DocRouter MCP Server

Overview

The DocRouter MCP Server is an implementation of Anthropic's Model Context Protocol (MCP) for docrouter.ai. It enables Claude and other MCP clients to securely access and interact with docrouter.ai workspaces, providing standardized interfaces for documents, prompts, extraction schemas, and extraction results.

MCP is an open protocol that standardizes how applications provide context to LLMs. Think of MCP like a USB-C port for AI applications. Just as USB-C provides a standardized way to connect your devices to various peripherals and accessories, MCP provides a standardized way to connect AI models to different data sources and tools.

Features

Architecture

The MCP server is implemented as a separate directory and process from the main docrouter.ai application. This separation provides several benefits:

• Isolation: The MCP server can run independently, reducing the risk of affecting the main application.
• Scalability: The MCP server can be scaled separately from the main application.
• Deployment Flexibility: The MCP server can be deployed on the same or different machines as the main application.
• Development Independence: Changes to the MCP server can be made without affecting the main application.

Directory Structure

packages/docrouter_mcp/
├── __init__.py                 # Package initialization
├── docrouter_mcp_server.py     # Main server implementation
└── requirements.txt            # Server dependencies

Installation

To install the MCP server, follow these steps:

1. Clone the repository

git clone https://github.com/analytiq-hub/doc-router.git
cd doc-router

2. Install required dependencies

pip install -r packages/docrouter_mcp/requirements.txt

Usage

Running the Server

You can run the MCP server using the MCP CLI:

python packages/docrouter_mcp/docrouter_mcp_server.py --url https://app.docrouter.ai/fastapi --org-id ORG_ID --org-api-token ORG_API_TOKEN

Testing the Server

To test the server functionality:

 npx @modelcontextprotocol/inspector python docrouter_mcp_server.py --url https://app.docrouter.ai/fastapi --org-id ORG_ID --org-api-token ORG_API_TOKEN

Integration with Claude

To use this MCP server with Claude Desktop:

1. Run the MCP server
2. In Claude Desktop, go to Settings > Developer, open claude_desktop_config.json
3. Add a new MCP server with the appropriate connection details

{
    "mcpServers": {
        "docrouter": {
        "command": "/Users//build/analytiq/doc-router/packages/docrouter_mcp/.venv/bin/python",
        "args": [
            "/Users//build/analytiq/doc-router/packages/docrouter_mcp/docrouter_mcp_server.py"
        ],
        "env": {
            "DOCROUTER_URL": "https://app.docrouter.ai/fastapi",
            "DOCROUTER_ORG_ID": "",
            "DOCROUTER_ORG_API_TOKEN": ""
        }
        }
    }
}

4. Adjust the path to the python executable and to the MCP server script to match your local setup
5. Url, org-id and org-api-token need to be specified as environment variables (Claude Desktop limitation)
6. Claude will now be able to access your docrouter.ai workspace through the MCP server

Interfaces

Document Tools

• get_docrouter_documents() - List all documents in the workspace
• get_docrouter_document(document_id) - Get a specific document by ID
• get_docrouter_document_ocr(document_id) - Get the raw OCR text for a document
• get_docrouter_document_ocr_page(document_id, page_num) - Get OCR text for a specific page
• get_docrouter_document_ocr_metadata(document_id) - Get OCR metadata for a document

Tag Tools

• get_docrouter_tags() - List all tags in the workspace
• get_docrouter_tag(tag_id) - Get a specific tag by ID
• search_docrouter_tags(query) - Search tags by name or description

Prompt Tools

• get_docrouter_prompts() - List all prompts in the workspace
• get_docrouter_prompt(prompt_id) - Get a specific prompt by ID
• search_docrouter_prompts(query) - Search prompts by name or content

Search & Extraction Tools

• search_docrouter_documents(query, tag_ids) - Search documents by name or content with optional tag filtering
• get_docrouter_extraction(document_id, prompt_id) - Get extraction results for a document using a specific prompt
• run_docrouter_extraction(document_id, prompt_id, force) - Run AI extraction on a document using a specific prompt

Help Tools

• docrouter_help() - Get comprehensive help information about using the DocRouter API
• docrouter_document_analysis_guide(document_id) - Get guidance for analyzing a specific document

Example Workflows

Here are some common workflows using the MCP server:

Finding Documents

search_docrouter_documents("invoice")

Viewing Document Content

get_docrouter_document_ocr("doc123")

Running Extractions

run_docrouter_extraction("doc123", "prompt456")

Viewing Extraction Results

get_docrouter_extraction("doc123", "prompt456")