# Hatchet alternative: node-flow vs Hatchet

> node-flow vs Hatchet — two Postgres-backed engines; node-flow defines workflows as JSON DAGs with a Conductor-compatible API, Hatchet defines them in code.

Source: https://node-flow.dev/docs/alternatives/hatchet

**node-flow and [Hatchet](https://hatchet.run) share a bet: Postgres is
enough.** Both avoid a separate broker and use PostgreSQL for durable state and
queueing. They differ in where the workflow lives. Hatchet workflows — tasks,
DAGs and durable tasks — are declared in code with its Python, TypeScript or Go
SDKs. node-flow workflows are JSON documents the engine reads, and workers in
any language poll for their steps.

## node-flow vs Hatchet at a glance

|                       | node-flow                                                                       | Hatchet                             |
| --------------------- | ------------------------------------------------------------------------------- | ----------------------------------- |
| Workflow is…          | A JSON DAG                                                                      | Code (tasks, DAGs, durable tasks)   |
| Languages             | Any, over HTTP                                                                  | Languages with a Hatchet SDK        |
| Worker connection     | HTTP long-poll                                                                  | Persistent connection to the engine |
| Infrastructure        | PostgreSQL 18                                                                   | Postgres, plus the Hatchet engine   |
| Operators             | `SWITCH`, `FORK_JOIN`, `DO_WHILE`, `SUB_WORKFLOW`, `DYNAMIC`, saga compensation | Expressed in code                   |
| Built-in system tasks | HTTP, SQL, gRPC, jq, sandboxed JS, LLM, MCP, email, webhooks — no worker needed | Tasks run in your workers           |
| Human approvals       | `HUMAN` task with inbox                                                         | Durable events plus your own UI     |
| Conductor API         | Yes                                                                             | No                                  |
| Managed offering      | Self-hosted                                                                     | Hatchet Cloud                       |

## Choose which

**Choose Hatchet** for a code-first, high-throughput task queue with fan-out,
concurrency keys and durable tasks inside services you control.

**Choose node-flow** when the definition itself should be data — reviewable,
diffable, editable in a visual editor, importable from BPMN — and when
non-engineers need to see and act on runs.

## Frequently asked questions

### Are node-flow and Hatchet both built on Postgres?

Yes. Both use PostgreSQL for durable state and queueing. node-flow’s only required dependency is PostgreSQL 18, which also holds its timers, transactional outbox and search.

## Next

* [Self-hosting](/docs/guide/self-hosting)
* [Open-source workflow engines](/docs/alternatives/open-source-workflow-engines)