MCPPoint Claude or Cursor at the whole corpus. Open, no key.

Scrape any website. Then search everything.

A web scraping API built for agents. Point it at a domain and it crawls every page — blog, docs, pricing, changelog — converts them to clean markdown, scores what each one actually is, and refreshes on a schedule. Then search across all of it at once, by meaning or by exact words, over MCP or HTTP.

claude mcp add stealpage --scope user --transport http https://stealpage.com/api/mcp

In any terminal

Search stealpage for how Stripe documents webhook retries, then answer with the source URLs and when each page was last fetched.

Then, in the same session

Check the MCP serverOpen · no key

Works with

ClaudeOpenAICursorVS CodeWindsurfGemini CLIJetBrains AIWarpRaycastn8nClaudeOpenAICursorVS CodeWindsurfGemini CLIJetBrains AIWarpRaycastn8nClaudeOpenAICursorVS CodeWindsurfGemini CLIJetBrains AIWarpRaycastn8nClaudeOpenAICursorVS CodeWindsurfGemini CLIJetBrains AIWarpRaycastn8n
Scrape

Any domain to clean markdown. Six fetch tiers, from plain HTTP up to a real browser, so a bot wall is an escalation rather than a failure.

Search

One index across every site you have crawled. Vector and full text fused on rank, filtered by what a page IS — technical, promotional, original.

Serve

MCP server, JSON API and .md endpoints. Your agent reads the corpus instead of guessing from training data.

405
Pages indexed
10
Companies
1,562
Embedded chunks
Split on headings, never a token count
27,855
Internal links mapped
The graph the crawl walks
89
Languages
A translation is a document, not a variant

Most cited by their own sites

See all

Body links only, never the navbar — counting the template ranks /login top on every site on the web.

Technical, and not a pitch

See all

Scored at index time. The word “technical” appears in almost no technical article and in every marketing page that wants to sound like one.

Published this month

See all

The corpus refreshes on a schedule and diffs every page against its own past, so new is a fact rather than a guess.

Reference documentation

See all

The half of a company's writing nobody links to from the blog, and the half an agent actually needs.

What shipped

See all

Changelogs, release notes and announcements — the only continuous record of what a company actually built.

Companies in the index

See all

Ordered by how much of them we hold, because that is what makes one worth reading here.