Bunkhouse overview

Bunkhouse adds an AI assistant to a web application you already run. One script tag loads a side pane, styled to fit your page, that discovers what the page can do and helps the signed-in user get work done.

What gets embedded

The loader script adds an Assistant launcher to your page. When a user opens it, the assistant can work with:

• Your existing UI: buttons, links, form fields and tables on the current page.
• Requests your app makes: API calls observed after the loader runs.
• Hints you publish: a hints file that describes operations discovery cannot see.
• WebMCP tools: tools your page declares for agents.

The model writes small programs that run in a sandbox on the embed origin, not in your page. Those programs can only call capabilities the assistant has discovered and that your configuration allows, and write actions are shown to the user for review before they run.

Before you start

You need:

• A web application you control, served from origins you can list exactly (for example https://app.example.com).
• The ability to add a script tag to its HTML, before your application’s JavaScript entry.
• For signed-in users, the ability to add a small endpoint to your application server. The installation guide explains the options.

Where to go next

1. Quickstart: create an account, register your application and place the loader.
2. Installation: choose a session mode and connect your existing authentication.
3. Skills and discovery: control what the assistant can find and do.
4. Security: keys, origins, private content and approvals.