Skip to main content
Workflows compose agents, teams, and callables into multi-step pipelines. Six composable step types let you build everything from linear chains to complex branching logic. 
from definable.agent import Agent
from definable.agent.workflow import Workflow, Step

researcher = Agent(model="gpt-4o", instructions="You are a research specialist.")
writer = Agent(model="gpt-4o", instructions="You are a technical writer.")

workflow = Workflow(
    name="research-pipeline",
    steps=[
        Step(name="researcher", agent=researcher),
        Step(name="writer", agent=writer),
    ],
)

result = await workflow.arun("Write about quantum computing")
print(result.content)
Each step receives the previous step’s output as context automatically.

​
Step Types

TypeBehaviorUse case
StepExecutes a single agent, team, or callableBasic building block
StepsExecutes steps sequentially, chaining contextLinear pipelines
ParallelExecutes steps concurrentlyIndependent analyses
LoopRepeats until end condition or max iterationsIterative refinement
ConditionIf/else branchingQuality gates
RouterN-way dynamic routingClassification and dispatch

​
Guides

Build Workflows

Create workflows with composable step types.

Run Workflows

Execute workflows and access step results.

​
Workflow Patterns

Sequential

Linear step chains with context passing.

Parallel

Concurrent execution with result combination.

Conditional

If/else branching based on step output.

Loop

Iterative refinement with end conditions.

Router

Dynamic N-way routing by selector function.

​
Resources