MCP Integration

Docsieve includes native support for the Model Context Protocol (MCP), allowing AI assistants like Cursor, Claude Code, and GitHub Copilot to query your documentation spaces directly from their workspace sessions.

Setting Up the MCP Server

You can run the MCP server using either the local open-source package or the hosted cloud API.

1. Using the Local Python Package

If you have installed docsieve via pip, start the stdio server directly:

python -m docsieve_hosted.mcp

2. registering the Server in AI Clients

Add the configuration block to your AI client’s settings:

For Claude Code (~/.claude.json)

{
  "mcpServers": {
    "docsieve": {
      "command": "python",
      "args": ["-m", "docsieve_hosted.mcp"]
    }
  }
}

For Cursor (settings.json or Custom MCP Settings UI)

1. Open Cursor Settings > MCP.
2. Click Add New MCP Server.
3. Set the name to docsieve.
4. Set the type to command.
5. Enter the command: python -m docsieve_hosted.mcp.

Available MCP Tools

Once connected, your AI assistant will have access to the following tools:

• search_briefs: Perform semantic searches across compiled documentation briefs.
• get_brief: Retrieve the full executive brief for a specific documentation space.
• query_pages: Fetch the raw parsed Markdown content of individual pages in the tree.
• check_changelog: Retrieve version diffs and changes detected during the last recrawl.