Loading tool details…
Master Rivet with our step-by-step tutorial, detailed feature walkthrough, and expert tips.
Download the Rivet desktop app from rivet.ironcladapp.com (macOS, Windows, Linux) or install the VS Code extension Create a new project — the file is a YAML you can commit to Git Drag in a Chat node, connect it to an Output node, and configure your OpenAI or Anthropic API key Install @ironclad/rivet
node in your Node.js app: npm install @ironclad/rivet
node Load the YAML project and execute a graph programmatically, streaming results back via callbacks Attach the remote debugger to your running service to observe execution live
💡 Quick Start: Follow these 3 steps in order to get up and running with Rivet quickly.
Explore the key features that make Rivet powerful for ai agent frameworks workflows.
Drag-and-drop nodes for LLM calls, code execution, control flow, and data transformation. The canvas maintains readability at high node counts through subgraph composition, comments, and grouping.
Every project is a plain YAML file. Diffing prompt changes in a PR looks like reviewing any other code file — a critical advantage over tools that lock graphs behind proprietary formats or cloud databases.
Connect the desktop editor to a running Node.js process over WebSocket. Each node lights up as it executes, showing inputs, outputs, LLM responses, and errors. This is Rivet's standout differentiator vs. text-based frameworks.
npm package that loads a Rivet project and executes graphs programmatically. Type-safe input/output contracts, streaming callbacks, and hooks for logging and metrics.
Built-in support for OpenAI, Anthropic, Google Gemini, AWS Bedrock, plus any OpenAI-compatible endpoint (Ollama, LM Studio, vLLM, Together, custom gateways).
Yes — Rivet is MIT-licensed open source with no paid tier. There's no hosted service; you run the desktop editor locally and deploy the Node runtime yourself.
No. Rivet's runtime is TypeScript-only via the @ironclad/rivet-node npm package. Python teams should look at LangGraph, Langflow, or CrewAI instead.
The @ironclad/rivet-node runtime opens a WebSocket server; the desktop editor connects to it. As your app executes a graph, each node lights up in the editor with its inputs, outputs, and any errors. This is Rivet's biggest differentiator vs. text-based agent frameworks.
Yes — that's the intended deployment. The desktop editor stays on your machine; the runtime library ships inside your Node.js service. There's no cloud dependency.
LangGraph if your stack is Python and you want the largest ecosystem. Rivet if your stack is TypeScript, you want a visual editing surface, and the remote debugger sounds valuable. For simple linear chains, both are overkill.
Now that you know how to use Rivet, it's time to put this knowledge into practice.
Sign up and follow the tutorial steps
Check pros, cons, and user feedback
See how it stacks against alternatives
Follow our tutorial and master this powerful ai agent frameworks tool in minutes.
Tutorial updated March 2026