Skip to main content
Vibe coding is the fastest way to ship OpenHome. Let Claude (or any coding agent) drive the build while you describe what you want in plain English. With the right briefing, you can go from idea to live Ability in an afternoon. There are two main surfaces:

Build Abilities with Claude

Scaffold, deploy, and test Abilities from your terminal using the OpenHome CLI. Claude writes the Python; you review and ship.

Build dashboards with Replit

Hand the Replit Agent a master prompt; get a working Flask + dashboard in one shot.

​
The rule of two tabs

Whichever surface you’re on, keep two browser tabs open side by side so you see both sides of the pipe at once:
Building an AbilityBuilding a dashboard
OpenHome Live Editor (Ability logs)OpenHome Live Editor (Ability logs)
Your terminal with Claude + the CLIReplit (server logs)
If a request fails, you see it in both within a second — the fastest feedback loop voice AI has ever had.

​
What makes OpenHome good for vibe coding

  • Clear sandbox rules — a tight list of allowed/blocked APIs means the LLM has unambiguous constraints to respect
  • Consistent SDK — every method on self.capability_worker or self.worker; easy to brief a model on
  • Deep source material — SDK Reference + Simple Abilities Cookbook fit in a single LLM context window
  • CLI + API — the OpenHome CLI gives agents a tool surface to drive the entire build → deploy → test loop

​
Next steps