Rivet.dev

Open-source orchestrator that runs agents and sandboxes as durable Actors

Advanced API
Screenshot of Rivet.dev, Open-source orchestrator that runs agents and sandboxes as durable Actors

What is Rivet.dev?

Rivet.dev is a coding tool. Rivet is an open-source, self-hostable orchestrator that runs agents, workflows and sandboxes as durable Actors with built-in SQLite and filesystem storage, fast cold starts, and deployment options from local to managed cloud and BYOC.

Rivet is an open-source orchestrator built for the workloads that AI agents generate and depend on. Its central idea is the Actor: a small, durable process that carries its own state and can stand in for an agent, a multi-step workflow, a sandbox for untrusted code, or a per-user backend. Each Actor receives a SQLite database and a POSIX filesystem, with storage tiered to S3, so state survives restarts and idle instances cost nothing while parked. The target audience is engineering teams building agent products: coding agents, agent-driven app builders, company-specific agents running inside a private cloud, personal assistants, and realtime collaborative apps. Actors live inside worker processes a team already operates rather than in separate containers or VMs, which is how the site justifies its density claims of roughly 72 KB per running Actor and cold starts around 12 ms, measured on FoundationDB with the Rust SDK. An experimental feature is involved in the cold-start figure. In practice, a developer installs the rivetkit package, defines Actors in TypeScript, and connects to a control plane that handles scheduling, routing and observability. Typed client SDKs, React hooks, HTTP and WebSocket addressing, actor-to-actor calls, cron scheduling and OpenTelemetry output come built in. Workers run on plain Node.js or Bun rather than a custom runtime, and Rust is supported too. The same binary and APIs run on a laptop and in production. Deployment is flexible. The Apache 2.0 control plane ships as a single Rust binary that can be self-hosted on Kubernetes, Docker Compose, virtual machines or platforms such as Railway and Render, backed by Postgres or FoundationDB. Rivet Cloud offers a managed option, and a bring-your-own-cloud mode has Rivet operate the control plane inside a customer VPC, including air-gapped setups. Among alternatives, Rivet positions itself against Kubernetes for stateful workloads and as a self-hostable answer to Cloudflare Durable Objects, citing the lack of memory and SQLite size caps. Teams wanting a simple hosted function runner will find it more infrastructure-oriented than that.

How do you use Rivet.dev?

  1. 1Read the docs overview
    Start with the main documentation to understand Actors and how they map to agents, workflows and sandboxes.
    Rivet.dev — Read the docs overview
  2. 2Install the rivetkit package
    Add rivetkit to a Node.js or Bun project with npm and follow the Actors quickstart to define a first Actor with typed state.
    Rivet.dev — Install the rivetkit package
  3. 3Build an agent or workflow Actor
    Use the agents or workflows guides to give an agent a durable process with memory, or to create multi-step operations that replay after failure.
    Rivet.dev — Build an agent or workflow Actor
  4. 4Run it locally
    Develop against the local open-source control plane, which uses the same APIs and behavior as production.
  5. 5Choose a deployment
    Self-host the Rust control plane, deploy to Rivet Cloud from the dashboard, or review the BYOC docs for running inside your own VPC.
    Rivet.dev — Choose a deployment

Pros and cons

Pros

  • Open source under Apache 2.0, and the managed service runs the same control plane you can self-hostAI
  • Each Actor includes SQLite and a POSIX filesystem tiered to S3, giving durable state without extra setupAI
  • Runs on plain Node.js or Bun with typed SDKs, so existing packages, tooling and tests keep workingAI
  • Flexible deployment: local, self-hosted single Rust binary, Rivet Cloud, or BYOC inside your own VPCAI
  • Built-in scheduling, WebSocket and HTTP addressing, multi-region routing and OpenTelemetry reduce extra infrastructureAI

Cons

  • Requires adopting the Actor programming model, which is a real learning curve for teams used to stateless servicesAI
  • Headline cold-start figure relies on an experimental feature not yet enabled by defaultAI
  • Performance figures are vendor-measured on FoundationDB with the Rust SDK and may differ in other setupsAI
  • Dynamic Apps is labelled Preview, so that part of the platform is not yet fully matureAI
  • The homepage lists no pricing, so cost for Rivet Cloud or BYOC cannot be judged without contacting the teamAI

How much does Rivet.dev cost?

Support

Documentation covers Actors, agents, workflows, agentOS and deployment, a changelog is published, and enterprise support is available through a talk-to-an-engineer option.

Learn more

Integrations

Works with TypeScript, Node.js, Bun, Rust, React and Effect. Storage backends include Postgres, FoundationDB and S3. Deploys to Kubernetes, Docker Compose, VMs, AWS ECS, Railway and Render.

Learn more

Features

Durable Actors with SQLite and POSIX filesystem storage, hibernation and fast wake, fault tolerance, multi-region routing, HTTP and WebSocket networking, typed SDKs, React hooks, cron scheduling, actor-to-actor calls, OpenTelemetry, and isolated execution of generated code.

Learn more

Frequently asked questions about Rivet.dev

  • Does Rivet.dev have an API?
    Yes, Rivet.dev offers an API.
  • How do you use Rivet.dev?
    The walkthrough on this page covers 5 steps: 1. Read the docs overview 2. Install the rivetkit package 3. Build an agent or workflow Actor 4. Run it locally 5. Choose a deployment.
  • What platforms does Rivet.dev support?
    Rivet.dev is available on Web App, MacOS, Windows and Linux.
  • What does Rivet.dev integrate with?
    Works with TypeScript, Node.js, Bun, Rust, React and Effect. Storage backends include Postgres, FoundationDB and S3. Deploys to Kubernetes, Docker Compose, VMs, AWS ECS, Railway and Render.
  • What are the limitations of Rivet.dev?
    Requires adopting the Actor programming model, which is a real learning curve for teams used to stateless services. Headline cold-start figure relies on an experimental feature not yet enabled by default. Performance figures are vendor-measured on FoundationDB with the Rust SDK and may differ in other setups.

Status

StatusActive
Views0
Outbound clicks0
Added10/8/2026

Platforms

Web AppMacOSWindowsLinux

Pricing

Freemium