A graph you can reason about.
Express branches, parallel work, loops, and human approvals in inspectable JSON.
Explore the workflow DSL Source available. Built on Postgres. Complex workflows.
Complex workflows.
Beautifully orchestrated.
A self-hosted workflow orchestration engine that runs on Postgres alone.
Durable workflows, workers in any language, and just one dependency.
Source available Self-hostable Postgres-powered
fulfil_orderv1
Interactive demoValidate an order, charge and reserve in parallel, then bring it all together.Illustrative execution · no server required
YOUR LOGIC. YOUR LANGUAGE.
TypeScriptPythonGoJavaRustHTTP
One worker protocol. Endless possibilities.
01 / THE DEVELOPER EXPERIENCEYou write the work.
You write the work.
Node Flow connects it.
From a simple sequence to a sprawling graph, the same clear model takes you from definition to execution.
Built for the unexpected.
Durable state, task leases, and configurable retries keep distributed work moving.
Understand executionEvery run tells a story.
Follow execution history, inspect task outputs, and retry from the point that matters.
Take control of a run02 / RADICALLY SIMPLE INFRASTRUCTUREBig workflows.
Big workflows.
Small footprint.
Queues, timers, workflow state, and the outbox. All backed by Postgres. One transactional foundation for the whole system.
One required infrastructure dependency Pure, deterministic decision engine Scale API, decider, and poller independently
Go inside the architecture PostgresONE SOURCE
OF TRUTH
Workflow stateTask queuesDurable timersTransactional outboxOF TRUTH
03 / A SHORT PATH TO YOUR FIRST RUNLess setup.
Less setup.
More shipping.
1
Start your local stack
Bring up Postgres, the server, and the dashboard with Docker Compose.
2
Define the workflow
Describe your tasks in JSON. Register the definition with the API.
3
Follow the complete quickstart Connect a worker. Watch it flow.
Lease a task, execute your logic, and report the result.
YOUR TERMINAL
# From a checkout of the repository
$
docker compose -f docker/docker-compose.yml upPostgresready
Node Flow server:3000
Dashboard:3100
$
Example output · Docker Compose starts three servicesTHE NEXT STEP IS YOURSBuilt for builders.
All documentation Built for builders.
Documented for humans.
Your first workflow
From a local stack to a completed run. Follow the code, step by step.
QuickstartLearn the building blocks
Branches, parallel tasks, waits, retries, and the complete workflow DSL.
Workflow guideExplore the API
Endpoints, schemas, and executable requests. Powered by Scalar.
API referenceMake it production-ready
Deploy, monitor, secure, and operate your own workflow infrastructure.
Self-hosting04 / HOW IT COMPARESA Conductor, Temporal and
All comparisons A Conductor, Temporal and
Trigger.dev alternative.
Netflix Conductor alternative
Same JSON model and a compatible API — without Redis or Elasticsearch.
node-flow vs Netflix ConductorOrkes Conductor alternative
RBAC, SSO and AI tasks, self-hosted and free to run commercially.
node-flow vs Orkes ConductorTemporal alternative
Workflows as readable data, workers without determinism rules.
node-flow vs TemporalTrigger.dev alternative
Workers in any language, running on compute you already have.
node-flow vs Trigger.devInngest alternative
Pull-based workers and server-side flow control on Postgres.
node-flow vs Inngestiii alternative
A focused orchestrator for the services you already run.
node-flow vs iiiAWS Step Functions alternative
JSON workflows anywhere, with no per-transition bill.
node-flow vs AWS Step FunctionsOpen-source workflow engines
Conductor OSS, Temporal, Airflow, Hatchet and more, side by side.
node-flow vs Open-source engines05 / QUESTIONSFrequently asked.
Frequently asked.
Plainly answered.
What is Node Flow?
Node Flow (node-flow) is a self-hosted workflow orchestration engine. You describe a process as a declarative JSON DAG, workers in any language poll for the steps they own, and the engine makes every step run, retry, time out and compensate as defined — across restarts, deploys and failures. PostgreSQL 18 is its only infrastructure dependency.
Is Node Flow an alternative to Netflix Conductor and Orkes?
Yes. Node Flow follows Conductor’s workflow model and serves a Conductor-compatible REST API at /conductor/api, so supported Conductor SDKs and workers connect with a base-URL change. It replaces Conductor’s Redis, persistence and Elasticsearch stack with a single Postgres database.
How is Node Flow different from Temporal or Trigger.dev?
Temporal and Trigger.dev make code durable: the workflow is a function in an SDK. Node Flow makes the workflow a JSON document the engine runs, and workers in any language implement individual steps. That suits processes that cross teams, services and languages, and that operators need to see and act on.
What infrastructure does Node Flow need?
Only PostgreSQL 18. Workflow state, task queues, timers, the transactional outbox and search all live in Postgres. There is no Redis, Cassandra, Elasticsearch or Kafka to run.
Which languages can workers be written in?
Any. The worker protocol is plain HTTP — lease a task, heartbeat, report a result. There is a TypeScript SDK and generated clients for Python, Go, Java and TypeScript.
Is Node Flow open source and free?
Node Flow is free to run, including commercially and inside products you sell. It is source available rather than OSI open source: the code is public and may be redistributed unmodified, but modified versions may not be distributed.
YOUR INFRASTRUCTURE. YOUR RULES.
Let it flow.
Your next workflow starts with a single step.
Start building with Node Flow
BEHIND THE FLOW
Meet the builderdwaipayan.in (opens in a new tab)Dwaipayan Sardar
Built with curiosity. Shared with the world.