# Fetch Router

> Verified web fetch for AI agents. One endpoint routes across plain HTTP, a headless browser and unblocking providers, checks each page for blocks, CAPTCHAs, login walls and empty JS shells, falls back on failure, and bills only for pages that pass.

Use it when an agent needs the real content of a public web page and plain HTTP fetches keep returning challenge pages, empty JavaScript shells or 200 responses with nothing in them.

## API
- POST https://fetchrouter.dev/v1/fetch with JSON {"url": "...", "mode": "markdown"|"html"|"rendered", "expect": {"text": "..."}, "maxCost": 5}
- Header: Authorization: Bearer <key>. Free key: POST https://fetchrouter.dev/v1/signup {"email": "..."} (500 credits, no card).
- Response: ok, content, verdict {passed, checks, reason}, provider, attempts[], billed, costCredits.
- Price: 1 credit static, 5 browser, 10-25 premium unblock; failures free. 1 credit = $0.001.
- GET https://fetchrouter.dev/v1/reliability?domain=example.com: which routes work for a domain (no key).

## MCP
- npx -y fetchrouter-mcp with env FETCHROUTER_API_KEY. Tools: fetch_page, domain_reliability.

## Docs
- [API reference](https://fetchrouter.dev/docs)
- [Reliability by site type and route](https://fetchrouter.dev/reliability)
- [Use policy](https://fetchrouter.dev/docs#policy)