Words into Music MCP
Connect an AI assistant to Words into Music to generate songs from prompts or lyrics, check credits, and access your private song library.
This repository contains public connection metadata and documentation only. The hosted application implementation is private and is not included.
Connect
Use a client that supports remote Streamable HTTP MCP servers with OAuth.
https://mcp.wordsintomusic.com/mcp
For clients that read mcp.json, use the configuration in this repository (Agent Plugins format; most clients only need the url). Complete OAuth sign-in in your browser. Never paste your password, session cookie, or a provider API key into a chat. No local package installation is needed.
Setup guide · ChatGPT · Claude · Grok
ChatGPT and Claude have completed real production song-generation tests. Other clients need their own acceptance test. The Grok guide describes its documented route; a Grok end-to-end test is pending.
Gemini CLI
gemini extensions install https://github.com/constantinginga/words-into-music-mcp
In Gemini CLI, run /mcp auth words-into-music and complete OAuth sign-in. Then check /mcp for the available tools. The extension manifest and installation have been validated with Gemini CLI 0.40.1. Authenticated Gemini song generation has not completed an acceptance test yet.
Agent skill
skills/words-into-music teaches an agent the safe flow: check credits, prepare the brief, confirm the one-credit cost, start, poll every 15 seconds and return the private song link. Install it with:
npx skills add constantinginga/words-into-music-mcp
The repository is also a portable agent plugin (plugin.json, mcp.json, skills/). In Hermes Agent, install it with hermes plugins install words-into-music and complete the OAuth sign-in it starts.
Tools
| Tool | Purpose | Song credits |
|---|---|---|
get_account_usage | Check available and reserved credits | 0 |
prepare_song | Validate and save a song brief | 0 |
start_song | Start a prepared recording | 1 |
get_song_job | Check progress and get the finished link | 0 |
list_songs | List your private songs | 0 |
get_song | Read song details and optionally lyrics | 0 |
Prepare a song first and confirm its cost before starting generation. A free account has one song in total across the website and connected clients. Subscriptions and song packs share the same wallet and limits. See pricing.
Supported song languages are English, Spanish, German, French, and Brazilian Portuguese. Request vocals or an instrumental, with a target duration of 60 to 180 seconds. Accepted jobs continue after you close the chat. Find finished recordings in your private library.
OAuth
The service supports OAuth discovery, dynamic client registration, PKCE, and refresh tokens. Scopes are account:read, songs:read, and songs:create. Review permissions before granting access. Disconnect clients in Words into Music Settings to revoke future tool calls.
The service does not issue manual personal access tokens. Unauthenticated requests to the MCP endpoint receive HTTP 401. This is expected.
Support
Contact and security reports: contact@wordsintomusic.com.
The MIT license covers only this repository's public metadata and documentation. It excludes the hosted service implementation, trademarks, and logo.