Running Workflows

How workflow execution works — from triggering a run to viewing results. Understand execution states, test runs, scheduled execution, and the execution history.

How Workflows Start

A workflow can be triggered in three ways:

Manual (Test Run): Click the Test Run button in the editor header. The workflow executes immediately with no trigger data. Use this during development and debugging.
Schedule: The scheduler fires at the time defined by the cron expression. Scheduled workflows run automatically without any user interaction.
Webhook: An external system sends an HTTP POST to the workflow's webhook URL. The request body becomes the trigger data, accessible in downstream nodes.

Execution States

Each execution progresses through a lifecycle of states:

State	Description
Pending	The execution has been queued but has not started processing yet.
Running	The execution engine is actively processing nodes.
Waiting	The execution is paused at an Approval Gate, waiting for human input.
Completed	All nodes finished successfully.
Completed with Warnings	All nodes finished, but one or more produced warnings.
Failed	One or more nodes encountered an unrecoverable error.
Cancelled	The execution was manually stopped by a user.

Node Execution States

Individual nodes within an execution also have states:

State	Description
Pending	Waiting for upstream nodes to complete.
Running	Currently executing.
Completed	Finished successfully. Output is available.
Failed	Encountered an error. Error details are available.
Skipped	Skipped because a condition branch was not taken.

When viewing the canvas during a live execution, each node highlights with the corresponding status color — blue for running, green for completed, red for failed.

Test Runs

Running a Test

Click Test Run in the editor header. The workflow is queued for immediate execution. You can watch nodes light up in sequence as they execute, with real-time status updates via WebSocket.

Test individual nodes

You can also test a single node in isolation. Click a node, then click the Test button in the configuration panel. This executes only that node with mock or previous input data.

After a test run completes, click any node on the canvas to view its output in the configuration panel. The output tab shows the full JSON response from the node's operation. For failed nodes, the error tab shows the error message and stack trace.

Execution History

The Execution History page lists every execution for your organization. Each entry shows:

Workflow name — click to navigate to the execution detail.
Trigger type — Manual, Scheduled, or Webhook.
Status — with a color-coded badge.
Start time — when the execution began.
Duration — total time in seconds (for completed executions).
Cost — AI token usage and estimated cost.

Filtering & Search

Use the filter bar at the top to narrow results:

Status filter — show only running, completed, failed, or cancelled executions.
Workflow search — type a workflow name to filter.
The result count updates in real time to show filtered vs. total executions.

Bulk Operations

Select multiple executions using the checkboxes, then use the bulk action buttons:

Archive — move selected executions to the archive. They no longer appear in the main list but are retained for audit purposes.
Delete — permanently remove selected executions and their node data. This action cannot be undone.

Crash Recovery

The execution engine saves a checkpoint after each node completes. If the server crashes or restarts during an execution, the engine resumes from the last checkpoint rather than starting over. This ensures that:

Already-completed nodes are not re-executed.
External side effects (sent emails, posted messages) are not duplicated.
The execution continues from where it left off with the same data context.