Agility Knowledgebase MCP Server

The Agility Knowledgebase MCP server gives AI assistants direct access to the Agility CMS documentation. It implements the Model Context Protocol (MCP) standard, allowing tools like Claude, ChatGPT, Gemini, Cursor, Windsurf, and other MCP-compatible agents to search and read documentation articles without leaving their interface.

This is separate from the Agility CMS MCP server, which provides content management operations (creating, updating, and managing content). The Knowledgebase MCP server is read-only and focused on searching the documentation you're reading right now.

Why Use It

When you're building with Agility CMS and using an AI coding assistant, the assistant can query the Agility documentation directly to answer your questions, look up API details, or find relevant guides — all without you needing to copy and paste documentation into your chat.

Example use cases:

  • Ask your AI assistant "How do I authenticate with the Content Fetch API?" and it searches the docs for you
  • Get full article content pulled into context while debugging an integration
  • Let ChatGPT or Gemini crawl the Agility docs as part of a broader research task

Server URL

https://docs.agilitycms.com/docs/api/mcp

Available Tools

search_docs

Searches across all Agility CMS documentation articles. Returns results formatted as markdown with full URLs, descriptions, categories, and content snippets.

Parameters:

  • query (string, required) — Your search query. Natural language works best.
  • page (number, optional) — Page number for pagination (0-based). Omit for the first page.

Returns: Up to 10 results per page, with total hit count and page information. Each result includes a link to the full article and an objectID you can pass to fetch_doc for the complete content.

fetch_doc

Retrieves the full content of a specific article by its ID. Use this after searching to get complete article text for in-depth analysis.

Parameters:

  • objectID (string, required) — The article's ID from search results.

Returns: Full article content formatted as markdown, including the title, full URL, description, section headings, category, and body text.

Setup

Claude Desktop

  1. Open Settings > Developer
  2. Click Edit Config to open claude_desktop_config.json
  3. Add the server:
{
  "mcpServers": {
    "agility-knowledgebase": {
      "url": "https://docs.agilitycms.com/docs/api/mcp"
    }
  }
}
  1. Restart Claude Desktop

Claude Code (CLI)

Add to your project's .mcp.json:

{
  "mcpServers": {
    "agility-knowledgebase": {
      "url": "https://docs.agilitycms.com/docs/api/mcp"
    }
  }
}

Or add it via the command line:

claude mcp add agility-knowledgebase --transport http https://docs.agilitycms.com/docs/api/mcp

ChatGPT

  1. Go to Settings > Connectors
  2. Click Add connector
  3. Enter the server URL: https://docs.agilitycms.com/docs/api/mcp
  4. Test the connection

Gemini

  1. Open Settings > Extensions
  2. Select Add Extension > Connect to MCP Server
  3. Enter the server URL: https://docs.agilitycms.com/docs/api/mcp
  4. Confirm the connection

Cursor

Add to .cursor/mcp.json in your project root:

{
  "mcpServers": {
    "agility-knowledgebase": {
      "url": "https://docs.agilitycms.com/docs/api/mcp"
    }
  }
}

Windsurf

Add to your Windsurf MCP configuration file:

{
  "mcpServers": {
    "agility-knowledgebase": {
      "url": "https://docs.agilitycms.com/docs/api/mcp"
    }
  }
}

Other MCP Clients

Any MCP-compatible client that supports Streamable HTTP transport can connect. Use the server URL:

https://docs.agilitycms.com/docs/api/mcp

No authentication is required. The server supports the standard MCP protocol (JSON-RPC 2.0 over HTTP) and does not require SSE support from the client.

Technical Details

  • Transport: Streamable HTTP (compatible with all major MCP clients including ChatGPT and Gemini)
  • Authentication: None required — the documentation is public
  • Rate limits: Standard Vercel function limits apply
  • Response format: Markdown over JSON-RPC 2.0

The server is built with the mcp-handler package and deployed on Vercel. It queries the same Algolia search index that powers the search bar on this documentation site.