Rivet
Open-source visual IDE for building complex AI agents and prompt chains — drag, drop, debug, embed.
About
Rivet is a desktop application for visually building AI agent graphs: you drag nodes for prompts, chat calls, vector retrieval, control flow, and subgraphs onto a canvas, wire them together, and run the graph live with a debugger that shows every intermediate value. It was built by Ironclad (the contract-management company) for its own AI work and open-sourced under MIT.
The split is what makes it practical: the Rivet Application is the visual builder, and Rivet Core is a TypeScript library you npm-install to execute exported graphs inside your own product. The README lists support for OpenAI, Anthropic Claude, and AssemblyAI models plus embedding/vector-database nodes. The caveats are age and focus: the model list in the README skews older (GPT-3.5/4 era), commits slowed after August 2026, and Ironclad's core business is legal tech, so Rivet is a side project that could drift. For teams that think in flowcharts, it is still the most approachable visual agent builder in the open-source set.
What it can do
- Visual drag-and-drop canvas for agent and prompt-chain graphs
- Live debugger showing every node's intermediate values
- Rivet Core TypeScript library to run exported graphs in your own app
- Nodes for chat, prompts, vector retrieval, control flow, and subgraphs
- Model support for OpenAI, Anthropic Claude, and AssemblyAI
- Prebuilt binaries plus full source for running from scratch
Setup
- Download the Rivet desktop app from the GitHub releases page (macOS, Windows, Linux).
- Build your agent graph on the canvas: prompts, chat, retrieval, and logic nodes.
- Add your model provider keys in the app settings and run the graph with the live debugger.
- Export the graph and execute it in your app with Rivet Core:
npm install @ironclad/rivet-core. - Iterate: edit the graph visually, re-export, redeploy.
Prerequisites
A desktop OS for the builder app (or Node.js + TypeScript for Rivet Core only); model provider API keys.
Use cases
Works with
Trust score — 7/10
Scored 7/10 (verified tier, metrics-v1). Evidence: 4,708 GitHub stars (GitHub API), MIT, pushed Aug 26 2026; the README corroborates the app/core split, the node model, and the provider list; official docs verify the install and embed flow. Strengths: the most approachable visual agent builder in open source, genuinely embeddable via Rivet Core. Docked for: (1) README model list skews older (GPT-3.5/4 era); (2) commits slowed after Aug 2026; (3) Ironclad's core business is legal tech, so this is a side project that could drift. License: MIT.
★ 4,708 GitHub stars
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 7.0/10
- Maintenance · recency, releases, issue load
- 7.0/10
- Openness · license + docs
- 10.0/10
- Manual review · sanity read + red-flag check
- 7.0/10
- · Adoption 7: starScore from 4,708 stars on Ironclad/rivet
- · Maintenance 7: pushed 2026-08-26 (recency 8); releases est 6; 107 open issues on 4.7k stars (issues 5)
- · Openness 10: MIT (0.7*10) + official docs verified at rivet.ironcladapp.com (0.3*10)
- · Manual 7: README corroborates the app/core split, node model, and provider list; docked for dated model list, slowed commits after Aug 2026, Ironclad side-project drift risk
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.