MiniMax Launches Open Source Music Skills for Agents to Compose and Sing
· Updated
MiniMax, an AI company building multimodal models, open-sourced three Music Skills for agentic workflows. These skills—minimax-music-gen, buddy-sings, and minimax-music-playlist—run natively in MMX-CLI alongside the Music 2.6 model. They allow agents to autonomously handle tasks ranging from one-line song generation to personalized playlist curation.
This release shifts music generation from a manual task to an Agent Skill (a reusable capability for autonomous systems). An agent can now decide to compose a song or sing as a persona to achieve a goal. The integration with Claude Code specifically enables coding agents to adopt vocal personalities.
Integrate these capabilities into any agent on the MiniMax platform by connecting a Token Plan and adding the skills to your environment. The buddy-sings skill is designed for Claude Code users who want agents to interact through first-person vocals. All code is available on GitHub.
What if your agent could write you a song, sing as your AI companion, and curate a playlist from your library natively?
🎶Today we're open-sourcing three Music Skills:
minimax-music-gen: one-line prompt to full track. Agent auto-picks original, instrumental, or cover.
buddy-sings: your Claude Code character reads its own persona and sings in first person.
minimax-music-playlist: agent scans your local music, learns your taste, builds your playlist.
All three are tuned to run natively in MMX-CLI alongside Music 2.6 and the rest of our multimodal models. Connect a MiniMax Token Plan, add MMX-CLI, and your agent picks up the Skills on its own.
Start here👉https://t.co/UEmNvPwChs
