Rivet.dev

Rivet is an open-source infrastructure platform designed for the era of AI agents. It gives developers a composable stack to orchestrate, operate, automate, and deploy agents — all from a single Rust binary that runs in your own backend, VPC, or on-prem. Think of it as a purpose-built operating system for agents: you get durable processes (Actors), secure code execution environments (agentOS), multi-step workflows that survive restarts, and the ability to deploy AI-generated apps (Dynamic Apps) for end users. Rivet is not an agent itself — it's the infrastructure that agents run on.

What makes Rivet stand out is how it bundles everything agents need into one stack. Actors give each agent a durable process with real-time streaming and memory. agentOS hands every agent its own isolated computer with a filesystem and session transcripts. Workflows let you write multi-step operations that keep running even if the server restarts. Dynamic Apps let you deploy the backends your agents build. The whole system is observable through a dashboard and a new MCP integration that lets AI clients like Claude Code and Cursor inspect actors and call actions inline. It works with frameworks like React and Next.js, runtimes like Node.js and Bun, and can be self-hosted on Kubernetes, AWS, GCP, or a plain VM.

Rivet is built for engineering teams that are shipping agent-based products and need production-grade reliability. If you're building a coding agent, a customer support bot, or any system where agents need to persist state, run code safely, and survive failures, Rivet gives you the primitives to do it without reinventing the wheel. It's also a good fit for teams that want to keep their agent infrastructure in their own cloud or on-prem for compliance reasons. The local dev experience matches production, so you can iterate fast and deploy with confidence.