Saturday. One repo, built with AI.
Built it with AI? Start with a check-up.
You made an app with an AI builder. It works, mostly. You're not sure what's inside it, or what you'd even ask an agent to do.
One big unknown, and no idea where to start
So the app stays as it is, and every change feels like it might break something you can't see.
Wire it up once.
One behaviour at a time, each in the real dashboard.
-
No agent yet? Setup picks one.
One that comes with the AI you already pay for, or a free one. If an agent is already signed in on this computer, Overhand uses that.
How it works: Overhand drives the agent through its own command line; the login stays with the agent.
- Claude Codefoundsigned in on this computer: used as it is
- Codexreadycomes with the AI you may already pay for
- OpenCode$0free, on a local model
-
A read-only check-up.
Overhand reads your project, changes nothing, and lists what's worth fixing, in plain words. Hand any finding to an agent with one click.
How it works: the check-up only reads. Nothing changes until you hand a finding over, and that arrives as a pull request.
-
Ask in your own words.
Say what you want, like "make the invite page friendlier". The agent plans first, in plain words, and nothing is built until you say go ahead.
How it works: type it in Ask or do anything (⌘K); the plan arrives in Needs you.
What you have at the end of setup
Your first win before lunch.
A list of what's worth fixing in your app, in plain words, and one of them already done as a pull request for you to read.
- 0files the check-up changes
- 1pull request: your first win
- 1click from any finding to an agent