

generate_drum_patternA
Destructive

Write a genre drum pattern into a Session clip on a Drum Rack track. REPLACES the clip's notes if the clip exists.

style: lofi | boom_bap | house | techno | trap | dnb | ambient. swing: beats to delay 8th-note offbeats (0.0-0.1; 0.06 ≈ MPC-ish). humanize: max random velocity deviation (0 = machine-perfect). Uses the standard Live drum map (36 kick, 38 snare, 42/46 hats...).

Inputs: clip_index (required), track_index (required), bars, seed, +3 more. Returns: result

generate_euclidean_drumsA
Destructive

Write a Euclidean (Bjorklund) rhythm for one drum voice into a Session clip. REPLACES the clip's notes if the clip exists.

Distributes pulses hits as evenly as possible across steps slots spanning one bar (steps=8 -> 8th notes, 16 -> 16th notes), repeated for bars. pulses=3, steps=8 is the tresillo. rotate shifts the pattern left by that many steps. instrument: a drum-map name (kick/snare/chat/ohat/clap/ride/...) or a raw MIDI number. humanize is max +/- velocity jitter.

Inputs: clip_index (required), track_index (required), bars, seed, +6 more. Returns: result

generate_genre_progressionA
Destructive

Build a genre-idiomatic chord progression and write it as voiced chords. REPLACES existing notes.

genre: lofi/boom_bap/hip_hop/trap/house/deep_house/techno/trance/dnb/rnb/ neo_soul/jazz/bossa_nova/ambient/synthwave/pop/afrobeats/amapiano/... (fuzzy matched). The genre's scale-degree grammar is transposed into key+scale, then voiced (see generate_voiced_progression for the style options). bars = number of chords (the grammar loops). Returns the chord symbols it used.

Inputs: genre (required), clip_index (required), track_index (required), key, +8 more. Returns: result

generate_grooveA
Destructive

Write a pocketed drum groove into a Session clip. REPLACES existing notes.

style: lofi | boom_bap | house | trap | jazz | funk. pocket: beats to lay every hit behind the beat (0.0-0.1). swing: off-beat hat delay (0.5 straight .. ~0.66). ghost: 0-1 probability of soft ghost snares between backbeats. humanize: max +/- velocity deviation. Uses the standard Live drum map.

Inputs: clip_index (required), track_index (required), bars, seed, +5 more. Returns: result

generate_melodyA
Destructive

Write a voice-led melody over a chord progression. REPLACES existing notes.

Notes come from key+scale, biased to chord tones on strong beats, chosen by nearest-pitch voice-leading (small steps), with optional chromatic approach notes. progression: dash-/comma-separated symbols. key: e.g. "C", "F#", "Bb". scale: major/minor/dorian/lydian/mixolydian/blues/bebop/minor_pentatonic/... density: low|medium|high (note rate). phrase_arc: rising|arch|ascend_descend| static (register contour). chromatic: 0-1 approach-note probability. low/high bound the register (MIDI). swing: 0.5 straight .. ~0.66.

Inputs: clip_index (required), progression (required), track_index (required), key, +9 more. Returns: result

generate_phaseA
Destructive

Write a Steve-Reich phasing pattern into a Session clip. REPLACES the clip's notes if it already exists.

Builds a short one-voice motif (each pitch a note_length-beat note, back to back) from pitches -- a space-/comma-separated list of MIDI numbers, e.g. "60 62 64 67 69". Then stacks repeats copies of that motif, each copy starting shift beats later than the last, so the copies drift against one another. A shift smaller than the motif length (len(pitches)*note_length) gives the classic overlapping phase texture; shift == motif length is a plain loop.

If source_clip_index is given, the motif is read from that clip on the same track instead of built from pitches. Returns a short summary.

Inputs: clip_index (required), track_index (required), shift, pitches, +4 more. Returns: result

generate_voiced_progressionA
Destructive

Write voice-led jazz chord voicings into a Session clip. REPLACES existing notes.

progression: dash-/comma-separated symbols, e.g. "Am9-Dm7-G13-Cmaj7". style: rootless (3-7-9-13 stack, no root), quartal (stacked 4ths), shell (root-3-7), or block (close-position). Smooth voice-leading is applied between chords. center_pitch biases register (60 = C3 in Live). voices caps notes per chord; humanize is max +/- velocity jitter.

Inputs: clip_index (required), progression (required), track_index (required), seed, +6 more. Returns: result

generate_walking_bassA
Destructive

Write a quarter-note walking bass line under a progression. REPLACES notes.

Root on beat 1, chord/scale tones through the bar, chromatic approach into the next chord's root on the last beat, with occasional soft ghost notes. progression: dash-/comma-separated symbols. key/scale set the passing-tone pool. octave: 1 = very low (C1), 2 = typical bass (C2). ghost: 0-1 offbeat ghost-note probability.

Inputs: clip_index (required), progression (required), track_index (required), key, +6 more. Returns: result

get_arrangement_clipsA
Read-only

List all clips placed in the Arrangement timeline for a track.

Returns each clip's name, start_time, end_time, length, and type.

Parameters:

  • track_index: The index of the track to inspect

Inputs: track_index (required). Returns: result

get_browser_items_at_pathA
Read-only

Get browser items at a specific path in Ableton's browser.

Parameters:

  • path: Path in the format "category/folder/subfolder" where category is one of the available browser categories in Ableton

Inputs: path (required). Returns: result

get_browser_treeA
Read-only

List the browser's top-level categories and their immediate entries. NOTE: returns a shallow listing (no deep hierarchy) - drill down with get_browser_items_at_path("instruments"), ...("sounds/Piano & Keys"), etc., or skip straight to search_browser when you know a name. category_type: "all", "instruments", "sounds", "drums", "audio_effects", "midi_effects".

Inputs: category_type. Returns: result

get_clip_infoA
Read-only

Full clip state: loop points, markers, playing position, per-clip time signature; for audio clips also file path, gain display, and warp markers.

Inputs: clip_index (required), track_index (required). Returns: result

get_clip_notesA
Read-onlyIdempotent

Return every MIDI note in an existing Session-view MIDI clip.

Each note includes pitch, start time, duration, velocity, and mute state; newer Live versions may add expression fields. This is read-only and fails for an empty slot or audio clip. Use get_clip_info for clip properties.

Inputs: clip_index (required), track_index (required). Returns: result

get_device_parametersA
Read-only

List all parameters of a device on a track: names, values, ranges, display strings.

Inputs: track_index (required), device_index (required). Returns: result

get_device_routingA
Read-only

Read a device's audio-input (sidechain) routing: current input type/channel and the available options. Only sidechain-capable devices (Compressor, Gate) have this. Use the options with set_device_routing to pick a sidechain source.

Inputs: track_index (required), device_index (required). Returns: result

get_drum_padsA
Read-only

List a Drum Rack's occupied pads (MIDI note, name, mute, solo).

Inputs: track_index (required), device_index (required). Returns: result

get_groovesA
Read-only

List grooves in the Groove Pool (index + name) and the global groove amount.

Inputs: no named parameters. Returns: result

get_group_infoA
Read-only

Report a track's group state: whether it is a foldable group track, whether it is inside a group, its fold state, and its parent group's name.

Inputs: track_index (required). Returns: result

get_locatorsA
Read-only

List arrangement locators (index, name, beat time) for structure-aware navigation. Jump with jump_to_locator.

Inputs: no named parameters. Returns: result

get_master_device_parametersA
Read-only

List parameters of a device on the Master track.

Inputs: device_index (required). Returns: result

get_master_metersA
Read-only

Read the Master track's output meters - values pinned near 1.0 during playback mean the master is clipping; pull set_master_volume down.

Inputs: no named parameters. Returns: result

get_rack_chainsA
Read-only

List an Instrument/Effect Rack's chains and the devices inside each - previously unreachable nested devices. Use set_chain_device_parameter to control them.

Inputs: track_index (required), device_index (required). Returns: result

get_return_device_parametersA
Read-only

List parameters of a device on a Return track: names, values, min/max, display strings. Call before set_return_device_parameter.

Inputs: device_index (required), return_index (required). Returns: result

get_return_tracksA
Read-only

List return tracks (index + name).

Inputs: no named parameters. Returns: result

get_scale_infoA
Read-only

Read the song's scale and tuning: scale name, root note, scale intervals, and the active tuning system (microtonal, Live 12.3+) if any. Set the scale with set_song_scale; tune device parameters (e.g. Wavetable) with set_device_parameter.

Inputs: no named parameters. Returns: result

get_session_infoA
Read-onlyIdempotent

Return global Live-set and transport state without modifying the set.

Includes tempo, song time signature, playback position/state, Arrangement loop, track/return/scene counts, scene names, key/scale, record mode, launch quantization, Master volume/pan, and Live version. Use get_track_info for a single track's clips/devices, or get_session_snapshot for a compact overview of every track.

Inputs: no named parameters. Returns: result

get_session_snapshotA
Read-only

One-call overview of the whole set: tempo, time signature, play state, and for every track its name, type, mute/solo/arm, volume, clip count, and device names, plus return-track names. Prefer this over many get_track_info calls.

Inputs: no named parameters. Returns: result

get_track_infoA
Read-onlyIdempotent

Return detailed state for one regular track without modifying the set.

Includes name/type, mixer and arm/mute/solo state, sends, freeze/color state, playing/fired slot indices, every Session slot's basic clip state, and the device-chain names/types. Use get_session_info for global transport state, get_device_parameters for parameter values, or get_clip_info for full clip properties.

Inputs: track_index (required). Returns: result

get_track_metersA
Read-only

Read a track's output level meters (post-fader): left/right 0.0-1.0ish (0.85 ≈ 0 dB, like faders). Poll during playback to judge balance/clipping - this is how you 'listen' without exporting.

Inputs: track_index (required). Returns: result

get_track_routingA
Read-only

Read a track's input/output routing (current + all available options) and monitoring state (0=In, 1=Auto, 2=Off). Routing options are identified by display_name - use those strings with set_track_routing.

Inputs: track_index (required). Returns: result

humanize_clipA
Destructive

Read a MIDI clip, apply timing + velocity humanization, and write it back. REPLACES the clip's notes with the jittered version.

timing: max +/- start-time deviation in beats. velocity: max +/- velocity deviation. Pitches are preserved; times stay non-negative. seed makes it reproducible.

Inputs: clip_index (required), track_index (required), seed, timing, +1 more. Returns: result

jump_byA

Move the playhead by a relative number of beats (negative moves back). Works during playback for on-the-fly navigation.

Inputs: beats (required). Returns: result

jump_to_cueA

Jump the playhead to the next locator/cue (direction >= 0) or the previous one (direction < 0).

Inputs: direction. Returns: result

jump_to_locatorA

Move the playhead to a locator (index from get_locators).

Inputs: locator_index (required). Returns: result

list_known_devicesA
Read-only

List the devices that describe_device has curated guidance for.

Inputs: no named parameters. Returns: result

list_recipesA
Read-only

List the built-in composition recipes (name, description, and default tempo/key/scale/bars), for use with apply_recipe.

Inputs: no named parameters. Returns: result

load_device_to_masterA

Load a device/effect from the browser onto the Master/Main track (e.g. a limiter for mastering).

Inputs: item_uri (required). Returns: result

load_device_to_returnA

Append one loadable browser device or preset to a return track's chain.

Obtain item_uri from search_browser. Loading selects the return track in Live and creates a new device, so repeated calls add duplicates. Prefer audio effects on returns; use load_instrument_or_effect for a regular track or load_device_to_master for the Master chain.

Inputs: item_uri (required), return_index (required). Returns: result

load_drum_kitA

Load a Drum Rack, then the first loadable kit at a browser path.

Use search_browser to obtain the rack URI and get_browser_items_at_path to verify the kit path. This is a two-step mutation: it appends the rack first, then loads the first loadable item found at kit_path. If path lookup fails, the new empty rack remains on the track; repeated calls add more racks. Use load_instrument_or_effect when one known browser URI is sufficient.

Inputs: kit_path (required), rack_uri (required), track_index (required). Returns: result

load_instrument_or_effectA

Load an instrument or effect onto a track using its URI.

Parameters:

  • track_index: The index of the track to load the instrument on

  • uri: The URI of the instrument or effect to load (e.g., 'query:Synths#Instrument%20Rack:Bass:FileId_5116')

Inputs: uri (required), track_index (required). Returns: result

preview_browser_itemA

Audition a browser item (a sample, instrument, or preset) through Live's preview tab WITHOUT loading it onto a track. item_uri comes from search_browser or get_browser_items_at_path. Use stop_browser_preview to stop.

Inputs: item_uri (required). Returns: result

quantize_clipA

Quantize a MIDI clip. grid: quarter/eighth/eighth_triplet/sixteenth/sixteenth_triplet/thirtysecond. amount 0.0-1.0 (use <1.0 to humanize toward the grid without full snap).

Inputs: clip_index (required), track_index (required), grid, amount. Returns: result

rack_variationA

Rack macro snapshots. action: "store" (save current macros as a variation), "recall" (restore variation index), "randomize" (randomize all macro knobs - instant sound-design exploration).

Inputs: action (required), track_index (required), device_index (required), index. Returns: result

record_sectionA
Destructive

Bounce a section of the arrangement to an audio file WITHOUT the Export dialog, so you can hear an internal signal. Creates a temporary audio track, routes its input to source ("Resampling" = the master output, or a track's name for that track's output), arms it, and records from start_beat to end_beat in REAL TIME (a 4-bar section takes about 4 bars of wall-clock; capped at 5 minutes). Returns the recorded WAV path. Set cleanup=False to keep the recording track. Runs the transport and adds an arrangement clip.

Inputs: end_beat (required), start_beat (required), source, cleanup. Returns: result

redoA

Redo the last undone action in Live.

Inputs: no named parameters. Returns: result

re_enable_automationA

Restore automation that was overridden by manual parameter changes. IMPORTANT: setting a device parameter that has an envelope overrides the envelope until this is called.

Inputs: no named parameters. Returns: result

replace_simpler_sampleA
DestructiveIdempotent

Replace a Simpler sample with an absolute local audio-file path. Requires Live 12.4+ AND runtime API support; otherwise fails before writing. Replaces musical content: confirm the target and preserve the original set. Readback is not an audio audition.

Inputs: path (required), track_index (required), device_index (required). Returns: result

save_setA

Save the current Live set to its existing file, if the Live API supports it (it does not on most versions). Returns whether the API call is available and whether the save succeeded. To save under a new name/path, use the GUI Save-As.

Inputs: no named parameters. Returns: result

search_browserA
Read-only

Search the Ableton browser by display name (case-insensitive substring). Returns loadable items with their URIs for load_instrument_or_effect / load_device_to_master / load_device_to_return.

category: optionally limit to one of "instruments", "sounds", "drums", "audio_effects", "midi_effects", "packs" (default: search all of them). Prefer this over walking get_browser_items_at_path when you know a name.

Inputs: query (required), category, max_results. Returns: result

session_diffA
Read-only

Report what changed in the set since the LAST time you called this tool: tempo/time-signature/play-state, tracks added, removed, renamed or reordered, and per-track volume, mute, solo, arm, clip-count, and device changes. The first call records a baseline (no diff). Track IDs are scoped to the running bridge; older bridges cannot identify renames reliably. Call it before and after an edit to verify it took effect.

Inputs: no named parameters. Returns: result

set_arrangement_overdubA

Toggle arrangement overdub: recording layers MIDI onto existing arrangement clips instead of replacing them.

Inputs: enabled (required). Returns: result

set_arrangement_timeA

Move the arrangement playhead to a specific position.

Parameters:

  • time: Position in beats from the start of the arrangement (e.g. 8.0 = bar 3 in 4/4)

Inputs: time (required). Returns: result

set_chain_device_parameterA

Set a parameter on a device INSIDE a rack chain (indices from get_rack_chains). Values clamp to the parameter's native range.

Inputs: value (required), parameter (required), chain_index (required), track_index (required), +2 more. Returns: result

set_clip_audioA

Set audio-clip properties. gain 0-1, pitch_coarse semitones (-48..48), pitch_fine cents, warping on/off, warp_mode 0=Beats 1=Tones 2=Texture 3=Re-Pitch 4=Complex 6=Complex Pro.

Inputs: clip_index (required), track_index (required), gain, warping, +3 more. Returns: result

set_clip_colorA
Idempotent

Set an existing Session clip's palette color without changing its content.

This is a visual organization change only. Use set_track_color to recolor the track header instead. Repeating the same color has no additional effect.

Inputs: clip_index (required), color_index (required), track_index (required). Returns: result

set_clip_grooveA

Attach a Groove Pool groove to a clip by index (swing/timing feel). Pass no groove_index to clear.

Inputs: clip_index (required), track_index (required), groove_index. Returns: result

set_clip_loopA
Idempotent

Update selected loop and launch markers on one existing Session clip.

Provide at least one optional field; omitted fields stay unchanged. start and end are loop-brace positions in clip-relative beats and end must be after start. start_marker controls where non-looped launch begins; looping enables or disables repetition. Use set_loop for the Arrangement's global loop region, not a clip, and get_clip_info to inspect current values.

Inputs: clip_index (required), track_index (required), end, start, +2 more. Returns: result

set_clip_nameA
Idempotent

Rename an existing Session-view clip without changing its notes or audio.

The slot must contain a clip. Use this for the clip label; use set_track_name for the track label. Repeating the same name has no additional effect.

Inputs: name (required), clip_index (required), track_index (required). Returns: result

set_clip_signatureA

Set a clip's own time signature (polymeter / odd-bar loops).

Inputs: numerator (required), clip_index (required), denominator (required), track_index (required). Returns: result

set_clip_trigger_quantizationA

Global launch quantization for firing clips/scenes. 0=None (instant, good for auditioning), 4=1 bar (default), 7=1/4 note, 13=1/32.

Inputs: value (required). Returns: result

set_crossfade_assignA
Idempotent

Assign one regular track to crossfader side A, neither side, or side B.

This configures which side affects the track; it does not move the crossfader or change track volume immediately. Use set_crossfader afterward to blend the A/B groups. Repeating the same assignment has no additional effect.

Inputs: assign (required), track_index (required). Returns: result

set_crossfaderA

Set the master crossfader (-1.0 = full A, 0 = center, 1.0 = full B).

Inputs: value (required). Returns: result

set_device_enabledA
Idempotent

Enable or bypass one device on a regular track.

true turns the device on; false bypasses its processing while preserving the device and parameter values. This changes audible signal flow but not clip content. Use set_track_mute to silence the whole track, or set_device_parameter to change one control.

Inputs: enabled (required), track_index (required), device_index (required). Returns: result

set_device_parameterA

Set a parameter after reading get_device_parameters. Numbers use the native range; strings use display units ('250 Hz', '-6 dB') or exact enum labels. Unsupported display mappings fail before writing. Returns Live's actual readback and warnings.

Inputs: value (required), parameter (required), track_index (required), device_index (required). Returns: result

set_device_routingA

Set a device's sidechain audio input. field is "input_routing_type" (the source track/return) or "input_routing_channel" (Pre FX / Post FX / Post Mixer); display_name is one of the options from get_device_routing. For ducking, also enable the device's Sidechain parameter via set_device_parameter.

Inputs: field (required), track_index (required), device_index (required), display_name (required). Returns: result

set_drum_padA

Mute/solo/rename a single Drum Rack pad by its MIDI note (e.g. 36=kick). Per-pad muting = instant beat variations without touching the notes.

Inputs: note (required), track_index (required), device_index (required), mute, +2 more. Returns: result

set_fold_stateA

Fold (collapse) or unfold a group track. Errors if the track is not a group track (see get_group_info).

Inputs: track_index (required), folded. Returns: result

set_groove_amountA

Set the global Groove Amount (0.0 to 1.0), which scales how strongly every clip's assigned groove is applied across the whole set.

Inputs: amount (required). Returns: result

set_loopA
Idempotent

Set the global Arrangement loop region and its enabled state.

start is an absolute beat and length is a positive beat duration. The region is updated even when enabled=false; playback simply will not repeat it. Use set_clip_loop for a Session clip's loop braces, not the Arrangement.

Inputs: start (required), length (required), enabled. Returns: result

set_master_device_parameterA
Idempotent

Set one enabled parameter on a device in the Master track's chain.

Call get_master_device_parameters first and use its parameter name/index and native min/max; numeric out-of-range values are clamped. Display strings with supported units or exact enum labels also work; unsupported mappings fail before writing. This affects the full mix. Read back the resulting display. Use set_device_parameter for a regular track or set_return_device_parameter for a return track.

Inputs: value (required), parameter (required), device_index (required). Returns: result

set_master_volumeA

Set the Main/Master track volume. Range 0.0-1.0 where 0.85 = 0 dB.

Inputs: volume (required). Returns: result

set_metronomeA

Turn Live's metronome click on/off.

Inputs: enabled (required). Returns: result

set_record_modeA
Idempotent

Arm or disarm Live's global Arrangement Record state.

true enables Arrangement recording; false disables it. Enabling alone does not record: arm the intended tracks with set_track_arm, then use start_playback. Use set_session_record or trigger_session_record for Session slots instead. Repeating the current state has no additional effect.

Inputs: enabled (required). Returns: result

set_return_device_parameterA
Idempotent

Set one enabled parameter on a device in a return track's chain.

Call get_return_device_parameters first and use its parameter name/index and native min/max; numeric out-of-range values are clamped. Supported display-unit strings and exact enum labels also work; unsupported mappings fail before writing. Use set_device_parameter for a regular track or set_master_device_parameter for the full-mix chain.

Inputs: value (required), parameter (required), device_index (required), return_index (required). Returns: result

set_scene_nameA
Idempotent

Rename one Session-view scene without changing any clips in the row.

Use set_clip_name for an individual clip. Repeating the same name has no additional effect.

Inputs: name (required), scene_index (required). Returns: result

set_sendA

Set a track's send level (send_index 0 = Return A, 1 = Return B, ...). Range 0.0-1.0.

Inputs: value (required), send_index (required), track_index (required). Returns: result

set_session_automation_recordA

Toggle recording of parameter changes into Session clip envelopes - an alternative to write_automation for captured knob rides.

Inputs: enabled (required). Returns: result

set_session_recordA

Toggle Session record (records into armed tracks' selected Session slots; also enables MIDI overdub into playing clips).

Inputs: enabled (required). Returns: result

set_simpler_playback_modeA

Set a Simpler device's playback mode: 0 = Classic, 1 = One-Shot, 2 = Slicing. Slicing chops the sample into playable slices, central to lofi/hip-hop sampling. Errors if the device at that index is not a Simpler.

Inputs: mode (required), track_index (required), device_index (required). Returns: result

set_song_scaleA

Set the song's key context. root_note 0-11 (0=C, 2=D...), scale_name e.g. "Major", "Minor", "Dorian". Read current values via get_session_info.

Inputs: root_note, scale_name. Returns: result

set_swing_amountA

Set the global swing amount (0.0 to 1.0) used when quantizing with swing.

Inputs: amount (required). Returns: result

set_tempoA

Set the tempo of the Ableton session.

Parameters:

  • tempo: The new tempo in BPM

Inputs: tempo (required). Returns: result

set_time_signatureA
Idempotent

Set the Live set's global time signature, such as 4/4, 3/4, or 6/8.

This changes bar/grid and metronome interpretation but does not rewrite clip notes. Use set_clip_signature instead when only one clip needs a different meter for polymeter. Repeating the same signature has no additional effect.

Inputs: numerator (required), denominator (required). Returns: result

set_track_armA

Arm or disarm a track for recording (fails on tracks that can't be armed, e.g. group/return tracks).

Inputs: arm (required), track_index (required). Returns: result

set_track_colorA

Set a track's color by index (0-69 in Live's palette).

Inputs: color_index (required), track_index (required). Returns: result

set_track_monitoringA

Set input monitoring: 0=In (always hear input), 1=Auto, 2=Off. Required for resampling/bounce workflows.

Inputs: state (required), track_index (required). Returns: result

set_track_muteA
Idempotent

Set one regular track's mute state (true mutes; false unmutes).

Muting silences that track in Live but preserves its clips, devices, and mixer settings. Use set_track_solo to audition a track relative to the rest, or stop_clip when only Session clip playback should stop.

Inputs: mute (required), track_index (required). Returns: result

set_track_nameA

Set the name of a track.

Parameters:

  • track_index: The index of the track to rename

  • name: The new name for the track

Inputs: name (required), track_index (required). Returns: result

set_track_panA
Idempotent

Set one regular track's stereo pan position; 0.0 is centered.

Live clamps the value to the track's native range. This changes mixer state, not the audio file or clip content. Use set_master_device_parameter for a device control, and set_crossfader for A/B crossfading.

Inputs: pan (required), track_index (required). Returns: result

set_track_routingA

Set track routing by display name. field: input_routing_type | output_routing_type | input_routing_channel | output_routing_channel. Enables sidechain sources, bus routing, and resampling (route a track's output into another track's input, arm, record). Get options from get_track_routing first.

Inputs: field (required), track_index (required), display_name (required). Returns: result

set_track_soloA
Idempotent

Set one regular track's solo state (true solos; false unsolos).

Solo changes monitoring relative to other tracks and does not alter clip or device content. Other tracks may remain soloed, so clear them separately when exclusive auditioning is required. Use set_track_mute to silence this track.

Inputs: solo (required), track_index (required). Returns: result

set_track_volumeA

Set track volume. Range 0.0-1.0 where 0.85 = 0 dB (unity gain).

Inputs: volume (required), track_index (required). Returns: result

setup_sessionA

Bootstrap a session in one call: set tempo and create named MIDI tracks, optionally loading an instrument found by name for each.

tracks: [{"name": "Drums", "instrument": "drum kit"}, {"name": "Bass", "instrument": "upright bass"}, ...]. "instrument" is a search_browser query (best match is loaded); omit it to create an empty track. Returns a per-track report with the loaded item names.

Inputs: tempo (required), tracks (required). Returns: result

start_playbackA

Start Arrangement playback from the current song time. NOTE: the response reports the PRE-command state (may say playing=false right after starting) - confirm with get_session_info.

Inputs: no named parameters. Returns: result

stop_all_clipsA

Stop every playing Session clip (does not stop the transport).

Inputs: no named parameters. Returns: result

stop_browser_previewA

Stop the browser preview started by preview_browser_item.

Inputs: no named parameters. Returns: result

stop_clipA

Stop playing a clip.

Parameters:

  • track_index: The index of the track containing the clip

  • clip_index: The index of the clip slot containing the clip

Inputs: clip_index (required), track_index (required). Returns: result

stop_playbackA

Stop playback. NOTE: the response reports the PRE-command state - confirm with get_session_info.

Inputs: no named parameters. Returns: result

switch_to_arrangement_viewA

Switch Ableton's main window to the Arrangement view.

Parameters:

Inputs: no named parameters. Returns: result

tap_tempoA

Tap the tempo once. Tapping repeatedly at a beat sets the tempo to that rate; a single tap nudges the transport. Returns the resulting tempo.

Inputs: no named parameters. Returns: result

transform_clipA
Destructive

Read a MIDI clip's notes, apply a classical motif transformation, and write the result. Overwrites the source clip unless dest_clip_index is given (in which case that destination slot is REPLACED instead).

op (one of): transpose - shift every pitch by amount semitones (integer; may be negative). Timing unchanged; out-of-range pitches clamped. invert - mirror pitches about the first note's pitch (up-motion becomes down-motion). amount is ignored. retrograde - play the motif backwards in time (onset order reversed, durations preserved). amount is ignored. augment - scale start times and durations by amount (a factor > 1 stretches/slows, 0 < amount < 1 compresses/speeds up). Requires amount > 0.

amount therefore means semitones for transpose and a time factor for augment. Returns a short summary of what was written.

Inputs: op (required), clip_index (required), track_index (required), amount, +1 more. Returns: result

trigger_session_recordA

Start Session recording into armed tracks; record_length in beats records a fixed-length loop (e.g. 16.0 = 4 bars) then auto-switches to playback.

Inputs: record_length. Returns: result

undoA

Revert the most recent undoable Live-set action.

This mutates the set and may remove newly created content or restore deleted content. It has no target parameter: Live chooses the top undo-history entry. Use redo to reapply an accidental undo. Prefer batch_commands for related edits so one undo reverts the group.

Inputs: no named parameters. Returns: result

write_automationA

Write clip automation for a device parameter. points = [{"time": beats, "value": v}, ...]. Replaces any existing envelope for that parameter on the clip.

Inputs: points (required), parameter (required), clip_index (required), track_index (required), +1 more. Returns: result

write_drum_gridA
Destructive

Write drums from an ASCII grid (16th-note steps). REPLACES existing notes.

grid: one line per instrument: "name: pattern" where pattern chars are 'X' (accent, vel 105), 'x' (normal, 85), 'o' (soft, 55), '.' or '-' (rest). Line length = number of 16th steps (16 = one 4/4 bar; 32 = two bars...). Instruments: kick snare rim clap chat phat ohat ltom mtom htom crash ride perc shaker, or a raw MIDI number ("36: x...").

Example (one bar of lofi): kick: X..x......X..... snare: ....x.......x... chat: x.o.x.o.x.o.x.o.

Inputs: grid (required), clip_index (required), track_index (required), swing. Returns: result

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.7/5.0

    Scored across 156 tools

    Disambiguation3/5

    Most tools target clearly distinct resources (track vs clip vs device vs return), and the descriptions are exemplary at cross-referencing alternatives. However, several overlapping clusters remain - four drum-pattern writers (generate_drum_pattern, generate_groove, generate_euclidean_drums, write_drum_grid), three chord generators, and two bass generators - where an agent must read carefully to avoid misselection.

    Naming Consistency4/5

    Naming is overwhelmingly systematic: snake_case, verb-first, with predictable family prefixes (get_, set_, create_, delete_, generate_, als_, adg_) across nearly all 156 tools. Minor deviations include noun-style commands (batch_commands, clip_operation, session_diff) and the jump_by vs jump_to_* split.

    Tool Count2/5

    156 tools far exceeds the well-scoped range, and despite Ableton Live's genuinely broad domain, there is significant redundancy (4 drum writers, 3 chord writers, 4 key detectors, 3 parameter-setter variants). This volume imposes heavy context costs and makes every tool selection a high-cardinality decision.

    Completeness4/5

    The surface is remarkably complete: tracks, clips, devices, transport, browser, automation, MIDI generation, live analysis, and offline .als/.adg inspection are all covered with full lifecycles. Minor gaps exist - no final render/export to disk, no arrangement clip move/trim/split, and automation can only be written/cleared, not read back - but agents can work around them.

    Maintenance

    ActivitySlowing
    ResponsivenessOutdated data

    Related MCP Connectors

    Related MCP Servers