Core concepts

Pipelines

A Pipeline is a reusable workflow made from named Activities. It defines the order and conditions under which Activities run, turning individual units of work into a review-driven process.

Design a workflow

Add one or more Activities as Pipeline steps. Steps can run in their listed order, or depend on another step reaching a successful or failed outcome. This lets a Pipeline continue along an appropriate path, such as requesting human approval after a successful review or asking for recovery work after a failure.

Each Pipeline has a name, description, and lifecycle status. Active Pipelines can run; disabled Pipelines remain available for later use; archived Pipelines remain part of history but cannot be changed or started.

Set shared defaults

A Pipeline can define a model and Harness default for its executable Activities. An Activity can override either choice when its work needs a different model or execution path. This keeps common workflow configuration in one place without preventing intentional exceptions.

Human Activities do not use the Pipeline Harness. They wait for a person to submit the required approval, rejection, or response.

Run a Pipeline

Starting a Pipeline creates an independent execution with the supplied work request. The platform freezes the workflow and its effective configuration for that execution, so later edits to the Pipeline or its Activities affect future runs only.

The API selects dependency-ready steps, while a compatible runtime performs repository work in its configured checkout. When a step reaches a terminal outcome, the Pipeline continues along the dependencies that match that outcome. The run completes when no further steps are eligible.

The platform introduction explains the shared execution lifecycle. The API reference defines the complete Pipeline contract.