Run Models

This gives Claude direct access to Replicate's prediction API for running thousands of AI models, from image generators to LLMs. You can fire off predictions, poll for results, or set up webhooks for async completion. The key workflow is straightforward: pick a model, fetch its schema to see what inputs it expects, create a prediction, and handle the output URLs. What's nice is the flexibility around synchronous vs async execution, plus the ability to chain models by passing output URLs directly as inputs to the next prediction. One gotcha: output URLs expire after an hour, so grab what you need immediately. Most useful when you're building multi-model pipelines or need Claude to orchestrate complex generation tasks without writing integration code yourself.

Install to Claude Code

npx -y skills add replicate/skills --skill run-models --agent claude-code