replicate

Overview

Replicate’s official MCP server for AI models and ML workflow

Transport

sse

Tools

• list_collections
• get_collections
• create_deployments
• update_deployments
• list_deployments
• get_deployments
• create_deployments_predictions
• list_hardware
• get_account
• create_models
• list_models
• get_models
• search_models
• list_models_examples
• create_models_predictions
• get_models_readme
• list_models_versions
• get_models_versions
• create_predictions
• list_predictions
• cancel_predictions
• get_predictions
• create_trainings
• list_trainings
• cancel_trainings
• get_trainings
• get_default_webhooks_secrets

Key Capabilities

• Run and manage AI model predictions through Replicate’s API
• Create and manage model deployments for production use
• Access and search thousands of AI models from the Replicate model library
• Train custom models and manage training workflows

How It Works

The server provides MCP-based access to Replicate’s AI model platform through:

1. Replicate API for running and managing model predictions
2. Model deployment and training workflow management
3. API key-based authentication via REPLICATE_API_TOKEN environment variable
4. Access to thousands of pre-trained models and custom model hosting