Skip to content

Static Workflows

A static workflow has its execution order fixed in advance. You describe the steps and their dependencies as a graph, and the executor computes the order once, at initialisation, using a topological sort.

How Static Workflows Run

  • Entry nodes (no parents) receive the workflow's initial input.
  • A node with one parent receives that parent's output.
  • A node with several parents (fan-in) receives a list of the parents' outputs.
  • One node can feed several others (fan-out).
  • Nodes are executed one at a time, in topological order.
  • The result is a map of nodeID → output for every node that ran.

Because every node in the graph runs and the order never changes, static workflows are predictable, easy to test and easy to audit.

Good Fits

  • Document or data processing pipelines with a fixed sequence of stages.
  • Compliance or approval processes where every check must always run.
  • Composite steps reused inside other workflows (as workflow nodes).

Building One

Follow the authoring guide, choosing static at step 5:

  1. Define the header
  2. Plan and write your nodes
  3. Define the graph

Then onboard the workflow and call it from an agent with the DSL Manager. The exact JSON fields are in the Schema Reference.

A complete, deployable static example — a five-agent, linear pipeline — is the Legal Contract Review Pipeline.