Monitor workflow runs and metrics

Review the summary

The workflow view shows total runs, success rate, average duration, and credits. Select a period that matches usage volume and decide whether simulations should be included.

Filter history

Filter by status, origin, version, agent, channel, and date. Keep simulations separate from real executions when evaluating user behavior.

Origins may include conversations, schedules, manual runs, simulations, and subworkflow calls.

Understand statuses

A run may be running, completed, failed, cancelled, waiting, or awaiting approval. Read the overall status together with the latest step and waiting or error reason.

Inspect details

The trace lists ordered steps with node type, status, duration, input, output, and error message. Open the related conversation when available to review the context that started the process.

To investigate a failure, locate the first incomplete step, inspect its non-secret input, verify dependency availability and permissions, compare a successful run from the same version, then correct and simulate the draft.

Use percentiles and usage

Percentiles reveal unusually slow runs that an average may hide. Review tokens and credits after adding agents, parallel branches, or generative steps.

Include the run identifier and workflow version when contacting support.