MCP 服务器

Medikode Medical Coding MCP Server

Enables AI assistants to access Medikode's medical coding platform for validating CPT/ICD-10 codes, performing chart quality assurance, parsing EOBs, calculating RAF scores, and extracting HCC codes from clinical documentation.

README

@medikode/mcp-server

Model Context Protocol (MCP) server for Medikode's AI-driven medical coding platform. This package enables AI assistants like Claude Desktop, Cursor, and ChatGPT to access Medikode's medical coding tools directly.

Medikode Dashboard - API Usage Trends

Medikode's AI-driven medical coding platform dashboard showing API usage trends and analytics

🌟 Features

5 Powerful MCP Tools: Validate codes, QA charts, parse EOBs, calculate RAF scores, and more
AI Assistant Integration: Works with Claude Desktop, Cursor, ChatGPT, and other MCP-compatible clients
Secure: Uses your existing Medikode API keys with the same security and access controls
Fast: Direct API access with caching for optimal performance
Easy Setup: Simple configuration with npx - no local installation required

🚀 Quick Start

Installation

npm install -g @medikode/mcp-server

Configuration

Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "medikode": {
      "command": "npx",
      "args": ["-y", "@medikode/mcp-server"],
      "env": {
        "MEDIKODE_API_KEY": "your_api_key_here"
      }
    }
  }
}

Cursor IDE

Add to your cursor_settings.json:

{
  "mcp": {
    "servers": {
      "medikode": {
        "command": "npx",
        "args": ["-y", "@medikode/mcp-server"],
        "env": {
          "MEDIKODE_API_KEY": "your_api_key_here"
        }
      }
    }
  }
}

🛠 Available Tools

1. `validate_codes`

Validates CPT/ICD-10 codes against clinical documentation.

Inputs:

chart_text (string, required): Provider note or chart excerpt
codes (array[string], required): CPT/ICD-10 codes to validate

Outputs:

valid (boolean): Whether codes are valid for the chart
recommendations (array[string]): Missing or conflicting codes

2. `qa_chart`

Performs a coding quality assurance check.

Inputs:

chart_text (string, required): Clinical documentation to review

Outputs:

issues_found (array[string]): Documentation or coding gaps
suggested_codes (array[string]): Recommended additional codes

3. `parse_eob`

Extracts structured data from Explanation of Benefits (EOB) documents.

Inputs:

eob_content (string, required): Raw EOB text (or PDF extracted text)

Outputs:

payer (string): Insurance payer name
claim_number (string): Claim reference number
total_billed (number): Total amount billed
total_allowed (number): Total amount allowed by payer
insurance_paid (number): Amount paid by insurance
patient_responsibility (number): Patient out-of-pocket amount

4. `score_raf`

Calculates RAF score and HCC capture from encounter documentation.

Inputs:

chart_text (string, required): Clinical encounter documentation

Outputs:

raf_score (float): Risk Adjustment Factor score
hcc_codes (array[string]): Hierarchical Condition Category codes

5. `multi_validate`

Composite workflow that validates chart coding and calculates RAF in one step.

Inputs:

chart_text (string, required): Clinical documentation
codes (array[string], optional): Optional codes to validate

Outputs:

validation_results (object): Results from validate_codes
raf_results (object): Results from score_raf

💡 Example Usage

Once configured, you can use the tools in your AI assistant:

User: "Validate these codes for this chart: 99213, I10, E11.9"

AI: I'll help you validate those codes using the validate_codes tool...
[Tool call to validate_codes]

Based on the validation results:
- Code 99213: Valid for established patient office visit
- Code I10: Valid for essential hypertension
- Code E11.9: Valid for type 2 diabetes without complications

🔑 Authentication

All tools require a valid Medikode API key. You can obtain one by:

Signing up at medikode.ai
Generating an API key in your account settings
Setting the MEDIKODE_API_KEY environment variable

📊 Usage Tracking

All MCP tool usage is tracked and appears in your Medikode dashboard alongside regular API calls. This includes:

Number of API calls made
Charts processed
EOBs parsed
RAF scores calculated

🔧 Troubleshooting

Common Issues

MCP Server Not Found

Ensure Node.js and npm are installed
Verify the package is available via npx: npx @medikode/mcp-server --help

Authentication Errors

Check that your API key is correct and active
Verify the MEDIKODE_API_KEY environment variable is set
Ensure your API key has the required permissions

Tool Not Available

Restart your AI client after configuration changes
Verify the MCP server configuration is correct
Ensure your AI client supports MCP

📚 Documentation

🛠 Development

Prerequisites

Node.js 18.0.0 or higher
npm or yarn
Medikode API key

Local Development

Clone the repository:

git clone https://github.com/medikode/mcp-server.git
cd mcp-server

Install dependencies:
```
npm install
```

Set up environment variables:

cp env.example .env
# Edit .env with your API key

Run in development mode:
```
npm run dev
```
Test the MCP server:
```
npm run test:routing
```

Building

npm run build

Testing

# Test WebSocket connection
node test-websocket.js

# Test ChatGPT integration
python test-chatgpt-integration.py

# Test environment routing
node test-environment-routing.js