DocRouter MCP Server — AI Assistant Integration

Model Context Protocol server for seamless AI assistant integration

View on GitHub

Overview

The DocRouter MCP Server enables AI agents to interact directly with DocRouter.AI for document processing workflows.

• Document management and retrieval
• AI-powered document processing
• Schema and prompt configuration
• Web forms configuration

Installation

Prerequisites

• Node.js 18+ installed on your system

Installation

npm install -g @docrouter/mcp

Verify installation:

which docrouter-mcp
docrouter-mcp --help

Configuration

Configure the MCP server for your AI assistant:

• Cursor IDE: Create .mcp.json
• Claude Desktop: Update claude_desktop_config.json
• Claude Agent: Configure MCP server

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

Replace your-org-api-token with your actual DocRouter organization API token. The organization ID is automatically resolved from this token and does not need to be configured.

Example Prompts

Create a Schema

Create a schema for extracting invoices with these fields:
- invoice_number (string, required)
- date (string, format: date)
- total_amount (number, required)
- vendor_name (string)
- line_items (array of objects). The line_items object: 
  - description 
  - quantity 
  - unit_price
  - total

Update a Schema

Update the invoice schema to add a new optional field:
- payment_terms (string)

Create a Prompt

Create a prompt for extracting invoice data from documents.

Configure it with the invoice schema, and tag it with a tag named 'green'

Update a Prompt

Update the invoice extraction prompt to ensure the date is in YYYY-MM-DD format."

What the MCP Server Can Do

The DocRouter MCP server exposes a rich set of tools for documents, OCR, LLM extraction, tags, prompts, schemas, forms, and chat. Instead of listing every tool here, refer to the MCP README and in-assistant help() commands for the live, authoritative list.

• Upload, list, update, and delete documents
• Access OCR text and structured OCR blocks
• Run LLM extraction and retrieve results
• Manage tags, prompts, schemas, and forms
• Run LLM chat and knowledge-base–style workflows
• Call help(), help_prompts(), help_schemas(), and help_forms() from your assistant for detailed usage

Learn More

• Claude Code Integration — Use the DocRouter MCP server from Claude Code.
• Cursor Integration — Use the DocRouter MCP server from Cursor.
• GitHub Copilot Integration — Use the DocRouter MCP server from GitHub Copilot.
• REST API — Learn about the underlying HTTP endpoints the MCP server calls.
• Knowledge Bases — Provide retrieval context for extraction and chat workflows.
• Knowledge Base Chat — Build chat experiences on top of DocRouter.

View on GitHub Get API Credentials