When to use it
You add steps whenever you need to model an action in the workflow. A journey that authenticates, creates a resource, and verifies the resource has three steps. Each step is distinct: it has its own request, its own value configuration, and its own assertions.Key concepts
HTTP request steps are the only step type currently available. An HTTP request step is built from a request template that defines the request shape (method, path, headers, body). You configure how the step gets its values and what assertions it must satisfy. Step values — each step can receive values from many sources: run inputs, declared variables, derived values, previous step outputs, hard-coded values, environment values, secret references, or “me” values. See Values and variables. Step outputs — a step can expose parts of its response as named outputs that downstream steps consume. A common pattern: step 1 captures a token from a response, step 2 uses that token in its authorization header. Assertions — each step can carry assertions (status code, body field checks, header checks, and so on) that determine whether that step passed or failed. See Assertions. Planned step types — the following step types are not yet available but are on the roadmap:The step types below are planned and not yet available.
- Connectors — integrate with external systems.
- Sub-journeys — embed one journey inside another.
- Scripts — run custom logic between steps.
- Set variables — assign or compute a variable mid-journey.
- Notes — attach documentation or context to a position in the journey.
- Manual checkpoints — require a human to confirm success before proceeding (used in Runbooks).
How it works
- In Build mode, you add a step to the journey canvas by selecting a template from the request template library.
- You configure the step’s value sources — which fields come from run inputs, which from previous step outputs, which are hard-coded, and so on.
- You add assertions to define what “success” looks like for that step.
- When the journey runs, steps execute in order. Each step’s response is evaluated against its assertions and its outputs are made available to subsequent steps.

