Rivet - Visual AI Agent Programming Environment

TL;DR

TL;DR: Rivet is an open-source visual IDE for building AI agent graphs with drag-and-drop nodes, with a companion TypeScript library for embedding workflows in your own apps.

Source and Accuracy Notes

⚠️ This section is MANDATORY. All links must be verified from actual source, not guessed.

• Project page: rivet.ironcladapp.com
• Source repository: github.com/Ironclad/rivet — MIT license verified
• Latest release: v1.11.3 (2025-08-08)
• HN launch thread: news.ycombinator.com/item?id=37433218

What Is Rivet?

Rivet is an open-source visual programming environment for building complex AI agents and prompt-chaining workflows. Developed by Ironclad, it comes in two parts:

• Rivet Application — a cross-platform desktop app (macOS, Linux, Windows) with a node-based graph editor
• Rivet Core — a TypeScript library (@ironclad/rivet-core) for embedding graphs in your own applications

The project describes itself as “the IDE for creating complex AI agents and prompt chaining, and embedding it in your application.”

Supported LLMs

Rivet ships with built-in support for:

• OpenAI GPT-3.5 and GPT-4
• Anthropic Claude Instant, Claude 2, and Claude 3 family (Haiku, Sonnet, Opus)
• AssemblyAI LeMUR framework for voice data

Embedding and Vector DB Support

• OpenAI Embeddings
• Pinecone

Additional Integrations

• AssemblyAI Speech-to-Text

Setup Workflow

Step 1: Download Prebuilt Binaries

Head to the releases page or use the direct download links from the README:

# macOS
curl -LO https://github.com/Ironclad/rivet/releases/latest/download/Rivet.dmg

# Linux (AppImage)
curl -LO https://github.com/Ironclad/rivet/releases/latest/download/Rivet.AppImage

# Windows
curl -LO https://github.com/Ironclad/rivet/releases/latest/download/Rivet-Setup.exe

Step 2: Install and Launch

Mount the DMG or run the AppImage, then launch the Rivet application. No account or API key is required for the core editor.

Step 3: Connect an LLM Provider

Open Settings → Providers and add your API credentials. Rivet supports OpenAI and Anthropic out of the box. Enter your API key for the provider you want to use.

Step 4: Create Your First Graph

1. Click New Project in the sidebar
2. Drag an LLM Node onto the canvas from the node library
3. Connect an input node and configure the model (e.g., GPT-4)
4. Add a Text Prompt node and write your prompt
5. Hit Run — the output appears in the debug panel

Deeper Analysis

Visual vs. Code-First Approaches

Rivet targets developers who want to prototype agent workflows visually before embedding them. The node-graph approach makes it easy to see the full flow — which models are called, in what order, with what context. The TypeScript library then lets you ship the same graph without the visual editor.

Comparison with LangChain / LlamaIndex

LangChain and LlamaIndex are code-first frameworks. Rivet is closer to a flow-based editor that sits above those layers. If you prefer wiring nodes visually, Rivet is faster to iterate with; if you want maximum flexibility in code, a framework is more appropriate.

Embedding Graphs in Your App

Rivet Core lets you run graphs programmatically:

npm install @ironclad/rivet-core

import { runGraph } from "@ironclad/rivet-core";

const result = await runGraph(graph, {
  provider: "openai",
  model: "gpt-4",
  input: { query: "Your question here" },
});

Practical Evaluation Checklist

• [x] Cross-platform desktop app (macOS, Linux, Windows)
• [x] MIT license — source freely available
• [x] No mandatory account or cloud service
• [x] Node-based visual editor for agent graphs
• [x] TypeScript library for embedding in apps
• [x] Prebuilt integrations for GPT-4 and Claude family
• [x] Prebuilt integrations for embedding/vector DB (OpenAI, Pinecone)
• [ ] Self-hosted inference option (relies on external API keys)
• [ ] Built-in RAG pipeline (requires manual wiring)

Security Notes

• API keys are stored locally and never sent to Rivet’s servers
• Graphs run entirely on your machine or in your own infrastructure
• The TypeScript library calls LLM providers directly using your credentials
• No telemetry or tracking in the open-source release (verify in source before using in sensitive environments)

FAQ

Q: Is Rivet free to use? A: Yes. Rivet is open-source under the MIT license. The desktop app, TypeScript library, and all integrations are free.

Q: Can I run it offline? A: The Rivet editor itself runs offline. However, calling LLM providers requires an internet connection and API credentials. Self-hosted model support depends on what providers you configure.

Q: How does it compare to building agents with code? A: Rivet’s visual graph editor makes it easier to see and debug complex prompt chains. The TypeScript library gives you the same power as code-first approaches, just with a visual prototype layer on top.

Q: Does it support local models? A: Rivet supports any LLM with an OpenAI-compatible API endpoint. You can point it at a local server like Ollama or LM Studio as long as it exposes a compatible API.

Conclusion

Rivet fills a gap between no-code agent builders and code-first frameworks. Its visual editor makes it easy to prototype complex prompt chains, and the TypeScript library means you can ship the result without a visual runtime dependency. If you want a visual way to build and embed AI agent workflows, Rivet is worth a look.