DocRouter MCP Server

Overview

The DocRouter MCP Server is a TypeScript-based server that implements the Model Context Protocol, enabling AI assistants like Claude Desktop, Cursor IDE, and other MCP-compatible tools to directly interact with your DocRouter organization.

This server provides seamless document processing capabilities within your AI workflows, allowing AI assistants to upload documents, retrieve analysis results, and manage your document processing workflows programmatically.

Document management and retrieval
OCR text extraction and metadata
AI-powered data extraction using prompts
Tag and prompt management
Advanced search functionality

Installation

Prerequisites

Node.js 18+ installed on your system
DocRouter Account with API access
DocRouter Organization ID and API Token

Global Installation (Recommended)

Install the package globally to make the docrouter-mcp binary available system-wide:

npm install -g @docrouter/mcp

Verify the installation:

# Check if binary is available
which docrouter-mcp

# Test the installation
docrouter-mcp --help

Local Installation

For project-specific installation:

npm install @docrouter/mcp

Note: Local installation requires using the full path to the executable in your MCP configuration.

Configuration

For Cursor IDE

Create .mcp.json in your project root:

{
  "mcpServers": {
    "docrouter": {
      "command": "docrouter-mcp",
      "env": {
        "DOCROUTER_API_URL": "https://app.docrouter.ai/fastapi",
        "DOCROUTER_ORG_ID": "your-org-id",
        "DOCROUTER_ORG_API_TOKEN": "your-api-token"
      }
    }
  }
}

For Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "docrouter": {
      "command": "docrouter-mcp",
      "env": {
        "DOCROUTER_API_URL": "https://app.docrouter.ai/fastapi",
        "DOCROUTER_ORG_ID": "your-org-id",
        "DOCROUTER_ORG_API_TOKEN": "your-api-token"
      }
    }
  }
}

Important: Replace your-org-id and your-api-token with your actual DocRouter credentials.

Available MCP Tools

Once configured, the following tools become available in your AI application:

Document Management

`get_docrouter_documents()`	List all documents
`get_docrouter_document(documentId)`	Get document by ID
`get_docrouter_document_ocr(documentId)`	Get raw OCR text
`get_docrouter_document_ocr_metadata(documentId)`	Get OCR metadata

Data Extraction

`run_docrouter_extraction(documentId, promptRevId, force)`	Run AI extraction
`get_docrouter_extraction(documentId, promptRevId)`	Get extraction results

Search and Discovery

`search_docrouter_documents(query, tagIds)`	Search documents
`search_docrouter_prompts(query)`	Search prompts
`search_docrouter_tags(query)`	Search tags

Management

`get_docrouter_prompts()`	List all prompts
`get_docrouter_prompt(promptRevId)`	Get prompt by ID
`get_docrouter_tags()`	List all tags
`get_docrouter_tag(tagId)`	Get tag by ID

Help and Guidance

`docrouter_help()`	Get help information
`docrouter_document_analysis_guide(documentId)`	Generate analysis guide

Example Workflows

1. Document Analysis Workflow

// Search for invoice documents
const documents = await search_docrouter_documents("invoice");

// Get OCR text for the first document
const ocrText = await get_docrouter_document_ocr(documents.documents[0].id);

// Find available prompts for invoice processing
const prompts = await search_docrouter_prompts("invoice");

// Run extraction with a specific prompt
const extraction = await run_docrouter_extraction(
  documents.documents[0].id,
  prompts.prompts[0].id
);

2. Document Discovery Workflow

// Get all documents
const allDocuments = await get_docrouter_documents();

// Get all available prompts
const allPrompts = await get_docrouter_prompts();

// Get all tags for categorization
const allTags = await get_docrouter_tags();

3. Guided Analysis Workflow

// Get a step-by-step guide for analyzing a specific document
const guide = await docrouter_document_analysis_guide("document-id");

// Follow the guide to analyze the document
// The guide will provide specific prompts and extraction strategies

Troubleshooting

MCP Server Not Connecting

Verify the binary exists: which docrouter-mcp
Check configuration syntax in your .mcp.json
Ensure environment variables are set correctly
Test the server manually: docrouter-mcp

Command Not Found

Reinstall globally: npm install -g @docrouter/mcp
Check your PATH: echo $PATH
Use full path in configuration if needed

Environment Variables Not Set

Verify variable names: DOCROUTER_ORG_ID, DOCROUTER_ORG_API_TOKEN
Check variable values are correct and not expired
Test API access with your credentials

Enable Debug Mode

Add debug logging to your environment:

"env": {
  "DOCROUTER_API_URL": "https://app.docrouter.ai/fastapi",
  "DOCROUTER_ORG_ID": "your-org-id",
  "DOCROUTER_ORG_API_TOKEN": "your-api-token",
  "DEBUG": "mcp:*"
}

Security Best Practices

Never Commit Credentials

Add .mcp.json to .gitignore if it contains real credentials

Use Environment Variables

Store credentials in environment variables instead of hardcoded values

Rotate API Tokens

Regularly rotate API tokens and limit permissions to minimum required access