Make music

Make your own music

Describe what you want to hear and let an assistant write the score. Soundroo renders it into music. Choose where you want to work.

With a coding agent

The way to make a whole collection. Your agent uses the Soundroo toolkit to compose ten or more tracks, render them and open a listening studio on your computer. Revise together until it sounds right, then ask it to publish.

The Soundroo toolkit is in private beta. This prompt works for people who have access to its repository; public cloning isn’t open yet.

Describe it
A mood, a place, instruments, tempo, or how you’ll listen.
Copy the prompt
It points your agent at the toolkit and its instructions.
Listen, then publish
Your agent sends a studio link. Publishing is a separate step you ask for.

Prompt for your coding agent

Make a Soundroo collection of 10 original tracks.

Brief:
(Describe the music here.)

Use the Soundroo toolkit at https://github.com/vincentsch/soundroo. Read its AGENTS.md and follow it. Compose original motifs and develop them; don't recreate existing melodies. Keep every score editable, render and check each track, then open the local studio and send me its URL.
When I say "publish", share the collection on soundroo.com and give me the links.

For Claude Code, Codex and other agents that can run commands on your computer. The toolkit renders and plays everything there, and nothing is published until you ask.

In ChatGPT or Claude

Add Soundroo as a connector, then ask for music in a chat. The assistant writes a compact score and Soundroo sends back a private preview link, usually within half a minute. Expect shorter, simpler pieces, one to three per conversation. The assistant can’t hear the result, so your notes guide each revision.

Connector URL

Claude

In Claude, open Settings → Connectors, add a custom connector and paste the URL above.
Connect it and sign in to Soundroo when asked. That creates your account.
In a chat, ask for a track. Claude replies with a preview link.

Claude Help Center

ChatGPT

Enable developer mode if your plan supports it. Open Plugins → Add custom MCP server and paste the URL above.
Sign in to Soundroo when asked. That creates your account.
In a chat, ask for a track. ChatGPT replies with a preview link.

ChatGPT Help Center

OpenAI and Anthropic decide which plans can add connectors, and workspace admins can turn them off. Their menus change, so their help centers have the current steps. You can disconnect Soundroo any time in Settings.

Try asking for

Use Soundroo to make a one-minute original track: a warm, slow funk groove with a vocoder hook
Render a private preview and send me the link so I can listen before we publish.

From your own code

Create an API key in Settings, send a score to the previews endpoint, listen to the private preview and publish it when it’s right. Scores render with Soundroo engine 0.4.0.

curl https://soundroo.com/api/v1/previews \
  -H "Authorization: Bearer $SOUNDROO_TOKEN" \
  -H "Content-Type: application/json" \
  -d @preview.json

API reference Score format