MCP Server for Replicate

What is MCP Server for Replicate?

MCP Server for Replicate is a FastMCP server that provides a standardized interface for accessing AI models hosted on Replicate’s API. It currently focuses on image generation, including model schema inspection, customizable parameters, and output resizing/optimization. This server is for developers who want to integrate Replicate models into MCP-compatible tools.

How to use MCP Server for Replicate?

Install dependencies with pip install -r requirements.txt, set your Replicate API token in a .env file as REPLICATE_API_TOKEN=your_token_here, then run the server using fastmcp dev.

Key features of MCP Server for Replicate

• Image generation with customizable parameters
• Model schema inspection for supported models
• Output resizing and optimization
• Early alpha stage with active development
• Roadmap includes text and video generation support

Use cases of MCP Server for Replicate

• Generating images from text prompts using Replicate-hosted models
• Inspecting the input/output schemas of available image models
• Integrating Replicate’s AI capabilities into MCP-enabled workflows

FAQ from MCP Server for Replicate

What models are currently supported?

Image generation models are supported, with features for schema inspection, generation with parameters, and output optimization.

How do I set up my API token?

Add REPLICATE_API_TOKEN=your_token_here to a .env file in the project directory.

What is the current status of the server?

It is in early alpha development; features and APIs may change significantly.

What are the planned future features?

The roadmap includes text generation, video generation, model version management, better error handling, caching, and rate limiting.

How do I run the server?

After setup, run fastmcp dev from the command line.