Your Entire Website as a Live AI Knowledge Base
Imagine asking Claude or ChatGPT a question about your docs site and getting back a cited, accurate answer that pulled from seventeen different pages — not a guess, not a hallucination, not a single-page fetch. That's what WebToMCP does: it indexes your entire website once and turns it into a live knowledge base any MCP-compatible AI can query on demand. Not a browser extension. Not a component-capture tool. A whole-site index with a single endpoint URL.
Not a component capture tool
There's a Chrome extension also called "Web to MCP" — it lets you click on a UI element in a web page (a button, a form, a nav item) and capture it as a reusable component for Cursor or Claude Code. It's a coding workflow tool. Good product, completely different job.
If you found this page looking for that extension, web-to-mcp.com is where you want to go.
If you found this page because you want to make your entire website's content queryable from Claude, ChatGPT, Cursor, or another AI assistant — with answers that cite real source URLs rather than guess — you're in the right place. Same three-word phrase, opposite workflow: we're not capturing single components from a page; we're indexing every page on a site so AI clients can answer content questions from the whole thing.
One-sentence contrast. web-to-mcp.com (the Chrome extension): capture one UI component from a page for a coding agent. WebToMCP (webtomcp.net, this site): index every URL of a website so AI assistants answer content questions with citations from the real pages.
What WebToMCP actually does
WebToMCP is a hosted indexing service for public websites. You give us a URL, we crawl every accessible page on it, build a hybrid vector + keyword search index on Cloudflare's edge, and return one MCP endpoint URL. Three steps, about five minutes end-to-end:
-
Crawl. We fetch every URL on your site that's within the path prefix you submit, respecting
robots.txt. JavaScript-rendered pages (React, Next.js, Shopify storefronts) are supported and metered per plan. Most sites under 500 pages finish in 1–10 minutes. -
Index. Each page gets extracted, cleaned, and chunked. Chunks are embedded into a vector index and indexed for keyword search. The two indexes run in parallel at query time — the hybrid approach handles both semantic queries ("what's your refund policy?") and exact-term queries ("what does the
--verboseflag do?"). -
Endpoint URL. When indexing finishes, you get one URL — something like
https://mcp.webtomcp.net/batches/<id>/mcp?key=wmcp_…— that speaks the Model Context Protocol. Paste it into Claude Desktop's connector settings, ChatGPT's custom connectors, or Cursor's MCP config. Done. The AI now queries your whole site instead of guessing.
We re-crawl weekly by default; manual force-refresh is one click away. Indexed content stays on Cloudflare, encrypted at rest, never shared with advertisers, never used to train models.
Who it's for
WebToMCP pays off whenever your site has more than a handful of pages and people would benefit from a cited, accurate AI answer instead of having to dig through them manually:
- Docs site owners. "What's the rate limit on the webhooks API?" "Which authentication methods does the SDK support?" Today Claude or ChatGPT either guesses from a stale training snapshot or reads one docs page per question. After indexing, it reads all of them.
- E-commerce stores. "What's the refund policy on digital goods?" "Do you ship to Singapore?" "What's included in the Pro bundle?" Questions support teams answer fifty times a day, all answerable from your site's content.
- Knowledge bases and wikis. Internal or external. If it's a public URL and it has content, it can be indexed and queried. Your Notion-published wiki, your Confluence space exported as a site, your Gitbook docs — all indexable.
- Content publishers and B2B SaaS blogs. "What integrations does Pro support?" "Summarize your pricing vs Competitor X." Answers that require reading across dozens of posts, all returnable as a single cited response.
It's overkill for a five-page brochure site. It's clear-cut value for anything content-heavy — docs, stores, wikis, changelogs, marketing blogs with deep product detail.
Try it without signing up
Three live demo endpoints — Cloudflare Workers docs, Tailwind CSS docs, Hono docs — are ready on the landing page. Copy any one into your AI client (no account needed) and ask something you'd normally have to search docs for. You'll see the citation-backed answer format that the indexed version of your own site would return.
Free tier: 1 knowledge base, ~500 AI questions per month, no credit card. Ready to index your own URL? Sign in with Google — the free tier doesn't expire.
Related reads
- What a website MCP endpoint is — and how to get one in 5 minutes
- Connect your website to Claude Desktop using MCP
- Connect your website to ChatGPT (every tier, including Free)
- All setup guides — Cursor, Gemini CLI, Codex.
Questions? developer@webtomcp.net. Or sign in with Google to index your own site (free).