opendaw-mcp 1.11.3

258 MCP tools for agent-native control of openDAW — a browser-based digital audio workstation

openDAW MCP

258 MCP tools for agent-native control of openDAW — a browser-based digital audio workstation.

This project wraps openDAW's internal box system and project API behind a Model Context Protocol server, allowing AI agents (Claude, GPT, Hermes, etc.) to create and manipulate music projects programmatically — tracks, instruments, effects, MIDI, automation, audio regions, rendering, and more.

How It Works

AI Agent ──MCP──▶ Python Server ──Playwright──▶ Headless Chromium ──▶ openDAW (Vite dev server)

The MCP server launches a headless Chromium instance loaded with openDAW, then communicates via page.evaluate() calls into the DAW's V8 context. Every tool performs real operations on the live project — no stubs, no mocks.

Features

• Track & Region CRUD — create/delete/move audio, note, and automation tracks with regions
• Instrument Control — Vaporisateur ( polysynth), Nano, Tape, Soundfont, Playfield (drum machine), MIDI output
• Effects — Delay, Reverb, Compressor, Equalizer, Saturation, Waveshaper, Stereo, Vocoder, NeuralAmp, Maximizer
• MIDI Effects — Arpeggio, Pitch, Velocity, Zeitgeist, Spielwerk (scriptable)
• Scriptable Devices — Apparat (instrument), Werkstatt (audio effect), Spielwerk (MIDI effect) with JS code compilation, full @param mapping metadata, and range-validated parameter setting
• Automation — event creation, interpolation modes, tempo/signature changes
• Audio Operations — region fades, gain, time/pitch stretch, warp markers, play modes
• Mixing — send/return routing, buses, volume, solo/mute, mixer state inspection
• Rendering — offline stem export with LUFS targeting, full mix render
• Clips — session view clip CRUD, clone, consolidate, playback settings
• Groove — groove shuffle amount and duration control
• Presets — export/import audio unit presets, effect chain presets
• Transfer — move regions and audio units between projects
• Project Info — tempo map conversion (PPQN↔seconds), duration, validation, sample listing
• Notes — pitch range, overlapping detection, advanced properties (chance, cent, playCount, playCurve)
• Modular System — voltage modules (Gain, Delay, Multiplier, AudioInput, AudioOutput), patch cable connections
• Piano Mode — global transpose, keyboard type, note scale, time range
• Project & Bus Metadata — creation date, signature, AU/track count, bus labels and colors
• Debugging & Control — screenshots, condition polling, raw JS evaluation

Quick Start

Prerequisites

• Python 3.11+
• Node.js 20+ (for openDAW dev server)
• Chromium (Playwright will install it)

Install

From PyPI:

pip install opendaw-mcp
playwright install chromium

From source:

git clone https://github.com/ameobius/opendaw-mcp.git
cd opendaw-mcp
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt
playwright install chromium

Set Up openDAW

Clone and build openDAW separately:

git clone https://github.com/andremichelle/openDAW.git
cd openDAW
npm install
npm run build   # or: npx turbo run build

Create a headless host page (see headless-daw/ for reference implementation).

Run

# Terminal 1: Start openDAW dev server
cd openDAW
npm run dev     # typically serves on http://localhost:5174

# Terminal 2: Start MCP server (stdio transport, default)
cd opendaw-mcp
source venv/bin/activate
python server.py

SSE Transport

For remote deployments and registry introspection (e.g. Glama):

MCP_TRANSPORT=sse FASTMCP_HOST=0.0.0.0 FASTMCP_PORT=8080 python server.py

CLI

python server.py --help        # show usage and env vars
python server.py --version     # print version and tool count
python server.py --list-tools  # list all 260 registered MCP tools

Docker

# Pull pre-built image from GitHub Container Registry
docker pull ghcr.io/ameobius/opendaw-mcp:1.10.0
docker run -p 8080:8080 ghcr.io/ameobius/opendaw-mcp:1.9.6
# MCP server available at http://localhost:8080/sse

# Or build from source
docker build -t opendaw-mcp .
docker run -p 8080:8080 opendaw-mcp

The Docker image bundles openDAW (built from source), Vite dev server, Chromium, and the MCP server. The entrypoint starts Vite, waits for it to be ready, then launches the MCP server in SSE mode.

Claude Desktop / MCP Client Config

Add to your MCP client config (e.g. claude_desktop_config.json):

{
  "mcpServers": {
    "opendaw": {
      "command": "python",
      "args": ["path/to/opendaw-mcp/server.py"],
      "env": {
        "OPENDAW_HOST_DIR": "path/to/headless-daw",
        "OPENDAW_URL": "http://localhost:5174",
        "OPENDAW_EXPORT_DIR": "path/to/exports"
      }
    }
  }
}

See mcp.json in the repo for a reference config.

Environment Variables

Variable	Default	Description
OPENDAW_HOST_DIR	../headless-daw	Path to the headless openDAW host directory
OPENDAW_URL	http://localhost:5174	URL of the running openDAW instance
OPENDAW_EXPORT_DIR	../exports	Directory for rendered audio exports
NODE_BIN_DIR	(from PATH)	Path to Node.js binary directory (if not on PATH)
MCP_TRANSPORT	stdio	Transport protocol: stdio or sse
FASTMCP_HOST	127.0.0.1	SSE server bind host
FASTMCP_PORT	8000	SSE server bind port

Architecture

HeadlessDawBridge

The HeadlessDawBridge class manages the Playwright lifecycle:

1. Launches headless Chromium with audio autoplay enabled
2. Navigates to the openDAW URL
3. Waits for window.DAW and factory globals to load
4. Injects DAW_HELPERS — JS utility functions that eliminate boilerplate across tools

DAW_HELPERS

JavaScript helpers injected into the DAW context:

• au(i) — get audio unit adapter by index
• track(auIdx, trackIdx) — get track adapter
• region(au, track, reg) — get region adapter
• instrumentAU() — get the first instrument audio unit
• modify(fn) — wrapper around editing.modify() for box mutations
• allAUs() — list all audio unit adapters

Tool Structure

Each MCP tool follows the pattern:

@mcp.tool()
async def tool_name(param: str) -> str:
    """Description."""
    async def _run():
        result = await bridge.page.evaluate("""...JS...""")
        return json.dumps(result)
    return await bridge.run(_run)

The bridge is a singleton — state persists within a single Python process. All box mutations go through editing.modify() as required by openDAW's transactional model.

Input Sanitization

All string parameters that are interpolated into JavaScript template literals are sanitized before evaluation — quotes, backslashes, and braces are stripped to prevent JS injection. Numeric parameters use proper int/float type annotations for FastMCP type coercion.

DSP Scripts

The scripts/ directory contains 17 example DSP scripts (12 Werkstatt + 3 Apparat + 2 Spielwerk):

Script	Device	Description
werkstatt_darksat.js	Werkstatt	Tape saturation with drive, bias, tone, mix, output gain
werkstatt_coldfold.js	Werkstatt	Wavefolding + bitcrush + slew rate reduction
werkstatt_paulstretch.js	Werkstatt	Extreme time-stretch via FFT/ISTFT overlap-add
werkstatt_envfollower.js	Werkstatt	Envelope follower with sidechain ducking
werkstatt_chorus.js	Werkstatt	Stereo chorus with LFO-modulated delay lines
werkstatt_reverb.js	Werkstatt	Algorithmic reverb with comb + allpass filters
werkstatt_shimmer.js	Werkstatt	Pitch-shimmer reverb with granular pitch shift
werkstatt_phaser.js	Werkstatt	Phaser with LFO-swept allpass filter chain
werkstatt_subcrusher.js	Werkstatt	Sub-bass enhancement with glide and distortion
werkstatt_lookahead.js	Werkstatt	Lookahead limiter with gain reduction metering
werkstatt_adsr_trim.js	Werkstatt	ADSR envelope trim for sustained samples (#241)
werkstatt_granular_stretch.js	Werkstatt	Granular time-stretch with Hann window + pitch shift (#201)
werkstatt_pitch_shift.js	Werkstatt	Real-time pitch shifter via delay-line sweep with crossfade (#188)
apparat_darkbass.js	Apparat	Bass synth with sub oscillator and filter envelope
apparat_coldlead.js	Apparat	Lead synth with detune and vibrato
apparat_subcrusher.js	Apparat	Sub-bass synth with glide and distortion
spielwerk_powerchord.js	Spielwerk	MIDI effect that generates power chord harmonies
spielwerk_arpeggiator.js	Spielwerk	MIDI arpeggiator with swing and octave range

Examples

The examples/ directory contains 9 Python scripts demonstrating the full workflow:

Example	Description
create_beat.py	Drum beat with Playfield
create_chord_progression.py	Chord progression with Vaporisateur
mix_workflow.py	Mixing: levels, effects, sends
render_stems.py	Stem export with LUFS targeting
automation_sweep.py	Filter cutoff automation
modular_patch.py	Modular system with patch cables
full_production_pipeline.py	Complete track: synth + drums + DSP + automation + render
scriptable_devices_demo.py	All 3 scriptable device types: Apparat synth + Werkstatt DSP + Spielwerk MIDI

Tool Catalog

See TOOL_CATALOG.md for the complete list of 260 tools with parameters and descriptions.

Orchestration Tools

High-level composers that combine multiple low-level operations into a single call — designed for agents to reduce token usage and round-trips:

Tool	Replaces	Example
create_notes_batch	10-50 × create_note	'[{"pitch":60,"start":0,"duration":0.5},...]'
create_drum_pattern	10-20 × create_note	'{"kick":"x...x...","hihat":"....o..."}'
create_chord_progression	15-50 × create_note	'[["C","min7"],["F","dom7"]]'
add_mastering_chain	3 × add_effect + 10 × set_param	style="warm", target_lufs=-14
create_genre_track	20-40 low-level calls	genre="lofi" → synth + drums + bass + chords

Mastering

The MCP server includes a full mastering chain for streaming-ready output:

1. Render — render_full (full mixdown) or export_stems (per-track stems)
2. Measure LUFS — measure_lufs (ITU-R BS.1770-4 K-weighting, gated mean squares)
3. Auto-gain — auto_gain (iterative: render → measure → adjust Maximizer threshold + output volume → re-render, converges ±1 LUFS)

# Render full mix
await server.mcp_opendaw_render_full("my_mix", 48000)

# Measure loudness
lufs = json.loads(await server.mcp_opendaw_measure_lufs("my_mix"))
# → {"lufs_integrated": -18.3, "true_peak_db": -3.71, ...}

# Auto-gain to Spotify target (-14 LUFS)
result = json.loads(await server.mcp_opendaw_auto_gain("-14", "mastered", 48000, "3"))
# → converges to -13.7 LUFS in 3 iterations

Platform targets: Spotify/YouTube -14 LUFS, Apple Music -16 LUFS, Tidal -14 LUFS.

Limitations

• Headless only — some UI-dependent features (file dialogs, popup-based model loading) are not available
• Single process — bridge state doesn't persist across Python process restarts
• Upstream coupling — tools depend on openDAW's internal box system, which can change between versions
• AU rename — requires adapter context not available in headless mode

Acknowledgments

• andremichelle — creator of openDAW, an incredible browser-based DAW
• Model Context Protocol — the protocol that makes agent-native tools possible

License

Apache-2.0 — see LICENSE

Changelog

v1.11.3 (2026-07-04)

• 1 new Werkstatt script: real-time pitch shifter via delay-line sweep (#188)
• Ruff lint fixes: removed unused imports/variables in examples and midi_parser
• 17 DSP scripts total (12 Werkstatt + 3 Apparat + 2 Spielwerk)

v1.11.2 (2026-07-04)

• 10 DSP bug fixes synced from upstream PR #283 CodeRabbit review:
  • darksat: undefined outR variable fix, DC blocker signal path corrected
  • chorus: delay buffer 2× for depth modulation, safe modulo for negative indices
  • coldfold: slew parameter /100 scaling removed (was disabling the parameter)
  • lookahead: gain reduction now applied to delayed signal (true lookahead)
  • reverb: comb filter indices advancing, per-comb damping state, M/S stereo width decode
  • shimmer: per-channel pitch shifter state (eliminates stereo crosstalk)
  • phaser: stable 1st-order allpass topology (2nd-order was unstable)
  • subcrusher: bidirectional glide (was diverging on upward glides)
  • arpeggiator: swing notes no longer dropped at block boundaries
• 14 DSP scripts total (9 Werkstatt + 3 Apparat + 2 Spielwerk) — all CodeRabbit issues addressed
• 2 new Werkstatt scripts: ADSR trim (sustained sample trimming, #241) + granular time-stretch (Hann window overlap + pitch shift, #201)
• 1 new Werkstatt script: real-time pitch shifter via delay-line sweep (#188)
• 17 DSP scripts total (12 Werkstatt + 3 Apparat + 2 Spielwerk)

v1.11.1 (2026-07-04)

• Scriptable device mapping info — list_script_params now returns full @param mapping metadata (min, max, mapping type, unit) via ScriptDeclaration.parseParams()
• Range validation — set_script_param now validates values against @param declarations: bool snaps to 0/1, int rounds+clamps, linear/exp clamps to [min, max]. Returns clamped flag and range info
• _clamp_script_param Python helper mirrors JS-side clamping logic
• +15 unit tests (93 total) — TestScriptParamClamping: linear/exp/int/bool/unipolar clamping, rounding, snapping
• +6 integration E2E tests — bridge startup, globals, track ops, scriptable compile, param clamping, latency benchmark (avg 4ms round-trip)
• 5 new Werkstatt DSP scripts — reverb (Schroeder plate), chorus (stereo dual-LFO), phaser (allpass cascade), lookahead compressor (soft knee), shimmer delay (granular pitch shift). Total: 12 scripts
• DAW_ScriptDeclaration added to headless-daw globals

v1.11.0 (2026-07-04)

• apply_mix_preset — 8th orchestration tool: batch volume/pan/mute/solo across all tracks. Named presets (lofi, house, balanced, wide) or custom JSON
• 258 MCP tools (250 low-level + 8 orchestration)

v1.10.2 (2026-07-04)

• 24 new unit tests for orchestration tools (78 total) — curve interpolation, chord theory, drum pattern parsing, song structure parsing
• TOOL_CATALOG fully synchronized — 0 discrepancies with server.py

v1.10.1 (2026-07-04)

• 260 MCP tools (253 low-level + 7 orchestration)
• create_song_structure — arrangement markers (intro/verse/chorus/bridge/outro) from JSON section list
• automation_sweep — smooth automation ramps with linear/exp/log curves. Replaces 10-30 create_automation_event calls
• PyPI v1.10.1 published — pip install opendaw-mcp
• Both orchestration tools tested end-to-end via Playwright bridge
• 54 tests, ruff clean, CI green

v1.10.0 (2026-07-04)

• 7 orchestration tools — high-level composers for agents:
  • create_notes_batch — batch MIDI note creation (JSON array, one round-trip)
  • create_drum_pattern — step-sequencer notation (x...x... → drum beat)
  • create_chord_progression — chord names → auto-voiced notes ([["C","min7"]])
  • add_mastering_chain — EQ + Compressor + Maximizer with style presets
  • create_genre_track — full genre starting point (house/techno/lofi/dnb/trap/ambient)
  • create_song_structure — arrangement markers from JSON
  • automation_sweep — smooth parameter ramps with linear/exp/log curves
• 260 total tools (250 low-level + 7 orchestration + 3 internal)
• set_metronome — dedicated metronome control (enabled, gain, beat_subdivision)
• Module-level lookup tables — TIDAL_RATE_MAP, DELAY_SYNC_MAP, WAVESHAPER_FUNCS, REVAMP_SECTIONS extracted for testability
• +23 new unit tests (54 total) — fraction maps, waveshaper funcs, revamp sections, safe_filename edge cases
• Official ScriptCompiler migration — set_script_device_code now uses the real ScriptCompiler from @opendaw/studio-adapters instead of custom @param/@sample parser. Benefits: declaration caching (WeakMap), proper sample file cleanup, label parsing, correct worklet wrapping
• Stems export fix — useInstrumentOutput changed from True→False. Stems now route through channel strip (effects, sends, volume/pan) as documented by naomiaro/opendaw-test
• export_dry_stem — new tool for freeze/flatten/re-amp workflows: captures raw instrument output before effects
• set_waveshaper_equation — 6 transfer functions (hardclip/cubicSoft/tanh/sigmoid/arctan/asymmetric)
• set_crusher_crush — sample-rate reduction with documented crush inversion semantics
• set_revamp_filter — 7 EQ sections (highpass/lowshelf/lowbell/midbell/highbell/highshelf/lowpass) with enabled/freq/gain/q/order
• set_tidal_rate — musical fraction string → Tidal LFO rate index (17 entries)
• set_delay_sync — musical fraction string → Delay synced time index (21 entries, includes "off")
• Effect lookup case-insensitive — add_effect("werkstatt") now works alongside Werkstatt
• naomiaro/opendaw-test research — 543 commits, 17 SDK doc chapters used as authoritative reference for effect parameters and box field names
• 54 tests, ruff clean, CI green

v1.9.6 (2026-07-03)

• measure_lufs refactored — 223 lines → 20 lines. Extracted _parse_wav() and _compute_lufs() helpers
• DRY: K-weighting coefficients — duplicated if/else branches (48kHz vs else) were identical, merged into single computation
• 9 new unit tests — WAV parsing (float32/mono/stereo/PCM16/invalid/no-data) + LUFS computation (silence/full-scale/low-level/stereo)
• Social preview banner — custom OpenGraph image for GitHub link previews
• awesome-mcp PR updated — title and body synced to 255 tools
• GitHub topics — 18 topics for discoverability
• 31 total tests, ruff clean, CI green

v1.9.5 (2026-07-03)

• CLI commands — --version, --list-tools, --help with full env var reference
• 31 unit tests — pytest covering _ok, _err, _wrap_eval, _unwrap_eval, _safe_filename, _safe_path, _parse_wav, _compute_lufs
• 3 bug fixes found by tests:
  • _ok() — {"success": False} in data overwrote the True flag (security fix)
  • _safe_filename() — case-sensitive extension stripping (.MP3 not stripped)
  • _safe_filename() — Windows backslash path traversal not handled on Linux
• CI enhanced — now runs pytest (54 tests) alongside syntax/AST/smoke/ruff checks
• PEP 561 — py.typed marker for type checker support
• Mastering pipeline example — full chain: render → measure LUFS → auto-gain → stems → MP3
• 16 examples total — all syntax-validated
• 255 total tools (added export_dry_stem for freeze/flatten workflows)

v1.9.4 (2026-07-03)

• Removed 2 duplicate tools (245 → 243)
  • delete_signature_event — superseded by delete_signature_change (richer: position match + index, returns updated event list)
  • list_aux_sends — superseded by list_sends (richer: target_bus_name, send_level_db, routing, send_pan via box-level access)
• TOOL_CATALOG.md regenerated from AST — all 255 tools with descriptions, 32 categories
• server.json Docker tag fixed — was stale 1.0.0, now matches release version
• 255 total tools

v1.9.3 (2026-07-03)

• DRY refactoring complete: 17 DAW_HELPERS, ~295 replacements, 0 raw enumeration patterns
  • New helpers: markerBoxes, sendBoxes, busBoxes, sampleBoxes, noteTrackBoxes, clipBoxes, rootClipBoxes, scriptParams, scriptSamples, chainBoxes
  • All pointerHub.incoming() enumeration patterns replaced across 245 tools
  • CONTRIBUTING.md updated with full 17-helper reference table
  • 6 DRY commits, 0 regressions, CI green

v1.9.2 (2026-07-03)

• DRY refactoring: 113+ tools migrated to h.allAUBoxes() / h.auBox() helpers
  • Replaced 133 occurrences of raw [...rootBox.audioUnits.pointerHub.incoming()].map(({box}) => box).sort(...) boilerplate with h.allAUBoxes() across 113+ MCP tools
  • Box-level helpers eliminate ~3000 lines of duplicated AU enumeration code
  • E2E verified: allAUBoxes returns sorted array, auBox(i) returns box by index, count matches raw, box identity matches
• Security hardening
  • Transport action enum validation (play, stop, toggle) — prevents JS injection
  • duplicate_effect chain_type enum validation (audio, midi)
  • _safe_filename() + _safe_path() helpers — os.path.basename() sanitization + path traversal protection on 6 render/export locations
  • _unwrap_eval bare except → json.JSONDecodeError
• 245 total tools (no tool count change — refactoring only)

v1.9.1 (2026-07-03)

• 2 new generic tools: Boolean & Integer effect parameter setters
  • set_effect_parameter_bool(unit_index, effect_index, parameter_name, value) — Generic boolean field setter. Covers Compressor (lookahead, automakeup, autoattack, autorelease), Gate (inverse), Maximizer (lookahead), StereoTool (invertL, invertR, swap), NeuralAmp (mono)
  • set_effect_parameter_int(unit_index, effect_index, parameter_name, value) — Generic integer field setter. Covers Vocoder (bandCount), StereoTool (panningMixing), Fold (overSampling), Crusher (bits), Delay (version). Device-specific tools are preferred when available.
• 245 total tools

v1.9.0 (2026-07-03)

• 6 new tools: Device-Specific Parameters & NeuralAmp Model Loading
  • set_neuralamp_model(unit_index, effect_index, model_json, label, pack_id) — Load NAM/Tone3000 model JSON directly into a NeuralAmp effect, bypassing the popup-based Select Flow. Creates NeuralAmpModelBox and links it via pointer
  • set_vocoder_modulator_source(unit_index, effect_index, source) — Set Vocoder modulator source: noise-white, noise-pink, noise-brown, self, or external
  • set_vocoder_band_count(unit_index, effect_index, band_count) — Set Vocoder filter band count (8-32)
  • set_stereo_tool_panning(unit_index, effect_index, panning_mixing) — Set StereoTool panning law (linear, equal-power)
  • set_fold_oversampling(unit_index, effect_index, oversampling) — Set Fold wavefolder oversampling (0=off, 1=2x, 2=4x)
  • set_crusher_bits(unit_index, effect_index, bits) — Set Crusher bit depth (1-16)
• 255 total tools

v1.8.2 (2026-07-03)

• 2 new tools: Audio Region Time Base & Waveform Offset
  • set_audio_region_time_base(unit_index, track_index, region_index, time_base) — Switch duration interpretation between 'musical' (PPQN, tempo-following) and 'seconds' (fixed wall-clock)
  • set_audio_region_waveform_offset(unit_index, track_index, region_index, offset) — Set waveform display offset for visual alignment
• PR #280 closed — andremichelle confirmed it's our bundler setup issue, not upstream. Closing as requested.
• 237 total tools

v1.8.1 (2026-07-03)

• 3 new tools: Warp Marker CRUD — create_warp_marker, delete_warp_marker, update_warp_marker
  • create_warp_marker(unit_index, track_index, region_index, position_beats, seconds) — Add warp marker to stretched audio regions
  • delete_warp_marker(unit_index, track_index, region_index, marker_index) — Delete non-anchor warp marker
  • update_warp_marker(unit_index, track_index, region_index, marker_index, position_beats, seconds) — Update warp marker position/seconds (-1 = unchanged)
  • Enables agent-driven tempo matching: programmatically pin audio regions to musical positions
• 235 total tools

v1.8.0 (2026-07-03)

• 3 new tools: MP3/FLAC Audio Conversion — convert_audio, render_full_format, export_stems_format
  • convert_audio(filename, format, bitrate, quality) — WAV→MP3/FLAC via system ffmpeg
  • render_full_format(filename, format, bitrate) — render + convert in one step
  • export_stems_format(filename_prefix, format, bitrate) — stems + convert each
  • Uses system ffmpeg (4.4.2), not browser WASM — more reliable in headless mode
  • E2E: WAV 1.01MB → MP3 0.11MB (ratio 0.106) → FLAC 0.19MB (ratio 0.194) ✅
• Bugfix: operator precedence in 14 division+nullish coalescing expressions
  • X / Quarter ?? 0 → (X ?? 0) / Quarter — prevents NaN→null for position/duration fields
• Improvement: get_track_info now includes exclude_piano_mode field
• 232 total tools

v1.7.4 (2026-07-03)

• 3 new tools: Engine Sleep/Wake + Loading Check — engine_sleep, engine_wake, query_loading_complete
  • engine_sleep() — suspend audio processing to save CPU during non-audio operations
  • engine_wake() — resume audio processing
  • query_loading_complete() — check if all audio samples are loaded and ready
  • E2E: sleep ✅, wake ✅, loading (loaded=false, is_ready=true) ✅
• Improvement: get_effect_chain + get_midi_effect_chain — now return short type names (Delay instead of DelayDeviceBox), plus class, minimized fields

v1.7.3 (2026-07-03)

• 2 new tools: Instrument Automation — add_instrument_automation, list_automatable_fields
  • Automate any instrument parameter: Vaporisateur cutoff/volume/ADSR, Tape flutter/wow, Playfield sample mute
  • Per-sample targeting via sample_index for Playfield
  • list_automatable_fields introspects Pointers.Automation support (18/23 on Vaporisateur)
  • Addresses upstream #269 (playfield mute automation) via MCP
  • E2E: Vaporisateur cutoff automated (3 events) ✅, 18 automatable fields ✅

v1.7.2 (2026-07-03)

• 3 new tools: Effect Duplication + Instrument Automation — duplicate_effect, add_instrument_automation, list_automatable_fields
  • duplicate_effect — duplicate single effect in-place with all params copied (audio or MIDI chain)
  • add_instrument_automation — automate any instrument parameter (Vaporisateur cutoff/volume, Tape flutter, Playfield sample mute, etc). Supports per-sample targeting via sample_index
  • list_automatable_fields — introspect which fields support Pointers.Automation (18/23 on Vaporisateur)
  • Addresses upstream issue #273 (Ctrl+D for audio effects) and #269 (playfield mute automation) via MCP
  • E2E tested: Delay duplicated with params ✅, Vaporisateur cutoff automated (3 events) ✅, 18 automatable fields detected ✅
• Bugfix: transport(action) now respects action parameter — was always toggling, now correctly handles "play", "stop", "toggle"
• Cleanup: removed unused region_type param from set_region_duration and set_region_mute

v1.7.1 (2026-07-03)

• 4 new tools: Engine Control — engine_panic(), get_engine_status(), schedule_clip_play(clip_ids), schedule_clip_stop(track_ids)
  • Panic button for stuck audio, real-time engine monitoring (CPU load, position, BPM, playing state), session view clip triggering
  • E2E tested: get_engine_status ✅, engine_panic ✅

v1.7.0 (2026-07-03)

• 2 new tools: DawProject Interop — export_dawproject(filename), import_dawproject(filename)
  • Export/import projects in .dawproject format (Bitwig, Ableton, rePitch compatible)
  • ZIP containing project.xml, metadata.xml, and audio samples
  • Enables cross-DAW workflow: create in openDAW → export to Bitwig, or import Bitwig project → render in openDAW
  • DawProject + DawProjectImport exposed as globals in headless-daw
  • E2E tested: export (2442 bytes, valid ZIP) → import (7 boxes, round-trip OK)

v1.6.2 (2026-07-03)

• 2 new tools: Studio Settings — get_studio_settings(), set_studio_setting(category, key, value)
  • Read/write StudioPreferences: engine, visibility, editing, debug, storage, time-display, pointer
  • Control auto-create-output-maximizer, overlapping-regions-behaviour, enable-beta-features, auto-delete-orphaned-samples, note-audition-while-editing, and more
  • StudioPreferences exposed as DAW_StudioPreferences global in headless-daw

v1.6.1 (2026-07-03)

• 4 new tools from DAW globals research — set_unit_minimized, list_aux_sends, capture_realtime, get_sample_info
  • Mixer minimize, aux send listing, realtime audio capture, sample metadata

v1.6.0 (2026-07-03)

• DAW_HELPERS refactoring — all 180 tools with const p = window.DAW migrated to const h = window.DAW_HELPERS pattern
  • Eliminated boilerplate: AU list enumeration, sort, editing.modify wrapping
  • Fixed 19+ pre-existing bugs: setPosition (api→engine), 8x missing .sort() on AU lists, 9x Quarter=960 hardcode → h.ppqn.Quarter, 2x Python/JS scope leaks
  • DAW_HELPERS provides: h.au(i), h.track(), h.region(), h.modify(), h.allAUs(), h.ppqn, h.uuid, h.rootBox, h.api, h.editing, h.boxGraph, h.tempoMap, h.rootBoxAdapter, h.project
• E2E verified: 23/23 tests passing

v1.5.2 (2026-07-02)

• Sanitization: all string parameters sanitized against JS injection
• Documentation: README badges, Docker, SSE, examples, mastering
• CI: GitHub Actions with AST tool count verification

v1.5.0 (2026-07-01)

• Modular system: 6 MCP tools for patchable modular synthesizer
• PianoMode: 6 MCP tools for piano roll view control
• Freeze/unfreeze: pre-render AU to save CPU
• Preset save/load: export/import AU as base64 preset
• Transfer regions/AUs: deep-copy with dependency tracking