Every app you build now works with AI agents
Bool now builds an agent interface alongside your screens. As it writes each feature, it also writes the typed actions an AI agent can call, using the WebMCP standard: add a stop to the itinerary, move a task, find a record. Open your published app in an agent-capable browser (ChatGPT's desktop browser, or Chrome with WebMCP enabled) and you can ask the agent to work in your app while you watch the same page change. Nothing to set up and no code to write. Every new app gets this from its first build.
See every action in the Site tools section
A new Site tools section in your project dashboard lists every action your app exposes to an agent, grouped by the page it lives on, with what each one does and what it takes. It fills in as Bool builds, so you can watch the interface take shape.
Try them yourself from the preview
Turn on the Site tools button from that section and it appears on the preview toolbar, showing the tools available on the page you're looking at. Run one by hand with the right control for each field, a date picker for a date, a menu for a choice, and watch your app respond.
Shape them in chat, like anything else
The actions are part of your app, so you change them the way you change everything: ask. "Let an agent assign cards" or "add a tool that finds recipes by ingredient" and Bool adds it alongside the feature.