What is Rivet?
Rivet is a visual IDE for designing, debugging, and testing AI prompt chains. Instead of writing prompt code, you build workflows by connecting nodes — LLM calls, conditionals, loops, data transforms, and subgraphs — with a real-time execution preview that shows intermediate results at every step.
Answer-Ready: Rivet is a visual IDE for AI prompt chain development by Ironclad. Build, debug, and test LLM workflows with drag-and-drop nodes, real-time execution preview, and team collaboration. Open-source with 3k+ GitHub stars.
Best for: Teams designing complex AI prompt chains who need visual debugging. Works with: OpenAI, Anthropic, Google, any OpenAI-compatible API. Setup time: Under 5 minutes.
Core Features
1. Visual Node Editor
[User Input] → [Chat Node (Claude)] → [If/Else] → [Output A]
↓
[Output B]Node types:
- Chat: LLM call with model selection
- Text: Static or template text
- If/Else: Conditional branching
- Loop: Iterate over arrays
- Code: Custom JavaScript
- Subgraph: Reusable sub-workflows
- Extract: Parse JSON/regex from LLM output
2. Real-Time Debugging
Click "Run" and see results at every node:
Node 1 (Input): "Summarize this article about AI"
Node 2 (Chat): "This article discusses three key trends..."
Node 3 (If): condition=true → taking branch A
Node 4 (Output): "Summary: This article discusses..."3. Prompt Versioning
- Save graph versions with descriptions
- Compare outputs between versions
- Roll back to any previous version
4. Team Collaboration
- Share graphs as
.rivet-projectfiles - Export as TypeScript for production integration
- Git-friendly project format
5. TypeScript Integration
import { runGraph } from '@ironclad/rivet-node';
const outputs = await runGraph(project, {
graph: 'main',
inputs: { userMessage: 'Hello!' },
context: {},
openAiKey: process.env.OPENAI_API_KEY,
});
console.log(outputs.response);6. Built-In Nodes
| Category | Nodes |
|---|---|
| AI | Chat, Text Completion, Embeddings |
| Logic | If/Else, Switch, Loop, Match |
| Data | JSON Parse, Regex, Split, Join |
| I/O | User Input, HTTP Request, Read File |
| Flow | Subgraph, Delay, Race, Parallel |
Use Cases
| Use Case | How |
|---|---|
| Prompt chaining | Multi-step LLM workflows |
| A/B testing | Compare prompt variants visually |
| Guardrails | Add validation nodes between LLM calls |
| RAG prototyping | Build retrieval + generation pipelines |
| Team review | Share visual graphs for prompt review |
FAQ
Q: Is Rivet free? A: Yes, open-source under MIT license. Desktop app is free.
Q: Can I use it with Claude? A: Yes, Anthropic Claude is fully supported as a chat node provider.
Q: How does it compare to Langflow? A: Rivet focuses on prompt design and debugging with fine-grained node control. Langflow is more about assembling LangChain components.