Replicate MCP Connector for Claude

Connect your conversational assistant directly to the Replicate ecosystem. This integration grants your AI the ability to interact programmatically with a vast library of open-source machine learning models without running them on your local hardware. From orchestrating complex image generations to spinning up specialized language models, you can command AI workflows directly from your chat.

What you can do

• Execute Predictions — Command the assistant to execute specific model versions on your behalf (create_prediction) by supplying a payload of variables. Monitor long-running processes by retrieving outputs and execution status reliably (get_prediction) or cancel them at will (cancel_prediction).
• Discover Models — Instruct the AI to intelligently scan the Replicate platform for models matching a specific use case using search_models. You can also explore trending and categorized models by leveraging the list_collections action.
• Analyze Model Metadata — Whenever you discover a new model, query its precise owner and name (get_model) to extract the exact schema and parameter requirements necessary for a successful execution. You can also view a log of your own executed tasks (list_predictions).

How it works

1. Install the Replicate platform extension module in your MCP.
2. Provide your personal Replicate API Token, extracted from your Replicate Account Settings panel. Deposit it securely into the configuration variables below.
3. Prompt your assistant organically: "Search for a high-quality video generation model, evaluate its parameter schema, and start generating a clip using the prompt 'a cat walking on Mars'."

Who is this for?

• AI Developers & Researchers — Explore and test novel open-source algorithms by generating quick predictions without modifying Python notebook code.
• Content Creators — Execute specialized image, audio, or video generation tasks by directly delegating workflows to your conversational assistant.
• Builders — Mix and match the output of various specialized models intelligently using natural language instructions.