Paste a URL. We crawl it, semantically index it, and give you a hosted endpoint every AI agent — and every automation tool — can connect to instantly.
Claude Desktop config
{
"mcpServers": {
"my-docs": {
"url": "http://sitemcp.clea.design/api/mcp/docs-example-com-k8x2p",
"transport": "sse"
}
}
}That's it. Your AI can now semantically search, read, and summarize your entire site.
Your website has everything your AI needs — but there's no standard way for agents to read it. Until now.
AI agents can't read your website
SiteMCP crawls every page and exposes it as structured, semantically searchable data any AI can query in real time.
Copy-pasting docs into prompts is broken
Stop stuffing context windows. Your agent calls search_site and gets exactly what it needs — nothing more.
Your knowledge base is invisible to AI tools
One URL. One endpoint. Every page becomes instantly queryable from Claude, Cursor, Windsurf, Zed, VS Code, or ChatGPT.
New content never reaches your AI tools
GitHub integration and auto re-crawl keep your endpoint in sync — automatically on every release or on a daily/weekly schedule.
The MCP protocol solves this — but building a server from scratch takes days. SiteMCP does it in 60 seconds.
Give your AI agents eyesThree steps from URL to live MCP endpoint
Enter any publicly accessible website or documentation URL. Set crawl filters (include/exclude patterns) to index exactly what you need.
Firecrawl extracts clean markdown from every page. OpenAI embeddings index it for semantic search. You get an email when ready.
One-click configs for 6 AI clients. REST API for ChatGPT. Slack, n8n, and GitHub integrations. All in one dashboard.
Not just an endpoint — a full platform for managing AI access to your web content, with every integration built in.
OpenAI embeddings power intent-aware search. Your agent finds what it means, not just what it says. Full-text fallback included.
See how agents use your endpoint — which tools they call, which sites they search, and what queries they run. 30-day history.
Generate a secret key per site. Only agents with your key can query the endpoint. Rotate or revoke anytime.
Schedule daily or weekly re-crawls. Your endpoint stays fresh without lifting a finger.
Connect your repo. Every time you publish a GitHub release, SiteMCP automatically re-crawls and re-indexes your site.
Browse to any website and click "Add to SiteMCP" — opens the dashboard with the URL pre-filled. Zero friction.
Paste your webhook URL. On every crawl complete or fail, SiteMCP fires a structured payload you can chain into any automation.
Get a formatted Slack message when a crawl completes or fails — right in your channel, with a link to the endpoint.
Copy the exact config for Claude Desktop, Cursor, Windsurf, Zed, or VS Code — pre-filled with your endpoint and API key.
One click generates a full OpenAPI 3.1 spec. Paste it into Custom GPT builder and your site is searchable from ChatGPT.
Include or exclude URL patterns. Index only /docs, skip /changelog — full control over what gets exposed to AI.
Optionally list your endpoint in the public MCP directory so other developers can discover and connect to it.
One endpoint, every AI client, every automation tool — all configured from a single dashboard.
One-click configs for every major AI coding assistant
Claude Desktop
claude_desktop_config.json
Cursor
.cursor/mcp.json
Windsurf
.codeium/windsurf/mcp_config.json
Zed
settings.json via mcp-remote
VS Code
.vscode/mcp.json
ChatGPT
GPT Actions / OpenAPI spec
Fire webhooks and get alerts when crawls complete
Slack
Crawl complete / failed alerts
n8n
HTTP webhook on crawl events
Zapier
HTTP webhook on crawl events
Make.com
HTTP webhook on crawl events
Resend — crawl + quota alerts
Keep your endpoint in sync with your development workflow
GitHub
Auto-crawl on release published
Chrome
One-click extension to add sites
REST API
Direct HTTP endpoints for any client
MCP SSE
Standard SSE transport for MCP
Need a different integration? The REST API works with any HTTP-capable tool.
View all integrations in the dashboardEvery MCP endpoint exposes these tools. Also available as REST endpoints for ChatGPT and HTTP clients.
search_sitesemanticSemantic search powered by OpenAI embeddings. Finds the most relevant pages by meaning, not keywords. Falls back to full-text automatically.
Input: { query: "how to install", limit: 5 }get_pageFetch the full markdown content of any crawled page by URL. Ideal for deep dives into specific topics or docs.
Input: { url: "https://docs.example.com/quickstart" }list_pagesPaginate through all crawled pages with their titles and URLs. Useful for discovery and exploration.
Input: { page: 1, limit: 20 }get_site_summaryHigh-level overview — page count, topics covered, and last crawled timestamp. Good for agent orientation.
Input: {}Also available as REST API
All 4 tools are mirrored at http://sitemcp.clea.design/api/rest/[slug]/search, /pages, /page, /summary — CORS-enabled, API key auth supported. Used by ChatGPT GPT Actions and any HTTP client.
One endpoint, infinite possibilities
Let your AI rep answer product, pricing, and feature questions sourced live from your docs — no hallucinations.
Build a site-aware Claude chatbot with zero backend. Just your URL and one line of JSON config.
Give Cursor or Windsurf instant access to any API docs without copy-pasting. Stays fresh on every release.
Generate an OpenAPI spec in one click and add your site as a GPT Action. Your knowledge, inside ChatGPT.
Turn your internal wiki or docs site into a queryable knowledge base for your entire AI workflow stack.
Trigger n8n or Zapier flows when a crawl completes — notify teams, update databases, post to Slack.
Index competitor sites and query their content from any MCP-compatible client. Research at AI speed.
Connect agents to industry blogs, research portals, or docs — make any web content programmable.
Start free, scale when you need it. All integrations included on every plan.
Most sites are ready in under 2 minutes. Large sites (100+ pages) may take up to 10 minutes. You'll get an email the moment it's done.
Claude Desktop, Cursor, Windsurf, Zed, VS Code, and ChatGPT (via GPT Actions). One-click configs are generated for all of them. Any custom SSE MCP client also works.
In site settings, click "Connect GitHub Repo" to get a webhook URL and secret. Add it to your repo under Settings → Webhooks → Releases. Every published release auto-triggers a re-crawl.
Instead of matching keywords, semantic search uses OpenAI embeddings to find pages that match the intent of the query — even if the exact words differ.
Paste your Slack incoming webhook URL or n8n/Zapier/Make endpoint into site settings. On every crawl complete or fail, SiteMCP fires a structured payload to it automatically.
Browse to any website, click the SiteMCP extension icon, and hit "Add to SiteMCP". It opens your dashboard with the URL pre-filled — no copy-pasting needed.
Yes. In the endpoint display, open the ChatGPT Actions tab — it generates a full OpenAPI 3.1 spec pointing at SiteMCP's REST API. Paste it into a Custom GPT's Actions builder and you're done.
Yes. Generate an API key per site. Only agents that present the key can query the endpoint. Without it, requests return 401.
Choose Off, Daily, or Weekly in site settings. A cron job checks schedules and triggers re-crawls automatically. You get an email on completion.
Yes. Set include patterns (e.g. /docs/*) and exclude patterns (e.g. /blog/*) in Crawl Filters. Applied on every crawl — including auto and GitHub-triggered ones.
Stop copy-pasting. Stop maintaining custom context. One URL gives Claude, Cursor, VS Code, ChatGPT — and your whole automation stack — live access to your content.
Start for freeNo credit card required. Free tier forever.