Documentation

MCP Setup & Tool Reference

Connect Claude Desktop, Cursor, Cline, or any MCP-compatible client to your IndexedAI endpoint. Your site id is shown in your dashboard after adding a site.

Setup guides

Claude Desktop

Open claude_desktop_config.json (Settings → Developer → Edit Config) and add:

{
  "mcpServers": {
    "your-domain.com": {
      "url": "https://indexedai.tech/api/mcp/{your-site-id}",
      "transport": "streamable-http"
    }
  }
}

Replace your-site-id and your-domain.com with values from your dashboard. Restart Claude Desktop after saving.

Cursor

Open Cursor Settings → MCP and add a new server with the config below, or paste directly into ~/.cursor/mcp.json:

{
  "mcpServers": {
    "your-domain.com": {
      "url": "https://indexedai.tech/api/mcp/{your-site-id}",
      "transport": "streamable-http"
    }
  }
}

Reload the Cursor window after saving. The server appears in the MCP panel.

Cline (VS Code extension)

Open the Cline sidebar → MCP Servers → Add Server. Use Streamable HTTP transport and paste the URL:

{
  "mcpServers": {
    "your-domain.com": {
      "url": "https://indexedai.tech/api/mcp/{your-site-id}",
      "transport": "streamable-http"
    }
  }
}

Or paste into cline_mcp_settings.json directly.

Tool reference

Every IndexedAI MCP endpoint exposes 3 static tools. Pro plans also get dynamic tools auto-generated from your site's crawl data.

get_indexnone

Returns the llms.txt index for the site — structured overview of purpose, sections, and important pages. Call this first before fetching individual pages.

// In Claude Desktop, Cursor, or any MCP client:
get_index()

// Returns:
// # YourSite
// > Site description
// ## Section
// - [Page](https://yoursite.com/page): Description

fetch_pageurl: string

Fetches the full text content of any publicly accessible page on the site's domain. Returns clean readable text with no HTML markup.

fetch_page({ url: "https://yoursite.com/pricing" })

// Returns the full text content of the pricing page,
// stripped of HTML, navigation, and boilerplate.

get_full_contextnone

Returns extended context from llms-full.txt — all major sections, API endpoints, pricing details, capabilities, and FAQ in one call. Use when you need a deep understanding of the site without fetching individual pages.

get_full_context()

// Returns the full llms-full.txt content:
// all sections, endpoints, pricing, and FAQ combined.

FAQ

Are there rate limits?

Free and Pro plans share fair-use limits. Heavy automated abuse will be rate-limited. For high-volume use cases, contact us.

How often does the index refresh?

TODO: Confirm refresh frequency. The index is generated when you first add your site. Manual re-scans are available from the dashboard. Automated periodic refresh is planned.

How do I remove my site?

Go to Dashboard → Sites, click the site, and select Delete. This removes the MCP endpoint and all associated scan data immediately.

Does the endpoint work without authentication?

Yes — your MCP endpoint is publicly accessible by URL. Anyone with the endpoint URL can query it. If you need to restrict access, contact us.

What transport protocol does the endpoint use?

The endpoint uses Streamable HTTP transport (the modern MCP standard). Use "streamable-http" in your client config, not "sse" or "stdio".

Ready to connect your site?

Create a free account