Skip to main content
Ableton Studio MCP — AI music production in Ableton Live with Claude: compose, arrange, mix, master, listen, music theory, techno drums

Ableton Studio MCP

The Ableton Live MCP server that lets Claude (or any MCP client) compose, arrange, mix, master, listen to and analyse music in Ableton Live 11 & 12.

By Olivier Cedrati (@cedrati).

License: MIT PyPI Remote Script Live Tests

Quickstart · Features · Tools · FAQ · Roadmap · Changelog · Issues


About

AI music production in Ableton Live. This is a Model Context Protocol (MCP) server and Ableton Remote Script that connect Ableton Live to Claude, Claude Code, Claude Desktop, Cursor and other MCP clients, so an AI assistant can work inside your Live set: write MIDI, program drums, build arrangements, mix with Live's devices, check loudness (LUFS) and true peak, and reason about harmony.

Ableton Studio MCP, written and maintained by Olivier Cedrati, is a full production assistant. Beyond writing notes, the assistant can mix (faders, sends, return and main-track effects, routing), listen to the result and judge it like a mastering engineer (LUFS, true peak, tonal balance, stereo, a spectrogram it can look at), and compose with real music-theory knowledge (scales and modes, style sheets, key and chord analysis, clip transformations, voice-led progressions).

It is developed on the feat/mixer-scenes branch of cedrati/ableton-studio-mcp. It is published on PyPI as ableton-studio-mcp. (The ableton-mcp package on PyPI is the original project and does not include any of this.)

No data is collected: no telemetry, no analytics, nothing sent anywhere (see Privacy).

Features

Area What the assistant can do
Mixing Volume, pan, sends, mute, solo, arm on tracks, return tracks and the main track. Load effects on return tracks and the main track (a shared reverb, a limiter). Read every parameter as Live displays it (Bell, 1.56 kHz, -6.0 dB).
Routing Set track inputs, outputs and monitoring: layer one MIDI part on several instruments, resample the main output, route to a return.
Listening & mastering analysis Record the main output in real time and get a report: loudness per ITU-R BS.1770 / EBU R128 (integrated, short-term, momentary, loudness range), true peak, PLR, tonal balance against a reference curve, resonances, stereo width per band, low-end mono, each track's meter level, a spectrogram image — and ranked findings with fixes for a delivery target (streaming, club, classical, broadcast, podcast).
Drums Program a Drum Rack from 19 electronic patterns (techno, house, garage, breaks, d'n'b...) or your own step grids: pads found by name, Euclidean and polyrhythmic layers, ratchets, ghost notes, per-note probability, fills, swing, Live's Groove Pool.
Long-form evolution & arrangement Curves over 16–64 bars (filter builds, reverb swells, slow LFOs, phrase steps) on extended clips; whole Arrangements built from Session clips by section (DJ-friendly techno, compact 64 bars, minimal, house) with locators — the clip envelopes play as track automation.
Music theory ~50 scales and modes with their character and typical styles; 24 style sheets (baroque to techno); key, mode and chord analysis of a clip with roman numerals; 17 clip transformations (mode change, diatonic transposition, harmonization, arpeggios, voicings, swing...); voice-led chord progressions; Live 12's song scale.
Clips Notes in Arrangement clips, partial note removal, loop brace and markers, warp (modes, markers, conform a loop to the tempo), clip automation envelopes.
Files Import Standard MIDI Files into clips; analyse any WAV/AIFF.
Session Duplicate and delete tracks; list, create, rename, duplicate, launch and delete scenes; delete Arrangement clips.
Devices & browser Know every device in the set and what it does (native devices from the Live 12 manual, Max for Live, plug-in families) and every installed plug-in; route a Compressor's sidechain; plug-in presets; Simpler sample loading and slicing; rack variations; drum choke groups. Plugins (VST, VST3, AU), User Library and Packs load like built-in devices.
Reliability Playhead moves, locator creation and device loading reply only once Live has applied them; version handshake between server and Remote Script; capability checks so a newer server never sends a command an older script would misread.

Everything is covered by automated tests (321) and checked end to end in Ableton Live 12. See the Changelog for the details of each release.


Quickstart

Requirements

  • Ableton Live 11 or 12 (Live 12 for the song scale; Live 10 works for the original features)
  • uv — on macOS: brew install uv
  • An MCP client: Claude Code, Claude Desktop, Cursor...

Python (3.10+) and the dependencies, including numpy, are installed automatically by uv.

1. Add the server to your MCP client

Claude Code
claude mcp add ableton -s user -- uvx ableton-studio-mcp
Claude Desktop — Settings → Developer → Edit Config
{
  "mcpServers": {
    "AbletonStudioMCP": {
      "command": "uvx",
      "args": ["ableton-studio-mcp"]
    }
  }
}
Cursor — Settings → MCP, as a command
uvx ableton-studio-mcp

Run the server in one client at a time.

2. Install the Remote Script into Live

uvx ableton-studio-mcp install

This copies the AbletonStudioMCP control surface into your User Library's Remote Scripts folder, found from Live's own settings (--list-targets shows where it would go, --target <folder> forces a location, --dry-run changes nothing).

Coming from 1.x? The control surface was called AbletonMCP before 2.0. Deselect it in Live and run uvx ableton-studio-mcp install --remove-legacy to delete the old folder: both listen on the same port.

The server and the Remote Script must come from the same version — always install both from this repository.

3. Enable it in Live

  1. Start (or restart) Ableton Live.
  2. Settings → Link, Tempo & MIDI → Control Surface: choose AbletonStudioMCP, with Input and Output set to None.

4. Check

Run the built-in check, which looks at Python, uv, the script installed in your User Library, the Control Surface setting and the script actually running in Live, and says what to fix:

uvx ableton-studio-mcp doctor

Or ask the assistant to run get_remote_script_info: it should report script_version 2.1.x and up_to_date: true.

Updating

  1. Fetch the new version and install its Remote Script: uvx --refresh ableton-studio-mcp install
  2. Restart Live — re-selecting the control surface does not reload the code.
  3. Restart your MCP client. If it still runs old code: uv cache clean ableton-studio-mcp (with the client closed).
  4. uvx ableton-studio-mcp doctor confirms that everything matches.

To try unreleased changes, use the development branch instead of PyPI: replace ableton-studio-mcp with --from git+https://github.com/cedrati/ableton-studio-mcp@feat/mixer-scenes ableton-studio-mcp in the commands above.


What you can ask

Compose and arrange

  • "Write a C minor organ fugue: subject, answer at the fifth, countersubject, stretto."
  • "Give me a deep house progression in A minor with offbeat stabs, then a rolling bassline."
  • "Analyse the clip on track 2: key, chords, roman numerals."
  • "Re-harmonise this melody in Dorian, add a third above, swing it at 58 %."
  • "Import this MIDI file and transpose it to F#."
  • "Program 8 bars of rolling techno on the 909 kit, with a 5-in-16 percussion layer and a clap roll every 4 bars."
  • "Open the acid bass filter over 16 bars, then lay out a 64-bar techno arrangement with a DJ intro and outro."

Mix and master

  • "Listen to the whole track and tell me if it's ready for streaming."
  • "What's loaded in this set, and which of my plug-ins would make a better acid line?"
  • "Duck the bass with the kick."
  • "Put a cathedral reverb on return A and send the pads to it."
  • "The low end is muddy — find out why and fix it."
  • "Layer the bass MIDI on a second synth one octave down."

Tool reference

Indexes start at 0. Times and lengths are in beats. Notes use MIDI numbers (60 = C4 in scientific pitch, shown as C3 in Live).

Session and transport

Tool Description
get_session_info / get_session_snapshot Tempo, time signature, tracks, transport; full state dump.
set_tempo · start_playback · stop_playback Transport.
get_scenes · create_scene · set_scene_name · duplicate_scene · fire_scene · delete_scene Scenes.
switch_to_arrangement_view · set_arrangement_time · create_locator · delete_locator Arrangement view, playhead, locators (create, rename, delete by name or position).
undo · redo Live's Undo and Redo. Each tool call that changes the set is one step, and tools that make many changes at once (drum patterns, progressions, arrangements, MIDI imports…) are one step too.
start_undo_group · end_undo_group Make a whole request one undo step (closed automatically after two minutes if left open).
get_remote_script_info Version handshake and capabilities.

Tracks, mixer and routing

Tool Description
create_midi_track · create_audio_track · set_track_name · duplicate_track · delete_track Tracks.
get_track_info Track, return track (track_type="return") or main track ("master"): mixer state, clip slots, devices.
set_track_volume · set_track_panning · set_track_send · set_track_mute · set_track_solo · set_track_arm Mixer; reply with Live's display value (-6.0 dB, 25L). track_type reaches return tracks and the main track.
get_track_routing · set_track_routing Input/output type and channel, monitoring (in, auto, off), matched by the names Live shows.
get_track_meters Live's output meter of every track, return and the main track.

Devices and browser

Tool Description
get_browser_tree · get_browser_items_at_path Browse instruments, sounds, drums, audio_effects, midi_effects, plugins, user_library, packs, samples, max_for_live, current_project.
load_instrument_or_effect · load_drum_kit Load by browser URI, on a track, a return track or the main track.
get_device_parameters · set_device_parameter Every parameter with its range, Live's display string and, for list parameters, the names of the values.
device_inventory Every device in the set and what it is for; optionally every installed plug-in by vendor.
device_knowledge Sheet for a native device (Live 12 manual), a bundled Max for Live device or a known plug-in; the native catalogue.
get_device_info Real device name, kind, latency, and device-specific state: sidechain, plug-in presets, Simpler, rack.
set_device_sidechain Route a Compressor's sidechain from another track and switch it on (ducking).
plugin_presets List or select a plug-in's presets (as exposed to Live).
simpler_sample Load a sample into Simpler; classic / one-shot / slicing by transient, beat, region; warp, crop, reverse.
rack_variation · set_drum_choke Rack macro variations (list, recall, store, randomize); Drum Rack choke groups.

Clips and notes

Tool Description
create_clip · create_audio_clip · set_clip_name · fire_clip · stop_clip · delete_clip Session clips.
get_clip_notes · add_notes_to_clip · clear_notes_from_clip · remove_notes Notes, in Session or Arrangement clips (location="arrangement"); remove_notes clears a time/pitch range only.
get_clip_loop · set_clip_loop Loop switch, loop brace, start/end markers.
get_clip_warp · set_clip_warp · edit_warp_markers Warp mode, markers, pitch, gain; original_bpm conforms a loop to the tempo.
get_clip_automation · set_clip_automation Clip envelopes for device and mixer parameters (Session clips).
get_arrangement_clips · duplicate_to_arrangement · set_arrangement_clip_name · delete_arrangement_clip Arrangement clips.
import_midi_file Import a .mid file (track/channel filters, transposition) or describe it.

Listening and analysis

Tool Description
listen_to_arrangement Records the main output between two song positions on a temporary track (in real time), samples every track's meter, then returns the mix report and the spectrogram image. The track and the recorded file are deleted afterwards, so listening never fills the project with audio (keep_track / keep_audio to keep them).
analyze_mix The same mastering report for any WAV/AIFF file.
analyze_audio_file Lightweight level report (no numpy needed).
record_arrangement · get_recording_status · stop_arrangement_recording · get_audio_clip_file Record armed tracks into the Arrangement between two positions; find the recorded file.

The mix report checks: integrated / short-term / momentary loudness and loudness range (validated on EBU reference signals), true peak (4× oversampled), PLR and crest factor, long-term spectrum and its tilt against a −4.5 dB/octave reference (sub, lows, low-mids, mids, high-mids, highs, air), narrow peaks with their note names, stereo correlation and width per band, low-end mono, DC offset. Each finding comes with a severity and a concrete fix using Live's own devices.

Drums and groove

Tool Description
drum_patterns Pattern library (techno four, rolling, hypnotic, minimal, dub, hard, acid, Detroit, house, deep house, tech house, afro house, UK garage, breakbeat, electro, drum & bass, dubstep, trap, EBM) with tempo, swing and production notes; Euclidean rhythms.
get_drum_map A track's Drum Rack pads and the role each plays (kick, clap, hats, perc...).
write_drum_pattern Program a clip from a style and/or grids per role: x hit, X accent, o ghost, 2-4 ratchet, ?/* probability; Euclidean layers, swing, fills, variation, humanize.
get_groove_pool · set_clip_groove · set_groove_params Live's Groove Pool: list grooves, apply one to a clip, adjust amounts and the global Groove amount.

Evolution and arrangement

Tool Description
write_clip_evolution Extend a MIDI loop to N bars and write long curves per lane (device parameter or volume/pan/send): linear, exp build-up, log, S-curve, swell, slow sine/triangle, steps, random walk.
build_arrangement Lay out the Arrangement from Session clips by section (template or custom), roles from track names, per-section clip variants, tiling, locators, dry_run. Envelopes of the placed clips become track automation.
arrangement_overview Map of which track plays in each block of bars.
arrangement_templates Templates (techno_dj, techno_64, minimal, house) and curve shapes.

Music theory and composition

Tool Description
music_scales ~50 scales and modes — church modes, harmonic/melodic minor families, pentatonics, blues, bebop, symmetric, Japanese, gamelan, flamenco — with character, characteristic note, styles; or one scale in a key: spelled notes, diatonic chords with roman numerals and functions, relative modes, Live's name.
music_styles 24 style sheets (baroque, classical, jazz, bossa nova, blues, gospel/neo-soul, funk, pop, rock, house, techno, trance, drum & bass, boom bap, lo-fi, trap, R&B, reggae, reggaeton, afrobeat, flamenco, film score, ambient, synthwave): tempo, metre, scales, harmony, progressions, groove and swing, bass, melody, arrangement, Live devices.
music_chord_ideas Reharmonisation ideas for a chord, with the reason for each.
analyze_clip_music Key and mode, chords with roman numerals (basic or jazz vocabulary), out-of-scale notes, range, rhythm, melodic motion.
transform_clip Chain of operations, in place or into an empty slot: transpose, transpose_diatonic, change_mode, snap_to_scale, invert, retrograde, time_scale, quantize, swing, humanize, accent, articulate, fold_range, harmonize, arpeggiate, strum, voice_chords.
write_progression Roman numerals (secondary dominants, borrowed chords, Neapolitan) or a style's progression, voice-led, played with an accompaniment pattern and a bass line.
get_song_scale · set_song_scale Live 12's song scale and Scale Mode.

Good to know

  • Live's API has limits. A script cannot save the set, export audio, create group tracks, or write Arrangement automation directly (tempo changes included). Track automation still reaches the Arrangement through Session clip envelopes, which build_arrangement places for you. Live's own analysers (Spectrum, EQ Eight's display, Max for Live meters) don't expose their readings — which is why listening works by recording the main output.
  • Listening takes real time: analysing a 3-minute song means playing it for 3 minutes.
  • Device loading, playhead moves and locators reply only once Live has applied them. Loading the same device that is already on a track resets it in place (the reply says so).
  • Undo works step by step. Live on its own merges everything a script changes into one undo step, so one Cmd+Z would wipe out a whole session of edits; Ableton Studio MCP gives each change its own step, and ask the assistant to undo to take back exactly its last action. Destructive tools (delete_track, delete_scene, delete_clip, delete_arrangement_clip) can be undone the same way, but saving your set before big experiments is still wise.

FAQ

Does Ableton Studio MCP collect any data? No. There is no telemetry, no analytics and no data collection of any kind, and the server makes no internet connection: everything stays between your MCP client and Live on your computer. See Privacy.

How do I control Ableton Live with Claude? Install this MCP server in Claude Code, Claude Desktop or Cursor, install the Remote Script into Live, and select AbletonStudioMCP as a control surface (see Quickstart). Then ask in plain language: "program 8 bars of rolling techno on the 909 kit".

Can an AI mix and master in Ableton Live? It can measure and act: listen_to_arrangement records the main output and reports integrated loudness (LUFS, EBU R128), true peak, loudness range, tonal balance, resonances and stereo width against a target (streaming −14 LUFS, club, classical, broadcast, podcast), with a spectrogram image the assistant looks at; the mixer, EQ Eight, Limiter, Compressor sidechain and other devices are then adjusted through the API.

Does it work with Ableton Live 12, Live 11, Push, Max for Live? Live 11 and 12 (Live 12 for the song scale and some device features). Max for Live devices and VST/AU plug-ins load like built-in devices; Push and other control surfaces keep working alongside it.

Which MCP clients are supported? Any MCP client: Claude Code, Claude Desktop, Cursor, and others that speak the Model Context Protocol.

Can it make techno and electronic music? Yes — drum programming from 19 electronic styles or step grids (Euclidean rhythms, ratchets, per-note probability), long filter builds over 64 bars, DJ-friendly arrangements with intro and outro, sidechain ducking, and a club loudness target.

What's next? The roadmap lists the planned work: sound design recipes, A/B comparison with reference tracks, finding samples by sound, audio-to-MIDI transcription, and more. Ideas and contributions are welcome in the issues.

What can't it do? Live's API doesn't let a script save the set, export audio, create group tracks, write Arrangement automation directly or use Live 12's MIDI Tools; see Good to know.


Troubleshooting

Start with ableton-studio-mcp doctor (see Check): it diagnoses most of the cases below.

Problem Fix
up_to_date: false, or "missing capability" errors The script in Live is older than the server. Re-run the install command, then restart Live.
The server still runs old code after an update Restart the MCP client; if needed uv cache clean ableton-studio-mcp with the client closed.
No connection Check that AbletonStudioMCP is selected as a Control Surface and that only one MCP client runs the server.
Timeouts on big requests Ask for the work in smaller steps.

How it works

MCP client (Claude…)  ⇄  MCP server (ableton_studio_mcp/, Python)  ⇄  TCP 127.0.0.1:9877  ⇄  AbletonStudioMCP control surface inside Live (ableton_studio_mcp/live_script/)
  • The control surface runs inside Live and executes commands through Live's Python API. It listens on localhost only (set ABLETON_MCP_HOST deliberately to change that).

  • The MCP server exposes the tools and does the heavy work outside Live: audio analysis, MIDI files, music theory. Its code is organised as:

    Module Role
    live.py The connection to Live: one socket, reopened when Live restarts, one request at a time, per-command timeouts, and the control surface's version and command list checked on every connection.
    app.py The MCP application and the @tool decorator: checks the commands a tool needs and turns any failure into one readable line.
    tools/ The tools, one module per domain: session, tracks, clips, devices, browser, routing, listening, midi, theory, drums, evolution.
    audio_analysis.py, mix_analysis.py, music_theory.py, drums.py, arrangement.py… The domain logic, independent of Live and fully unit-tested.
    install.py, doctor.py, server.py Installer, diagnostics, command-line entry point.
    live_script/AbletonStudioMCP/ The control surface that runs inside Live: transport.py (socket, newline-delimited JSON), registry.py (the command table: each handler's signature is its contract, and the advertised capabilities are generated from it), surface.py (everything runs on Live's main thread from update_display: no threads, non-blocking sockets; reads of a sequence answered in the same pass in under 1 ms, each change given its own display update), and one module per domain (session, tracks, clips, notes, warp, automation, devices, groove, browser, recording, snapshot). It runs in Live's own Python (3.7 in Live 11), so it avoids newer syntax.
  • Messages are one JSON object per line: {"type": ..., "params": ...} → {"status": ..., "result" | "message": ...} (protocol 2; the server still understands 1.x control surfaces, which send no newline).

Development

git clone -b feat/mixer-scenes https://github.com/cedrati/ableton-studio-mcp
cd ableton-studio-mcp
uv run --with pytest pytest -q             # 321 tests, no Live needed
uv run ableton-studio-mcp install          # install your working copy into Live
uv run ableton-studio-mcp doctor           # check the whole chain

The Live control surface lives in ableton_studio_mcp/live_script/AbletonStudioMCP/ and is shipped inside the package. Versions: __version__ (ableton_studio_mcp/__init__.py), version in pyproject.toml and in server.json are the release number; the control surface's SCRIPT_VERSION only has to be in the same series (major.minor). A release that changes the control surface starts a new series (2.1.0), so users know to reinstall it; a patch release of the server alone (2.0.1) doesn't touch Live. The tests and the release workflow check all of this. Add an entry to the Changelog. GitHub Actions runs the tests on Python 3.10, 3.12 and 3.13 on every push and pull request.

Releasing: push a tag vX.Y.Z matching that version. The release workflow checks the versions, runs the tests, builds the package, publishes it to PyPI (trusted publishing, no token stored), creates the GitHub release and publishes server.json to the official MCP Registry as io.github.cedrati/ableton-studio-mcp (GitHub OIDC; the mcp-name: comment at the top of this README proves the PyPI package is ours).


Privacy

No data is collected. None. Ableton Studio MCP has no telemetry, no analytics, no usage statistics, no crash reporting and no account.

  • Nothing leaves your computer. The server only talks to Ableton Live on your own machine (127.0.0.1) and makes no internet connection.
  • Your music, your Live sets and your prompts stay with you. Tools do not receive your prompts, and nothing about your sessions is recorded or sent.
  • No tracking in Live. The Remote Script only answers the commands it is sent; it does not watch or log what you do.
  • The telemetry and research-dataset recording inherited from the original ableton-mcp were removed from the code entirely in 1.17.0, along with their online-database dependency (Supabase). Nothing is merely switched off: the code is gone.

The only files the tools write are on your machine (recordings you ask for, temporary spectrogram images, the Remote Script itself): the full list is in TERMS.md.


Credits and license

  • Author: Olivier Cedrati (@cedrati). Report issues on cedrati/ableton-studio-mcp.
  • Thanks: Ableton Studio MCP began as a fork of ableton-mcp by Siddharth Ahuja, which first connected Live to AI assistants through MCP. Since 2.0 it is an independent codebase, rewritten from the ground up (versions 1.x remain available under their original license).

MIT License — Copyright (c) 2026 Olivier Cedrati. See LICENSE. This is a third-party integration, not made or endorsed by Ableton.

Metadata

Release files for ableton-studio-mcp 2.1.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for ableton-studio-mcp 2.1.0
File Size Uploaded Provenance Actions
ableton_studio_mcp-2.1.0.tar.gz 209.5 kB With provenance Details for ableton_studio_mcp-2.1.0.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for ableton-studio-mcp 2.1.0
File Interpreter ABI Platform Actions
ableton_studio_mcp-2.1.0-py3-none-any.whl Python 3 none any Details for ableton_studio_mcp-2.1.0-py3-none-any.whl

Total release size: 379.4 kB

Release files / ableton_studio_mcp-2.1.0.tar.gz

Download URL ableton_studio_mcp-2.1.0.tar.gz
Size 209.5 kB
Tags Source
SHA-256 checksum
How to use checksums
8b21b69ae4dc352c3b84098a9f112cb20480c6855e23a7680486c4c2582dbb52
BLAKE2b-256 checksum
How to use checksums
8e778e14a9ce4f8def3ee96db75ed1ab29ece6415d0165884a5d128dbc975432
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 21, 2026.

Transparency log

Release files / ableton_studio_mcp-2.1.0-py3-none-any.whl

Download URL ableton_studio_mcp-2.1.0-py3-none-any.whl
Size 169.9 kB
Tags Python 3
SHA-256 checksum
How to use checksums
28d868229027970f3e088b38ff352ae6511c08879d9c4ae9eac9d48744e70745
BLAKE2b-256 checksum
How to use checksums
25f2770d5e7be6afeb623ccdbe970edcdaf58efc44d3906c4ec9681b97c36a41
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 21, 2026.

Transparency log

Security for ableton-studio-mcp 2.1.0

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

Full provenance

Both files in this release have been uploaded with provenance metadata.

Select from 2 files

Attestation for ableton_studio_mcp-2.1.0.tar.gz

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 21, 2026.

Transparency log

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page