How ProofKit Works
Agent-first architecture — from schema to deployed app in a single session.
What gets installed
ProofKit is four pieces that work together to connect your AI agent to your FileMaker file:
The Add-on
Installed into your FileMaker file. Provides the Web Viewer and the bridge between your app and FileMaker scripts/data.
The Plug-in
Runs on the machine. Handles communication between the MCP server and the add-on inside FileMaker.
The MCP Server
Exposes your FileMaker schema, scripts, and data to AI agents through the Model Context Protocol.
Agent Integrations
Skills, rules, and context that teach your AI agent how to build ProofKit apps effectively.
How the pieces connect
A chain from your editor to your FileMaker file:
The feedback loop
This is what makes ProofKit different from "just using ChatGPT to write code." The agent doesn't just write — it deploys, verifies in a real browser, sees what it built, and can inspect, iterate, and fix common issues in the same loop.
TypeGen and type safety
The agent can work from the shape of your data at compile time. TypeGen reads your FileMaker schema and generates TypeScript types — field names, value lists, relationships. Many field-name mistakes are caught before you run the app.
The opinionated stack
ProofKit ships with a curated set of tools that AI agents work best with:
- React + TypeScript — the language models know best
- shadcn/ui — beautiful, accessible component library
- TanStack Query + Router — caching, routing, and data fetching
- Vite — lightning-fast builds and hot reload
Already using Svelte, Vue, or another framework? ProofKit can accommodate your existing stack too.