MCP Server Replicate

MCP Server Replicate

A FastMCP server implementation for the Replicate API, providing resource-based access to AI model inference with a focus on image generation.

Features

• Resource-based image generation and management
• Real-time updates through subscriptions
• Template-driven parameter configuration
• Comprehensive model discovery and selection
• Webhook integration for external notifications
• Quality and style presets for optimal results
• Progress tracking and status monitoring
• Secure API key management

Available Prompts

The server provides several specialized prompts for different tasks:

Text to Image (Primary)

Our most thoroughly tested and robust prompt. Optimized for generating high-quality images from text descriptions with:

• Detailed style control
• Quality presets (draft, balanced, quality, extreme)
• Size and aspect ratio customization
• Progress tracking and real-time updates

Example:

Create a photorealistic mountain landscape at sunset with snow-capped peaks, quality level: quality, style: photorealistic

Other Prompts

• Image to Image: Transform existing images (coming soon)
• Model Selection: Get help choosing the right model for your task
• Parameter Help: Understand and configure model parameters

Prerequisites

• Python 3.11 or higher
• A Replicate API key
• UV for dependency management

Installation

Installing via Smithery

To install MCP Server Replicate for Claude Desktop automatically via Smithery:

npx -y @smithery/cli install @gerred/mcp-server-replicate --client claude

Installing Manually

You can install the package directly from PyPI:

uv pip install mcp-server-replicate
uvx install mcp-server-replicate
pip install mcp-server-replicate

Claude Desktop Integration

1. Make sure you have the latest version of Claude Desktop installed

2. Open your Claude Desktop configuration:

   # macOS
   code ~/Library/Application\ Support/Claude/claude_desktop_config.json
   
   # Windows
   code %APPDATA%\Claude\claude_desktop_config.json
   

3. Add the server configuration using one of these options:

   {
     "globalShortcut": "Shift+Alt+A",
     "mcpServers": {
       "replicate": {
         "command": "uv",
         "args": ["tool", "run", "mcp-server-replicate"],
         "env": {
           "REPLICATE_API_TOKEN": "APITOKEN"
         },
         "cwd": "$PATH_TO_REPO"
       }
     }
   }

   Copy

4. Set your Replicate API key:

   # Option 1: Set in your environment
   export REPLICATE_API_TOKEN=your_api_key_here
   
   # Option 2: Create a .env file in your home directory
   echo "REPLICATE_API_TOKEN=your_api_key_here" > ~/.env
   

5. Restart Claude Desktop completely

You should now see the 🔨 icon in Claude Desktop, indicating that the MCP server is available.

Usage

Once connected to Claude Desktop, you can:

• Generate images with natural language:

  Create a photorealistic mountain landscape at sunset with snow-capped peaks
  

• Browse your generations:

  Show me my recent image generations
  

• Search through generations:

  Find my landscape generations
  

• Check generation status:

  What's the status of my last generation?
  

Troubleshooting

Server not showing up in Claude Desktop

1. Check the Claude Desktop logs:

   tail -n 20 -f ~/Library/Logs/Claude/mcp*.log
   

2. Verify your configuration:

   • Make sure the path in claude_desktop_config.json is absolute
   • Ensure UV is installed and in your PATH
   • Check that your Replicate API key is set

3. Try restarting Claude Desktop

For more detailed troubleshooting, see our Debugging Guide.

Documentation

• Implementation Plan
• Contributing Guide
• API Reference
• Resource System
• Template System

Development

1. Clone the repository:

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

2. Install development dependencies:

   uv pip install --system ".[dev]"
   

3. Install pre-commit hooks:

   pre-commit install
   

4. Run tests:

   pytest
   

Contributing

We welcome contributions! Please see our Contributing Guide for details.

License

This project is licensed under the MIT License - see the LICENSE file for details.