Skip to main content

Workflows your users can watch.

Rivet runs durable workflows inside Actors that keep their state in SQLite, stream every change to users and run on any cloud.

Choose Rivet for

  • Yes: Durable agent loops
  • Yes: Live progress streamed to users
  • Yes: Workflow state you can query with SQL
  • Yes: Long runs with growing state
  • Yes: Forking agent runs
  • Yes: Any cloud, your VPC or on-prem

Choose Cloudflare Workflows for

  • Yes: Apps already on Cloudflare Workers
  • Yes: Background jobs and data pipelines
  • Yes: Python, until Rivet supports it

Every Rivet workflow runs inside an Actor, in the same process as your backend.

Add durable workflows to your backend

Paste the prompt into your coding agent, or follow the quickstart.

Read the Workflows quickstart

Why Cloudflare Workflows keep state away from users

A workflow runs apart from the Agent or Durable Object your users connect to.

  • No: A separate instance from your AgentProgress is copied to the Agent over RPCYes: A workflow runs inside each ActorOne identity, one SQLite database
  • No: Real-time communication not supportedRelay through an Agent; subscribe() streams only to WorkersYes: Steps broadcast to connected clientsProgress streams live over WebSockets
  • No: State is a status and step outputsNo queries, so app screens keep a copy in D1 or the AgentYes: Query workflow state with SQLActions read the Actor’s SQLite while the workflow runs
  • No: Finished runs are deleted after 7 daysBy default, 30 at most; copy results out to keep themYes: The Actor keeps its dataUntil you destroy it

Why Cloudflare Workflows struggle with long-running agents

Agents run for hours or days, and their context keeps growing.

  • No: 10,000 steps per run by default25,000 at most, so long agent loops must start new instancesYes: Loops prune old history automaticallyHistory stays small however long the loop runs
  • No: 1 MiB per step result, 1 GB per instanceLLM context and tool output go to R2, passed by keyYes: Steps write large data to the Actor’s SQLiteNo fixed storage cap; older data moves to S3
  • No: 30 seconds of CPU per step by default5 minutes at most; heavier steps must be splitYes: No CPU cap per stepSteps run in your Node.js or Bun process
  • No: Every step is billed$0.80 per 100,000 past 500,000 a month, plus CPU and storageYes: Rivet Cloud bills raw hardware and storageNot per step

Why Cloudflare Workflows lock you in

Workflows run only on Cloudflare, inside the Workers runtime.

  • No: Runs only on CloudflareMoving clouds means porting every workflowYes: Runs on any cloud or on-premRivet Cloud, your VPC or your own servers
  • No: Workers runtime, 128 MB of memoryNo native npm packages; large in-memory state has to be splitYes: Runs in your Node.js, Bun or Deno processNo fixed memory cap; native packages work on Node.js
  • No: Jurisdictions, but no network isolationData stays in a region, not in your VPCYes: Runs inside your VPCWith bring your own cloud or self-hosting

Rivet Workflows vs Cloudflare Workflows

FeatureCloudflare WorkflowsRivet
ForkingPartly: Restart from a stepReruns the same instance; no way to copy itYes: Fork an Actor in 20 msWith its SQLite, schedules and workflow progress
SchedulingYes: Cron schedules and sleepsUp to 100 schedules per accountYes: Timers and cron in each ActorSchedules survive sleep, restarts and upgrades
LanguagesYes: TypeScript, JavaScript and PythonPartly: TypeScript and RustBoth SDKs are stable
DebuggingYes: Dashboard and Local ExplorerStep history, retries and errorsYes: Run inspectorEvery step’s inputs, outputs and errors
Durable executionSame on bothSteps, retries, sleeps and rollback
WaitingSame on bothSleeping runs use no CPU
Local developmentSame on bothOne local dev server
Managed cloudSame on bothRivet Cloud, or Cloudflare

Run your next durable workflow on Rivet.

Durable workflows in Actors that keep state in SQLite and stream every change to users, on any cloud.

Read the Workflows quickstartTalk to an engineer