MCP Server for Replicate

A FastMCP server implementation for interfacing with Replicate’s API. This server provides tools for accessing various AI models hosted on Replicate through a standardized interface.

Current Status: Early Alpha

This project is in early alpha development. Features and APIs may change significantly.

Currently Supported

• Image generation models with:
  • Model schema inspection
  • Image generation with customizable parameters
  • Output resizing and optimization

Roadmap

Planned Features

1. Text Generation

   • Support for text completion models
   • Chat model integration
   • Streaming support for real-time responses

2. Video Generation

   • Support for video generation models
   • Video output handling and optimization
   • Progress tracking for long-running generations

3. Additional Features

   • Model version management
   • Better error handling and retries
   • Caching for frequently used models
   • Rate limiting and queue management

Setup

1. Install dependencies:

pip install -r requirements.txt

2. Set up your Replicate API token in .env:

REPLICATE_API_TOKEN=your_token_here

3. Run the server:

fastmcp dev server.py