Docs

From zero to a durable workflow in ten minutes.

Install the SDK, write one function, run the local dev server and watch your first step retry itself. Everything below runs on the free plan.

01 · Quickstart

Your first workflow, in your language.

Workflows live in your repo, deploy with your app and run on your cloud or ours. TypeScript, Python and Go SDKs share one runtime and one dashboard.

workflows/onboarding.ts
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17

STEP 01
Install the SDK
$ npm i @orrindel/sdk
or pip install orrindel · go get orrindel.dev/sdk
STEP 02
Run the dev server
$ npx orrindel dev
Local dashboard on :8288, with replay and traces
STEP 03
Deploy with your app
$ git push
Workflows register on boot. No separate workers.

02 · Guides

Guides for the workflows people ask about most.

Each guide is a working example with tests, written by the engineers who built the feature.

Guide · 12 min

Build a refund agent with approvals

Classify tickets, check charges and pause for a lead before refunds over $200.

Guide · 8 min

Fan out over a dataset

Process 10,000 rows in parallel with a concurrency limit per provider.

Guide · 6 min

Schedule work with cron and sleep

Weekly reports, trial reminders and follow-ups without a cron box.

Guide · 10 min

Replay failed runs after a fix

Deploy a fix, select the failed runs and resume them from the broken step.

Guide · 7 min

Give an agent typed tools

Turn functions into tools the model can call, with approval on the risky ones.

Guide · 9 min

Trace and budget LLM spend

Tokens and cost per step, with daily budgets and alerts in chat.

03 · Reference

Six calls cover almost every workflow you’ll write.

No new language and no YAML. You write normal functions and wrap the parts that talk to the outside world.

step.run()01

Run it once, keep the result

Wrap any call. The result is saved, so a retry or replay reads it back instead of calling again.

const acct = await step.run("fetch-account",
  () => crm.get(event.customerId))

step.sleep()02

Wait days without a server

Pause a run for minutes or months. Nothing polls and nothing is billed while it sleeps.

await step.sleep("trial-ends", "14d")
await step.run("send-offer", sendOffer)

step.waitForEvent()03

Pause until something happens

Hold a run until a matching event arrives, like a paid invoice or a signed form, with a timeout.

const paid = await step.waitForEvent("paid",
  { event: "invoice.paid", timeout: "7d" })

retries04

Retry with backoff, by default

Failed steps retry on their own: 3 times, with exponential backoff. Errors you mark final stop at once.

step.run("charge", charge, {
  retries: 3, backoff: "exponential" })

concurrency05

Limit how much runs at once

Cap parallel runs per key, like one per customer, so a burst from one account never starves the rest.

orrindel.workflow({ id: "sync",
  concurrency: { limit: 2, key: "event.accountId" } })

step.invoke()06

Call another workflow

Compose workflows. The parent waits for the child's result, and both show up in one trace.

const lead = await step.invoke("enrich",
  { function: enrichLead, data: { email } })

Questions

Things people ask on day one.

No. Workflows are ordinary functions in your codebase. You wrap the calls you want to make durable in step.run and deploy as usual. Most teams move one flow in an afternoon.

Each step.run, sleep and waitForEvent is one step. Retries of a failed step are free, sleeping costs nothing, and steps restored from memo during a replay are never billed.

Step inputs and outputs are stored in the region you pick (us-east, eu-west or ap-south) and encrypted with per-workspace keys. Enterprise plans can run the whole runtime in your own cloud account.

A queue delivers messages. Orrindel keeps the state of the whole run: which steps finished, what they returned, what is waiting and why. That is what makes retries, waits of several days and replays possible without extra code.

Yes, and you decide how carefully. Mark a tool as needing approval and the run pauses for a person. Mark it idempotent and a retry can never run it twice.

Runs pause and resume where they stopped. Nothing is lost and nothing runs twice. Our status history and incident write-ups are public.

Help

Stuck? Ask the engineers who built it.

Community chat is answered by the core team on weekdays. Paid plans get a shared channel with a named engineer.

Start building →

Book a demo →