Roadmap

Next

• Automation: read and draw automation envelopes, so the AI can shape filter sweeps, volume fades, and other changes over time.
• Copy all parameters from one device to another (e.g. apply the EQ curve from track A to tracks B and C)
• Improved undo/redo support

In consideration:

• Directly support Vercel's AI Gateway in the built-in chat UI
• Fetch available models from Ollama and LM Studio/Bionic servers instead of manually configuring in the built-in chat

Changelog

See the list of releases for more detailed information.

2.4 - Fewer dead ends (September 2026)

When something can't be done, the AI now finds out what and why for each target, and keeps going with the rest.

• One consistent way for every tool to report a target it couldn't do, and why
• Read tools take a list, so several clips, tracks, scenes or devices can be read in one call
• Address a device by what it is instead of where it sits: t0/inst for a track's instrument, t0/afx1 for its second audio effect
• t0[5|1] finds the clip playing at bar 5, even if it started earlier
• Attach images in the built-in chat, so a sketch of a song can become MIDI clips
• Library searches, including similar samples, show what Live's browser shows

The new Producer Pal remote script (experimental) does what the Live API alone can't:

• Put Producer Pal into a freshly opened Live Set
• Add plug-ins and Max for Live devices
• Add a device from a preset, including racks
• Swap a preset onto a device already in the Set

2.3 - Consistency (September 2026)

The tools were built one at a time, each with its own idea of how to name a thing, what to return, and what to refuse, so Producer Pal would sometimes refuse something reasonable or say it had done something it hadn't. This release set down ten principles for how a tool should behave and brought every tool to them.

• Writes report what actually landed, read back off the object. A value Live ignored or snapped no longer comes back as a complete success
• Device parameters read and write in the units Live shows: 12 ms, 0.4 s, 1/16, -1.68 st. A value in the wrong unit or out of range is refused instead of quietly written wrong. Dozens of parameters across Live's devices were being read in the wrong units, or not at all
• An arrangement clip is named by where it starts: t0[5|1], or t0[loc:Chorus]. A locator name works anywhere a bar|beat position does
• Arrangement clips move and copy across tracks and onto take lanes, audio included, even where Live's API has no way to do it directly
• Drum pads and rack chains go deeper: a pad stacking several chains reads, writes, splits and copies one layer at a time, pads are spelled by pitch everywhere, and any rack chain can be copied with its devices and mixer
• A track's or a rack chain's sends can be set together in one call, and report the level each one landed on
• Every result says where its object is, and every warning names its object by both path and id, in the spelling you used
• Batch operations do the shared work once instead of once per item. Splitting 40 arrangement clips at 32 points went from 13.4s to 4.6s
• Built-in chat: leaving a conversation mid-response asks first, deleting a voice conversation is undoable, and several other ways a conversation could be lost are fixed
• Command-line flags and the Claude Desktop extension's settings apply to that client alone instead of changing what everyone connected sees

2.2 - Performance and paths (August 2026)

• Producer Pal no longer leaks Live API objects, so Ableton stays fast over a long session instead of slowing down and filling its log file
• Drum kit, Live Set, and arrangement operations do a fraction of the work they did; heavy arrangement edits could freeze Live
• One shorthand syntax locates anything in a Live Set: t2/s3 is a clip slot, t2/l0 a take lane, t1/d0/pC1 a drum pad. path replaces slot, and clip results report it. Breaking for anything reading slot or trackIndex
• Every id param is just id, and a param sent as null reads as unset
• Drum Racks: copy a pad, delete a single chain, read and write chain mixers, and nested racks work
• Copy or move a clip to another track in the arrangement, and split arrangement clips at song positions
• Calls that run long report what landed instead of failing outright
• Built-in chat: the input is a markdown editor, and the per-turn tool-step budget is 25 and configurable

2.1 - Delegation (August 2026)

• Subagents: the built-in chat can delegate self-contained tasks to nested assistants that run in parallel, and be resumed for follow-up work
• Presets: named bundles of provider, model, tool set, and notation, including what subagents run as
• Skills fragments can be switched off individually, and are dropped automatically for tools you've turned off
• Notation and the tool set are pinned to each conversation
• Project context is backed up next to your Live Set, so it survives a device upgrade
• Audio clips: control warping when creating a clip, and correct timing for unwarped clips
• Voice mode on GPT Realtime 2.1, with cheaper transcription

2.0 - Personalization (July 2026)

• Global context: add your own reference material and custom instructions across all Live projects
• Global memory: Producer Pal can adapt over time to your needs and interaction style
• Customize the built-in skills and the built-in chat's system instructions
• Alternative MIDI syntaxes
• Context manager in the built-in chat UI for working with all the above

1.4 - MIDI Transforms, REST API, and Live API Access (February 2026)

MIDI transforms: math expressions for transforming note properties

• Ramps, curves, and LFO shapes (arrangement-relative or clip-relative)
• Randomization with arbitrary ranges, or choose from a set of values
• Helpers: swing(), quant(), legato(), seq(), clipseq(), step(), wrap(), reflect()
• Context variables like note index and clip position in the arrangement

REST API and agent skills:

• HTTP-based tool calls without MCP setup
• Drop-in agent skill for Claude Code, Codex CLI, Gemini CLI, and other SKILL.md-compatible runtimes

New tools and broader Live API coverage:

• ppal-library for searching the Live Library: samples, MIDI clips, and installed plugins (filterable by category, type, and folder)
• ppal-live-api for direct Live API access to cover gaps in specialized tools
• Simpler sample loading via ppal-update-device (requires Live 12.4+)
• Device-specific parameter control for 9 native devices (Drift, Wavetable, Simpler, Meld, Compressor, EQ Eight, Hybrid Reverb, Roar, Spectral Resonator), with auto-discovered device-specific actions and degree (°) unit display
• Take lane support

Other improvements:

• Split arrangement clips at specified positions
• Multi-object create / update / duplicate operations. transforms on update-clip and duplicate is a single string broadcast across every clip/copy. Use clip.index arithmetic or clipseq() inside the string for per-clip variation, or make separate calls for structurally-distinct edits.
• Per-project notes: improved UI, now always enabled by default (disable the ppal-context tool to prevent AI edits)
• Configurable Ollama and LM Studio URLs for remote hosting
• Experimental voice control in the chat UI (OpenAI realtime and Google Gemini Live), with adjustable voice volume up to 125%
• Many built-in chat UI improvements: conversation history, token usage reporting, grouped tool calls, message editing, encrypted API keys, a "Continue" button to resume long tasks, and more

1.3 - Device Control (January 2026)

• Full device control: add/delete/move native devices on any track, read/write parameters, insert into rack chains
• Rack macro and variation management
• A/B comparison for device parameters

Also added support for:

• Arrangement locators
• MIDI clip quantization

1.2 - Audio clip, mixer, and improved Arrangement support (November 2025)

• Audio clip support with a read-samples tool to scan folders for samples (since absorbed into ppal-library)
• Track mixer control: gain, panning, and sends
• Arrangement clip positioning and length control

1.1 - Built-in chat UI (November 2025)

Direct API integrations for Google Gemini, OpenAI, Mistral, OpenRouter, LM Studio, Ollama, and more.

1.0 - Support for more LLMs (October 2025)

Expanded features and support for multiple AI platforms.

0.9 - Public Beta with Claude Desktop (July 2025)

Initial public release with Claude Desktop support and a focus on MIDI clip manipulation and basic Live Set management.