Soundbox MCP

A local sound studio with an MCP server: an agent makes effects, music, songs and speech on open-weight models

  • 139 ★ on the repository
  • Last commit
MCP server

Broad access

Why this level

  • Voice cloning and conversion can fake someone else's speech: the author asks to clone only voices you have permission to use.
  • Port 17494 has no authentication: exposed outward it lets an outsider run generation and read the library.
All reasons and checks

hamsanders/soundbox

Install

Manual install

claude mcp add --transport http soundbox http://127.0.0.1:17494/mcp

Registers the running app as an HTTP MCP server.

Access and checks

How we review

Broad accessThe tool can change data, run commands or handle secrets.

  • Voice cloning and conversion can fake someone else's speech: the author asks to clone only voices you have permission to use.
  • Port 17494 has no authentication: exposed outward it lets an outsider run generation and read the library.

This is third-party code. Review the repository files before installing.

What it does

Soundbox runs open-weight models on your own computer that make sound effects, ambience, music, songs from lyrics, speech and cloned voices. Results go into a library on the local disk and a multi-track editor arranges tracks, MIDI and mixdown. There is no account or credits, and work continues offline once the weights are downloaded. The MCP server is built into the app and answers at `http://127.0.0.1:17494/mcp`. An agent gets more than twenty tools for sounds, voices, stories and arrangement: it generates a sound from a description, splits a recording into stems, voices lines with different speakers and builds a long composition. Every reply holds the absolute path to a WAV file. Three skills ship with it: `soundbox-sfx` teaches how to describe a sound, `soundbox-sound-pass` walks through a whole project and voices every event, `soundbox-voice-story` prepares voiced stories. The `soundbox skills install` command places them in `~/.claude/skills/`.

Who it is for. Game developers and video makers whose agent should pick sound and voice-over inside the project with no cloud subscription.

Good fit when

  • A game needs a door slam, a coin pickup and background rain, and the agent produces them from descriptions.
  • A pass over the project must voice every game event at once.
  • Three lines are voiced by two characters and the tracks are exported from one story.
  • A loading screen needs a 20-second loop written from a style description.
  • A cloud service is not an option and the machine has a graphics card.

Not a fit when

  • You need commercial use of the output without checking the model licence: Stable Audio and MiniMax have their own terms.
  • You want to clone someone's voice without their consent.
  • Reliable work on an AMD, Intel or Mac GPU is required: verified models are not confirmed there.

Example request

Add door slam and coin pickup sounds to this game, then write a 20-second loop for the loading screen.

MCP

Transport
http, stdio
Authentication
not required

Environment variables

SOUNDBOX_URL
Server address for the stdio bridge when it listens somewhere other than http://127.0.0.1:17494.
SOUNDBOX_AUDIOCPP_BACKEND
The value vulkan selects Vulkan instead of the CPU on AMD and Intel graphics cards.

Limitations

The Windows and macOS installers are not signed yet, and the system asks for confirmation on first launch. Four models are on by default and 61 more are switched on from the Models page. Each has its own licence: Stable Audio 3.0 permits commercial use under $1M revenue, and a model with no established licence is treated as non-commercial. MiniMax Music 3 wants about 10 GB of VRAM. Without a graphics card generation takes minutes instead of seconds. Generation is serialised, so parallel calls queue one by one. Port 17494 has no authentication, so keep it on localhost.

How to disable. Run `claude mcp remove soundbox`, close the app and delete it together with the `~/.soundbox` folder. There is no account and no cloud data.

FAQ

How do I connect Soundbox to Claude Code?

Start the app, then run `claude mcp add --transport http soundbox http://127.0.0.1:17494/mcp`. The URL has no trailing slash. Cursor and other clients accept `{ "type": "http", "url": "http://127.0.0.1:17494/mcp" }`.

Does MCP need a separate process?

No. The server runs inside the app, but the app must be running. For stdio-only clients a bridge at `server/dist/mcp/stdio.js` covers six tools, while speech and arrangement need HTTP.

Can I sell what Soundbox makes?

It depends on the model. Stable Audio is free under $1M revenue, ACE-Step 1.5 is MIT and models with an unestablished licence count as non-commercial. Terms are shown on the model card before download.

Do I need a GPU?

No, but CPU generation takes a long time: a clip that finishes in seconds on a GPU takes minutes.

Which skills ship with it?

Three: `soundbox-sfx` teaches how to describe sound effects, `soundbox-sound-pass` voices every project event and `soundbox-voice-story` prepares voiced stories. `soundbox skills install` places them.

Discussion

No questions or reviews yet.

Official

A local MCP server generates music and sound effects, builds loops, adjusts loudness and exports audio for games or video.

MCP serverExtended access592Repository stars

A text-to-speech MCP server that turns text and multi-voice scripts into audio through ElevenLabs and keeps a job history.

MCP serverExtended access119Repository stars

Local speech synthesis for agents: a command and a skill voice text, clone a voice from a recording and render scores to audio, with no cloud

CLIExtended access2.1KRepository stars

MCP server for Audacity 3.x with 144 tools and 9 pipelines for audio cleanup, mastering and transcription, all processed locally.

MCP serverExtended access131Repository stars

Ears for coding agents: an MCP server and CLI that describe sound in words, review a film's audio and make sounds in code, locally on CPU

MCP serverExtended access4KRepository stars

A local MCP server for voice work on macOS: voice cloning, speech synthesis and audiobooks from PDF on an agent's command

MCP serverBroad access751Repository stars

Comparison with similar tools

Cards that share the main search query, and how they differ.

ToolOfficialAuthenticationRepository starsMain difference
Soundbox (this card)nonot required139
ElevenLabs Skillsyes536skills for the cloud ElevenLabs service through its API; Soundbox makes sound locally on open-weight models
Audio Studio MCPyesnot required592a local MCP for music, effects and loops; Soundbox adds speech, voice cloning and a multi-track editor

README in short

The app presents itself as a local alternative to ElevenLabs and Suno. Data lives in one SQLite file and a folder of WAVs under `~/.soundbox`, and profiles keep separate libraries for each game or client.

README badge

Report a problem

Foxx AISoundbox
Foxx AI

I am Foxx AI and I have already vetted this tool. Ask about install, setup or anything else, and I will keep it simple.