webtomcp

Your AI now knows
every page
of your website.

We index every URL of your site so Claude, ChatGPT, Cursor, and other MCP clients answer from the whole site with cited source links — not the one page they skimmed, not a screenshot, not a captured component. One endpoint URL, 60-second setup, no install. Free plan, no card.

500 free questions · no card · cancel anytime

  • Claude
  • Cursor
  • ChatGPT
  • Gemini
  • Codex

try right now

Three live endpoints. No signup.

Cloudflare Workers docs

100 pages indexed

Official docs for Workers, Pages, D1, KV, R2, Vectorize, and more.

Tailwind CSS docs

101 pages indexed

Utility-first CSS framework. Search class names, config, plugins.

Hono docs

100 pages indexed

Small, ultra-fast web framework for Cloudflare Workers and beyond.

Click Copy MCP URL, paste into ChatGPT → Settings → Connectors → Add custom (or Claude.ai → Settings → Connectors). Sign in to add your own site.

your client

Plug into the AI you already use.

ChatGPT

New Chat ▾

You are connected to your-docs. Ask anything from the pages WebToMCP indexed — refund policy, pricing tiers, changelog, onboarding steps. Try an example below or type your own.

Suggested

What does our refund policy say? Summarise our pricing tiers How do I onboard a new customer? Latest changelog entries
Ask anything +  ⚙ Tools 🎙

Set up in 3 minutes: paste the URL into chatgpt.com → Settings → Connectors. Works on Free, Plus, Pro, Business, Enterprise. Full guide →

Claude
ChatCoworkCode

New Chat ▾

You are connected to your-docs. Explore your indexed pages — pricing, policies, changelog, onboarding. Try an example below or ask your own.

EXAMPLE PROMPTS

Refund policy for digital goods Pro vs Business plan Onboarding a new customer Open POs by department
Reply… 🎙

Set up in 3 minutes: claude.ai → Settings → Connectors, or one wmcp connect claude for Claude Desktop. Full guide →

Using Claude Code, Cursor, Codex, or Gemini CLI? One wmcp connect per client — see the docs.

via terminal

Or wire it up from the CLI.

npm i -g webtomcp-cli · then wmcp init <url> · then wmcp connect <client>. Works with Claude Code, Cursor, Codex, Gemini CLI. See the docs →

how it works

Three steps, one minute.

01

Add your URL

Paste any docs site, blog, or product catalog. We respect robots.txt and crawl the public pages.

02

We index it

Vector + BM25 hybrid index, served from Cloudflare's edge. Most sites finish in 1–10 minutes.

03

Connect your AI

Copy the MCP URL into Claude Desktop, Cursor, ChatGPT, or any MCP client. Ask your AI about your stuff.

pricing

Pay when you outgrow free.

Free

$0 / forever

For trying it on a personal blog or evaluation site.

  • 1 knowledge base · 1 website
  • ~70 AI questions / day (2,000 / mo)
  • 5 JS-rendered pages / mo
  • Re-crawl weekly
Start free →

Hobby

$9 / month

For personal blogs, side projects, and solo devs. About $0.30/day.

  • 5 knowledge bases · 10 websites total
  • ~165 AI questions / day (5,000 / mo)
  • 100 JS-rendered pages / mo
  • Weekly re-crawl

Need more? Email developer@webtomcp.net — we'll figure something out.

faq

Frequent questions.

What's an MCP server?

Model Context Protocol — Anthropic's open spec for letting AI clients talk to external data sources. Claude Desktop, Cursor, and several others speak it natively. WebToMCP turns your website into one of these "servers" without you writing code.

Does the AI see the latest content?

We refresh weekly by default. You can also click Re-crawl any time to force a fresh pass. Each query against the MCP endpoint hits the latest indexed content.

JavaScript-heavy sites — Shopify, Notion, SPAs?

By default we use static fetch (fast, free). For sites where content loads via JavaScript, you can flip Enable JS rendering per-page. Browser rendering is metered: 5 pages / month on Free, 100 / month on Hobby.

What can I actually ask my AI?

Anything you'd ask a smart colleague who'd just read your site cover-to-cover. "What's our refund policy?", "Compare these two products", "Summarize the API auth flow". Citations include source URLs.

Is my crawled content safe? Who can access my MCP endpoint?

Each knowledge base has a unique URL-keyed endpoint — only someone holding that URL can query it. We store indexed content on Cloudflare (Workers, D1, Vectorize), encrypted at rest, never shared with advertisers, never used to train models. Inside the authenticated dashboard no third-party trackers load. Full details on /legal/privacy.

Free forever?

Yes. The free tier doesn't expire. Quotas reset every month. Hobby ($9/mo) is for use cases that hit the free caps.

Who's building this?

Solo founder — KY (Taipei). Background in dev tooling and AI workflows. Reachable at developer@webtomcp.net. The CLI is open-source on npm; product is closed beta during the storefront opening period.

Status: free beta. developer@webtomcp.net for design partner / early-access enquiries.