AI Agent News/stories/rivet-releases-open-source-actor-orchestrator-for-agent-workloads-built-on-s3.md

Infrastructure

Rivet releases open-source actor orchestrator for agent workloads built on S3

Rivet is an open-source orchestrator that runs each agent, session, user or tenant as a long-lived actor with its own SQLite database, realtime connections and scheduling. The project says its actors start in milliseconds and store durable state on S3 at SSD-level latency.

Source: github.com/rivet-dev/rivet

Rivet describes itself as an orchestrator for agentic workloads built on the actor model. Each Rivet Actor is a long-lived process with in-memory state, a SQLite database, WebSockets and scheduling, addressed by key. Developers can run actors on plain Node.js, Bun or Rust.

Rivet says its actors share a node process rather than each booting a sandbox, which gives a 4.99 ms p99 cold start and 72 KB of memory per actor. Its published comparison lists a sandbox at 1.70 s and 512 MiB. Idle actors hibernate and wake on the next request, and SQLite writes backed by S3 take 1.07 ms at p99.

An object storage bucket is the only required dependency, and Rivet works with Amazon S3, Google Cloud Storage, Azure Blob Storage and S3-compatible stores such as MinIO and Cloudflare R2. Rivet Cloud, self-hosting and a bring-your-own-cloud option are available. The project is licensed under Apache 2.0.

Why it matters if you build agents

Teams that keep one process per agent or session can use actors with durable state in place of a sandbox for many workloads, though the sandbox comparison comes from Rivet's own benchmark. Code that an agent writes and runs should go to agentOS, which Rivet says is built for that purpose. Teams using Cloudflare Durable Objects now have an open-source option they can self-host.

Get stories like this every weekday

One short email with the day's agent tooling news.

Subscribe