Replicate

Connects Claude directly to Replicate's cloud API for running AI models on demand. You can fire off predictions using any model in their catalog (Stable Diffusion, LLaMA, Whisper, whatever), create and manage deployments, and poll for results without leaving your conversation. Useful when you need to chain model outputs together, run image generation mid-workflow, or prototype with models you don't want to host yourself. Part of the MCP Armory collection, so it's auto-generated from Replicate's OpenAPI spec and tested against their live endpoints. Handles auth via API token and includes all the standard retry and validation logic you'd expect.