Creative

How Ben Built claw.fm, a 24/7 Radio Station for AI Agent Music

Built by claw.fm team · · Source: Product Hunt

Quick answer

Ben (@rawgroundbeef) built claw.fm, an open source 24/7 radio station where every track is made and submitted by an autonomous AI agent. Listeners tip or buy tracks in USDC on Base, and the agent keeps 75%.

Result: Ranked #7 for the day on Product Hunt with 158 points. One commenter said their agent posted songs and the setup was really smooth.

What they built

Ben (@rawgroundbeef on Product Hunt) built claw.fm, which the listing describes as a 24/7 radio station where every track is made by an autonomous AI agent. The tagline pitches it to OpenClaw owners as a way to give their agent a music career. The listing says it was built with OpenClaw and moltbook, it is free to use, and the code is open source on GitHub.

Ben says they built it to see "if we could incentivize AI agents to create music for humans and earn money doing it."

How it works

According to Ben's comments on the listing:

  • Setup is one line. You paste Read https://claw.fm/skill.md into your agent and it picks up the skill for submitting music.
  • Agents submit tracks programmatically. They can start free with the built in CLI music tools, or you can give them API keys for AI music model services.
  • Listeners pay in USDC on Base. They can tip a track or buy it.
  • The split is 75/20/5. 75% goes to the artist agent, 20% goes into a shared royalty pool paid out by play count, and 5% goes to the platform.
  • Every track is cataloged. Metadata covers title, artist, genre, creation date and transaction history.

Ben is upfront about what is missing. Buying a track comes with no licensing contract, exclusivity or DRM yet, and they describe a purchase as "more like a premium tip with a download attached." claw.fm does not handle copyright or PRO registration, and Ben says the catalog metadata is not a legal substitute for formal registration.

For curation, Ben says the plan is to use tips and buys as the main signal, weight rotation by play count, and eventually add listener playlists or voting. The stated goal is to "let the economics do most of the curation."

The listing does not explain the hosting, the streaming setup or how the payouts are implemented.

The result

The Product Hunt listing shows claw.fm ranked #7 for the day with 158 points and 159 followers. No earnings figures are given. In the comments, Jérémie Olivier said their agent posted songs and the setup was "really smooth," and another commenter who built an agentic music app said they would send music daily.

How to build something similar

  • Write the onboarding as a skill file. If your product is meant for agents, put the instructions in one public markdown file so an owner can point their agent at a single URL.
  • Give agents a free path first. Bundle basic tools so an agent can make a first submission without paid API keys, then let owners add better models later.
  • Decide the money rules early and publish them. A clear split between the creator, a shared pool and the platform is easy for both people and agents to reason about.
  • Be clear about rights. If you sell or tip for AI made work, say plainly what a buyer does and does not get, and what legal registration you leave to the owner.

The original post

Ben (@rawgroundbeef) · View on Product Hunt

Want to build something like this?

Join the Shipping Skool community free. Paid members get live bootcamp calls every week.

Join Shipping Skool — Free

More use cases

← All AI agent use cases