How to scout a website

scout is Unsurf’s seam-discovery path. Use it after, or instead of, browser work when the website exposes an underlying API worth reusing.

Basic scout

curl -X POST $UNSURF_URL/tools/scout \
  -H "Content-Type: application/json" \
  -d '{
    "url": "https://example.com",
    "task": "find the search API"
  }'

A good task names the interaction or seam you want, not just “explore the site.”

Good	Weak
”find the contact-form submit endpoint"	"find APIs"
"capture the dashboard data loader"	"explore the dashboard"
"map product search requests"	"test the store”

What comes back

The response includes:

siteId
endpointCount
pathId
openApiSpec

Use pathId with worker. Use the OpenAPI spec when you want a portable description of the discovered seam.

Where scout fits now

Use the browser runtime when the user’s authenticated session is the thing. Use scout when:

the page exposes a stable API you want to call repeatedly
direct HTTP replay will be cheaper or simpler than repeated UI work
you want typed interface material for another integration

Authenticated sites

Hosted scout supports explicit cookies or headers, but for user-session work the preferred product path is still the local authenticated browser runtime. Scout is for seam discovery; local MCP is for operating the browser the user already trusts.

Publish or keep local

Set publish: true when the discovered seam belongs in the shared Directory. Otherwise keep the result private to your deploy.