Rivet is an open-source visual programming environment for building AI agents with large language models. Instead of writing prompt chains as tangled code, you assemble them as a node-based graph, run them in the editor, and then execute the same graph directly inside your application. It was built by Ironclad, the digital contracting company, to develop its own Contract AI assistant. Rivet is free to download, with a debugger that follows prompt chains while they run and graphs saved as plain YAML files.
Rivet is made for software engineers, AI engineers and product teams who build LLM features and want to iterate faster than they can in code alone. Teams appreciate that complex agent logic becomes visible, so colleagues who did not write the chain can still follow it. Technical leads use it to compare speed and quality tradeoffs while building AI experiences, and developers working in Node or TypeScript can embed graphs into production services. Because graphs are text files, they fit into the same repository and review process a team already uses for code.
Typical use cases include prototyping a multi-step agent that combines prompts with existing APIs, then promoting it to production without a rewrite. Ironclad used it to build Contract AI, a virtual assistant that answers questions across the contract lifecycle using contract search, workflow and data visualization capabilities. Others use the remote debugger to watch a live chat application and see what each node produced. Teams collaborate by versioning YAML graphs in Git, reviewing changes in pull requests, and trying the example chat application to learn how a graph is developed and debugged.
