Requests and responses
MCP and REST share one definition of each operation, so the field names, limits and costs are the same on both.
Operations
| MCP tool | REST | Cost | Sign-in needed |
|---|---|---|---|
| find_technologies | GET /v1/technologies | Free | Not on REST |
| count_sites | GET /v1/counts | Free | Not on REST |
| search_sites | GET /v1/sites | 1 credit per site returned | Yes |
| lookup_domain | GET /v1/domains/:domain | 2 credits, free when not covered | Yes |
| get_balance | GET /v1/balance | Free | Yes |
| buy_credits | POST /v1/checkout | Free to call | Yes |
REST paths are relative to https://api.whichsitesuse.com. Over MCP every tool needs you signed in; over REST the two free search tools also answer without a key.
Inputs
On REST, GET operations take their inputs as query parameters and /v1/checkout takes a JSON body. List inputs can be comma-separated (technologies=shopify,klaviyo) or repeated (technologies=shopify&technologies=klaviyo). Over MCP, pass real arrays and objects.
The response envelope
Every successful response, MCP or REST, wraps the result in the same envelope. Over MCP it comes back as the tool's structured content and as JSON text. Keep the attribution line when you publish results.
{
"data": { ... },
"meta": {
"crawl_date": "YYYY-MM-DD",
"credits_used": 0,
"credits_remaining": ...,
"attribution": "Technology data from HTTP Archive (httparchive.org), detected with its open-source Wappalyzer fork. Core Web Vitals from the Chrome UX Report."
}
}
credits_remaining is your balance after the call, or null for a REST call made without a key. crawl_date is the HTTP Archive crawl the answer came from.