Overview

Opik MCP Server provides a standardized interface to the Opik platform through the Model Context Protocol (MCP). This implementation allows developers to access Opik's capabilities directly from their development environment, making it easier to manage AI/LLM workflows.

Installation

Prerequisites

• Node.js v20.11.0 or higher
• npm or yarn package manager
• An Opik account with API access

Setup Instructions

1. Clone the repository

   git clone https://github.com/comet-ml/opik-mcp.git
   cd opik-mcp
   

2. Install dependencies

   npm install
   

3. Build the project

   npm run build
   

4. Configure environment variables Create a .env file based on the provided example:

   cp .env.example .env
   

   Edit the .env file with your Opik API credentials and preferences.

Usage

Starting the Server

The MCP server can be started with different transport mechanisms:

Standard I/O (recommended for IDE integration)

npm run start:stdio

Server-Sent Events (experimental)

npm run start:sse

IDE Integration

Cursor IDE

To integrate with Cursor IDE, create a .cursor/mcp.json file in your project directory with the following configuration:

{
  "mcpServers": {
    "opik": {
      "command": "/path/to/node",
      "args": [
        "/path/to/opik-mcp/build/index.js",
        "--apiUrl",
        "https://www.comet.com/opik/api",
        "--apiKey",
        "YOUR_API_KEY",
        "--workspace",
        "default",
        "--debug",
        "true"
      ]
    }
  }
}

Replace /path/to/node with your Node.js executable path and /path/to/opik-mcp/build/index.js with the actual path to the built index.js file. Also, replace YOUR_API_KEY with your actual Opik API key.

Available Tools

The Opik MCP Server provides access to the following tools:

Prompts Management

• Create, list, update, and delete prompts
• Organize prompts by project/workspace
• Version control for prompts

Projects/Workspaces Management

• Create and manage projects
• Organize resources within workspaces
• Set access controls and permissions

Traces

• Track execution traces for LLM interactions
• Analyze performance and behavior
• Debug issues with detailed trace information

Metrics

• Gather performance metrics
• Monitor usage statistics
• Generate reports on model behavior

Docker Deployment

For containerized deployment, you can use the provided Dockerfile:

# Build the Docker image
docker build -t opik-mcp .

# Run the container
docker run -p 3000:3000 --env-file .env opik-mcp

Troubleshooting

If you encounter issues:

1. Check that your API key is valid and has the necessary permissions
2. Ensure the correct API URL is configured
3. Verify that the workspace exists and is accessible
4. Enable debug mode for more detailed logs

For additional help, join the Comet Slack community or refer to the Opik documentation.