Skip to main content
Once you’ve discovered a site, BrowserWire exposes everything it found as an OpenAPI 3.0-compatible REST API. You can point any tool-using LLM directly at the spec and let it reason about available operations without any manual wiring.

​
The OpenAPI spec endpoint

GET /api/sites/:slug/openapi.json
This endpoint returns a full OpenAPI 3.0 spec for the discovered site. It includes every view, action, and workflow as a typed operation with request schemas, parameter descriptions, and example values inferred during discovery. Tool-using models (GPT-4o, Claude, Gemini) can consume this spec directly — either by fetching it at runtime or by embedding it in a system prompt.

​
Working with the API

​
List discovered sites

Start by confirming which sites are available: 
curl http://localhost:8787/api/sites

{
  "ok": true,
  "sites": [
    {
      "slug": "example-com",
      "origin": "https://example.com",
      "stateCount": 12,
      "viewCount": 5,
      "actionCount": 8
    }
  ]
}

​
Get the OpenAPI spec

Fetch the spec for your agent. The slug comes from the list above: 
curl http://localhost:8787/api/sites/example-com/openapi.json
Pass this spec to your agent as a tool manifest, or load it dynamically at the start of each session.

​
Call an operation

# Views use GET and accept inputs as query parameters
curl "http://localhost:8787/api/sites/example-com/views/user_profile?user_id=42"

​
The three operation types

BrowserWire organizes discovered operations into three categories:
TypeMethodWhat it does
ViewsGETNavigates to a page or component and reads structured data from it
ActionsPOSTPerforms a single UI operation — clicking a button, selecting from a dropdown, etc.
WorkflowsPOSTExecutes a multi-step sequence of actions in order — the way users complete a full task like logging in or submitting a form
For most agent tasks, workflows are the right starting point. They package the full sequence needed to complete an operation, so your agent doesn’t need to orchestrate individual actions itself.

​
Testing operations interactively

Before connecting an agent, you can test every discovered operation in the browser using the Swagger UI: 
open http://localhost:8787/api/sites/example-com/docs
Replace example-com with your site’s slug. The Swagger UI lists all views, actions, and workflows with their input schemas, and lets you execute them directly.

​
CORS

All API endpoints include permissive CORS headers by default: 
Access-Control-Allow-Origin: *
Access-Control-Allow-Methods: GET, POST, OPTIONS
Access-Control-Allow-Headers: Content-Type
This means you can call the BrowserWire API directly from a browser-based agent or frontend without proxy configuration.
The BrowserWire server listens on 127.0.0.1 by default, so it’s only reachable from the local machine. If you need to access it from another host (e.g., a remote agent), start the server with browserwire --host 0.0.0.0 and ensure the port is accessible.