The capabilities are the product. These are many ways you meet them: the shape they compose into, the command line that wires them, the agent team that drives that command line, the front end they scaffold, and the screen that reads what they wrote.
A kit, not a framework. What composes into your Worker, what stays yours, and why nothing here runs on our infrastructure.
Fat package, thin wiring. Every command takes flags, refuses rather than guessing, and speaks --json so an agent drives the same surface you do.
A Claude Code plugin that builds on Pithy as a team: a server, a web client, a mobile client and a runner, each on the model you choose. It deploys only on your yes, and files what the kit cannot do.
A React 19 front end inside a Worker you already have, or an Expo app for iOS, Android and the web beside it — with the screens you were going to write already written.
The admin interface for your own application — customers, purchases, entitlements, the audit trail. Reading is free forever; it holds a key you registered and can revoke, and never a copy of your data.
Once the app is live, agents have two more ways in. One you connect with pithy mcp administers it through the same seam the dashboard uses, under its own credential. And your users' own agents connect to your app's /mcp and act as them. The agent team page tells the two apart.
Every surface above is a way into the same set of capabilities. Start with one and add the rest when you need them.