July 2, 2026 · 5 min read
Claude Desktop supports MCP — the Model Context Protocol — which lets Claude use external tools during a conversation. One of the most useful things you can do with this: point Claude at a website and ask it questions without ever leaving the app. Here's how.
AgentReady is an MCP server that indexes websites and lets you query them. Once you add it to Claude Desktop, you can ask Claude things like "what does stripe.com say about idempotency keys?" or "how does supabase.com handle row-level security?" and get cited answers from the actual content.
Open your Claude Desktop config file. On macOS it's at:
~/Library/Application Support/Claude/claude_desktop_config.jsonOn Windows:
%APPDATA%\Claude\claude_desktop_config.jsonAdd AgentReady to your mcpServers object:
{
"mcpServers": {
"agentready": {
"command": "npx",
"args": ["-y", "@agentreadyweb/mcp"]
}
}
}If you already have other MCP servers configured, just add the agentready entry to the existing object — don't replace the whole thing.
Close and reopen Claude Desktop. On the next launch it'll download and start the MCP server via npx. You should see a tools icon in the chat interface indicating MCP tools are active.
Before you can query a site, it needs to be indexed. You can do this two ways:
Via the portal: Go to agentready.it.com, paste a URL, and click the button. Takes about 60 seconds.
Via Claude: Just ask Claude to submit a site. Say something like "use AgentReady to index docs.stripe.com". Claude will call the submit_site tool and index it inline. (As of recent updates, ask_site also auto-indexes sites it hasn't seen before, so you can skip this step and just ask your question.)
Now ask Claude anything about the indexed site:
"What does docs.stripe.com say about webhook signatures?"
"How do I set up SSO on supabase.com?"
"What are the rate limits on the Anthropic API?"Claude will call ask_site with the domain and your question, and return an answer grounded in the actual site content — with a source link so you can verify.
Any publicly accessible website with crawlable HTML content. This covers most documentation sites — Mintlify, Docusaurus, ReadMe, GitBook, plain HTML, custom sites. Sites that require login or are pure JavaScript SPAs (single-page apps with no server-rendered HTML) don't work well yet.
The AgentReady directory lists sites that are already indexed and ready to query — you can start there without submitting anything.
Cursor also supports MCP. Add the same config block to your Cursor MCP settings and you get the same capability — ask questions about your dependencies' docs without switching context.