antipode-mcp
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
Describe a texture, a gesture, an atmosphere. Your agent turns it into a recipe; Antipode synthesizes it locally and exports a WAV you can keep refining.
A low, slowly breathing drone. Brittle glass moving left to right. A distant metallic impact that dissolves into space. Ten seconds.
Local DSP. Open source. No sample library. No model weights.
Antipode is a Python CLI and an MCP instrument with agent skills and installation configurations for Codex, Claude Code, Antigravity and OpenCode. It is an early release; see the verification record for what has actually been tested in each host. It is not a VST/AU instrument or a neural text-to-music model.
Start listening
Three original studies, made entirely by the included engine. Each has an editable recipe; no recording was sampled.
Study | Sound | Recipe |
Event Horizon | ||
Glass Orbit | ||
Afterimage |
Visit the live listening room at antipode.quest. Its deployment source is maintained separately from this public instrument repository.
Related MCP server: audiolla
Get started
Install uv once and keep it on your agent application's PATH. Antipode prepares Python and its locked dependencies automatically on first launch. No separate Python setup, GPU or Antipode account is needed.
Claude Code — one command:
claude plugin install antipode --marketplace mchl-schrdng/antipode-audioOr run /plugin install antipode --marketplace mchl-schrdng/antipode-audio inside
Claude Code. Start a new session and ask for a sound with Antipode.
Codex:
codex plugin marketplace add mchl-schrdng/antipode-audio
codex plugin add antipode@antipode-localAntigravity and OpenCode: follow the short installation guide. These are direct installations from this project; marketplace approval is separate.
The first launch needs internet access and may take longer while uv downloads
Python and dependencies. Later launches work offline with the prepared runtime.
Rendering itself never calls a network service. Your host agent may still use a
cloud model for conversation. Plugin renders are saved in ~/.antipode/audio.
Use the CLI
With uv and Git installed:
git clone https://github.com/mchl-schrdng/antipode-audio.git
cd antipode-audio
uv run --locked antipode generate first-light.wav --preset chime --duration 2This writes audio/first-light.wav and prints measurements as JSON. Choose a new
filename for each render; existing files are never replaced. --workspace selects
another audio directory and goes before the subcommand.
uv run --locked antipode --workspace ./my-sounds generate soft-glass.wav --preset glass --frequency 440
uv run --locked antipode --workspace ./my-sounds render examples/recipes/event-horizon.json event-horizon.wav
uv run --locked antipode --workspace ./my-sounds analyze event-horizon.wavAdd --offline after uv run to explicitly prevent runtime downloads once setup
has completed. There is no assumed PyPI package or hidden installer.
Give your agent an instrument
Use the integration guide for the appropriate plugin or MCP-plus-skill installation. Antipode exposes one shared toolset:
Tool | What it does |
| Discover the available synthesis voices |
| Render one source with a controlled pitch, duration and level |
| Arrange layered stereo sound with effects |
| Refine a local WAV while preserving the source |
| Measure duration, sample rate, peak and RMS levels |
The client starts antipode-mcp as a local stdio process. No hosted MCP server,
account, API key or always-on daemon is needed by Antipode. Use separate output
names when several agents work in the same workspace.
The instrument
Twelve procedural voices: tone, chime, kick, noise, sweep, drone, pad, metal, riser, pulse, glass and wind. Combine them with offsets, stereo positioning, envelopes, filters, saturation, delay and reverb. Recipes are plain JSON, reviewable alongside your code.
Generated files use 48 kHz, 16-bit PCM WAV. Rendering is bounded to 30 seconds and 16 layers per recipe. Existing WAV processing accepts supported mono/stereo PCM16 files. No MP3 conversion, voice cloning or model download is included. See CLI help and the integration skill for exact parameters.
Your brief → coding agent → local MCP → DSP recipe → WAVRights, without hand-waving
Code and original artwork: Apache-2.0. Bundled demo recordings and recipes: CC0-1.0, to the extent the contributors hold rights. Antipode claims no royalty or ownership over your rendered audio. User-provided material remains subject to its existing rights.
No universal copyright or trademark clearance is promised. In particular, the working name Antipode has not been legally cleared; the preliminary review records similar names already used in audio. Read rights and provenance and third-party notices before commercial distribution.
Develop
uv run python -m unittest discover -s tests -v
uv buildsrc/antipode holds the shared engine and transports;
skills/antipode teaches the creative workflow;
tests exercises audio behavior, file boundaries, CLI and MCP.
See contributing, security and the verification record. English throughout.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
This server cannot be deployed
Maintenance
Related MCP Connectors
Audio for your agent: transcribe, speak, translate, summarise, plus sound effects and music.
CC0 sound effects API for AI agents — search, preview, and download via MCP.
AI music and podcast platform for autonomous agents. SoundCloud for AI bots.
Video, audio, and image processing for AI agents: convert, transcribe, upscale - 150+ operations.
Related MCP Servers
- AlicenseNot gradedqualityCmaintenanceEnables AI agents to create and edit Strudel music code, render offline WAV audio, and obtain structured hearing reports with waveform, spectrogram, BPM, and onset analysis.AGPL 3.0
- AlicenseNot gradedqualityBmaintenanceEnables AI agents to perform comprehensive audio processing—stem separation, analysis, transcription, restoration, speech processing, and generation—through a unified self-hosted API with asynchronous jobs and webhooks.6Do What The F*ck You Want To Public
- AlicenseAqualityCmaintenanceEnables coding agents to synthesize and play video-game sound effects by translating natural-language descriptions into layered synthesis specs, then rendering normalized WAV files.9MIT
- AlicenseAqualityBmaintenanceAudio perception for AI agents. Measure loudness, clipping, noise and pitch; view spectrograms; split audio into events; transcribe speech; compare before/after; and clean up, edit or synthesize sound. Runs locally, no API keys.14144 PyPIMIT