PageIndex

PageIndex MCP exposes PageIndex’s reasoning-based, tree-structured document retrieval directly to MCP-compatible agents and LLMs. It lets you chat with long PDFs and structured documents without relying on vector databases, while preserving full document context and clear, human-like reasoning paths.

How to use

You can connect to PageIndex MCP in two primary ways: via a remote MCP server over HTTP for direct access, or via a local stdio server that you run on your machine. Use the HTTP option when your client can directly reach a remote endpoint. Use the stdio option when you want a local, self-contained setup that handles uploads and authentication locally. In both cases you can upload PDFs or supply document URLs, and you’ll interact with PageIndex through your MCP client just like you would with other MCP services.

How to install

Prerequisites: make sure you have Node.js version 18.0.0 or newer installed on your system.

Option A: Run a local MCP server (stdio) to upload PDFs locally.

{
  "mcpServers": {
    "pageindex": {
      "command": "npx",
      "args": ["-y", "pageindex-mcp"]
    }
  }
}

Option B: Connect directly to PageIndex via HTTP (remote server). Use this when your client and network can reach the remote MCP URL.

{
  "mcpServers": {
    "pageindex": {
      "type": "http",
      "url": "https://chat.pageindex.ai/mcp"
    }
  }
}

Option C: If your client cannot directly use HTTP MCP servers, use mcp-remote as a bridge to the remote endpoint.

{
  "mcpServers": {
    "pageindex": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://chat.pageindex.ai/mcp"]
    }
  }
}

Notes on usage and capabilities

Option A provides a local PDF upload capability, allowing you to add PDFs to PageIndex and chat about them from any MCP client. Option B connects to the hosted PageIndex MCP server and supports PDF processing via URLs. Option C offers a bridge if your client does not support direct HTTP MCP servers.