Replicate MCP Server

A Model Context Protocol server implementation for Replicate. Run Replicate models through a simple tool-based interface.

Installation

npm install -g mcp-replicate

Or install from source:

git clone https://github.com/deepfates/mcp-replicate
cd mcp-replicate
npm install
npm run build

Features

Models

• Search models using semantic search
• Browse models and collections
• Get detailed model information and versions

Predictions

• Create predictions with text or structured input
• Track prediction status
• Cancel running predictions
• List your recent predictions

Image Handling

• View generated images in your browser
• Manage image cache for better performance

Configuration

The server needs a Replicate API token to work. Set it as an environment variable:

export REPLICATE_API_TOKEN=your_token_here

Usage

If installed globally:

mcp-replicate

Or run directly with npx:

npx mcp-replicate

Or if installed from source:

npm start

The server uses stdio to communicate via the MCP protocol. You'll need an MCP client to interact with it.

Available Tools

Model Tools

• search_models: Find models using semantic search
• list_models: Browse available models
• get_model: Get details about a specific model
• list_collections: Browse model collections
• get_collection: Get details about a specific collection

Prediction Tools

• create_prediction: Run a model with your inputs
• get_prediction: Check a prediction's status
• cancel_prediction: Stop a running prediction
• list_predictions: See your recent predictions

Image Tools

• view_image: Open an image in your browser
• clear_image_cache: Clean up cached images
• get_image_cache_stats: Check cache usage

Examples

Here are some examples using an MCP client:

Search Models

// Find text-to-image models
const results = await mcpClient.invoke("search_models", {
  query: "text to image models"
});

Create a Prediction

// Run a model
const prediction = await mcpClient.invoke("create_prediction", {
  version: "model_version_id",
  input: {
    prompt: "A friendly robot making pancakes"
  }
});

Check Prediction Status

// Get prediction results
const status = await mcpClient.invoke("get_prediction", {
  prediction_id: "prediction_id"
});

Browse Collections

// List featured collections
const collections = await mcpClient.invoke("list_collections", {});

Development

1. Install dependencies:

npm install

2. Start development server (with auto-reload):

npm run dev

3. Check code style:

npm run lint

4. Format code:

npm run format

Requirements

• Node.js >= 18.0.0
• TypeScript >= 5.0.0

License

MIT