@mrazmat/mcp-replicate

MCP Replicate Server

A Model Context Protocol (MCP) server for running machine learning models on Replicate.

Setup

1. Install the package:

npx -y @mrazmat/mcp-replicate

2. Set your Replicate API key as an environment variable:

export REPLICATE_API_KEY="your-api-key-here"

3. Configure Claude Desktop by adding this to your claude_desktop_config.json:

{
  "mcpServers": {
    "replicate": {
      "command": "npx",
      "args": ["-y", "@mrazmat/mcp-replicate@0.1.5"],
      "env": {
        "REPLICATE_API_KEY": "your-api-key-here"
      }
    }
  }
}

Important Usage Notes

When asking Claude to generate images, use these guidelines:

1. Do NOT include SHA hashes/version numbers after the model name

   • Correct: "black-forest-labs/flux-schnell"
   • Incorrect: "black-forest-labs/flux-schnell:462ce631b56105646c5442fb7..."

2. Format your prompts clearly. Example:

   Generate an image using the Flux model on Replicate of a serene Japanese garden with cherry blossoms.
   Use "black-forest-labs/flux-schnell" as the model.
   

3. Specify detailed parameters if needed:

   Generate an image using Flux of two Asian leopard cats playing in a backyard.
   Use model "black-forest-labs/flux-schnell" with:
   - prompt: "two Asian leopard cats playing in a backyard, detailed fur, natural lighting"
   - negative_prompt: "deformed, unrealistic, cartoon-like"
   - guidance_scale: 7.5
   

Example Models

Some popular Replicate models you can try:

• "black-forest-labs/flux-schnell" - High quality image generation
• "stability-ai/sdxl" - Stable Diffusion XL
• "meta/llama-2-70b-chat" - For text generation

Development

# Install dependencies
pnpm install

# Build the project
pnpm build

# Watch for changes
pnpm watch

# Run in development mode
pnpm dev

License

MIT