I built an AI Music MCP that lets agents generate fully produced songs directly from text instructions. The goal was simple: add “music” as a missing modality to the MCP ecosystem.
The MCP currently supports three main generation modes:
Idea → Track: describe a mood or scene and get a finished 4+ minute song
Lyrics → Song: provide your own lyrics + style, and the tool returns a fully arranged track
Instrumental Mode: generate jazz, orchestral, lofi, cinematic, etc. without vocals
It also works across genres (classical, electronic, folk, rock, etc.) and supports prompts in many languages.
Each generation call returns full metadata, including:
MP3 download URL
title
lyrics (if applicable)
style/genre tags
cover art
duration
created timestamp
The idea is to let agents compose complete, usable music without needing any training, audio pipelines, or GPU setup. Just MCP calls.
If you're experimenting with creative agents or building AI products that could benefit from native music generation, here’s the project: https://www.musicmcp.ai/
Happy to hear feedback, ideas, or edge cases you'd like supported.