Flows

Creating and editing flows

Build workflows that process data, call AI models, and take actions with Flows. For an overview of the concepts, read What are Flows?.

Before you begin

Sign in to the Runtype dashboard before you create or edit a flow.

Create a flow

To create a flow, follow these steps:

  1. On the Flows page, click Create Flow.
  2. In the Create New Flow panel, choose Start from scratch, Use an example, or Import flow.
  3. Complete the selected creation path. The flow editor opens.
  4. To rename the flow, click the flow name in the header.

You can also create a flow from a Product when you add a Capability. The same flow editor opens.

Add steps

Steps are the units of work in a flow. They can call an AI model, fetch data, make API requests, or run code. For the available step types, see Flow step types overview.

To add a step, follow these steps:

  1. Click the + control where you want to insert a step.
  2. In the step picker, browse or search for a step type.
  3. Select a step type. The step card opens.

If the search leaves one result, press Enter to add it.

Pass data between steps

Flows run top-level steps in the order shown. To pass data between steps, reference an earlier step’s output variable with a template such as {{summary_result}} or {{ticket_analysis.category}}.

You can also reference system variables. For example, use {{_record.metadata.field}} to read a Record metadata field.

Reorder steps

Drag a step in the step list and drop it at its new position. The flow runs steps in the order shown.

Enable or disable steps

To skip a step without deleting it, click Enable step or Disable step on its step row. The disabled step is skipped during execution.

Edit a step

To edit a step, click it in the step list or editor. Update its configuration in the step card, then click Save in the toolbar.

Delete a step

Before you delete a step, update references to its outputVariable in other steps. Then, click Delete step on the step row.

Test a flow

To test a flow, follow these steps:

  1. In the toolbar, click Run flow. The Run Flow sheet opens.
  2. In the Run Flow sheet, choose Test.
  3. In the Test view, enter the input values for the flow.
  4. Click Run Flow. The execution results appear in the sheet.

Review the execution results for each step. They include the step’s input, output, status, and execution time when available.

You can test a flow before saving it. The test uses the current unsaved steps.

Save and publish a flow

The toolbar shows Create for a new flow and Save for an existing flow. Click Create or Save to save your changes.

To publish an existing flow, follow these steps:

  1. Click the chevron next to Save.
  2. If you have unsaved changes, click Save & Publish. Otherwise, click Publish.

Publishing records an immutable snapshot of the flow. Products that use the flow run its saved definition, so saving is what changes their behavior. For more information, see Flow versioning and publishing.

For an existing flow, use the same dropdown to open View versions, Conversations, or Activity. The Activity view shows who changed the flow and when. For more information, see Audit trail.

Published versions are locked. Edit the flow to create a new draft.

Generate a flow with AI

Use the AI flow generator to create steps from a description. In an empty flow, describe what you want to build in the generator on the editor page. For a flow that already has steps, click Generate flow in the toolbar and describe what you want to add or modify.

If the flow already has steps, choose whether to replace them or add the generated steps to the flow. Review the generated steps before you apply them.

Error handling

Configure how each step responds to errors with the error-handling control on its step card. Choose one of these modes:

  • Continue on error: Skip the failed step and continue the flow.
  • Stop on error: Stop the flow when the step fails.
  • Use fallbacks: Try configured alternatives before the step reports failure.
Prompt steps behave differently

Run Task and Run Agent steps continue to the next step after an error. Their error-handling setting controls the fallback chain, not whether the flow stops. An error caused by exhausted platform credit or an unavailable required saved MCP connection still stops the flow. Required MCP connection failures bypass prompt fallbacks, including fixed messages. For instructions on stopping a flow after another prompt failure, see Error handling for prompt steps.

For prompt steps, the fallback list supports these types:

  • Retry: Run the same model again.
  • Different model: Switch to a backup model with optional temperature and max tokens overrides.
  • Fixed message: Return a pre-written reply without calling a model. Use it last to return a reply if earlier fallbacks fail.

Fallbacks run in the order you list them until one succeeds.

For prompt steps, choose when the fallback chain runs:

  • The step errors: Run fallbacks when a provider error, timeout, or thrown failure occurs.
  • The reply is empty: Run fallbacks when the model completes without visible text. A reasoning model can use its full token budget without returning visible text.

You can enable both triggers. Use a backup model when a provider or model is unavailable. Add a fixed message as the final fallback to return a reply when other fallbacks fail.

Next steps

Use these pages to continue working with flows: