Turn a failed pipeline into the next action

A failed pipeline is rarely one isolated error. The useful clues live across the block, the run history, the logs, the pipeline graph, and often the data itself. This workflow keeps that investigation in one place. Start from a failed run, bring the relevant context into Mage's AI Sidekick, inspect what changed, and move into data exploration or a chart when the answer needs evidence.

Start with the run that needs attention

Open the failed block and its run details. Sidekick can work from the pipeline, block, page, dashboard, and editor context already in front of the team, so the investigation begins with the work rather than a blank prompt.

Ask with the context that matters

Use run history, logs, block context, and attached files or folders to ask a focused question: what changed, which step failed, or what the error means. The August release adds smoother streamed answers, saved chat history, and clearer tool activity so teams can follow the work as it happens.

Check the underlying data

When the next question is about the data rather than the error, open the new Data Explorer. Preview connected data, inspect remote SQL or Parquet results, and keep the investigation tied to the active workflow.

Turn evidence into a next step

Sidekick can create previews and chart artifacts from block outputs. That gives a team a way to examine the result, share a concrete view, and decide whether to adjust code, retry the run, or change the schedule.

What this workflow enables

• Investigate a failed run without collecting context across separate tools
• Keep the pipeline, error, logs, and working data in one debugging flow
• Move from an answer to a data preview or chart when the team needs evidence
• Retry or prioritize the right work once the next action is clear

The August Mage Pro release notes include the full release. For the broader product workflow, explore AI Sidekick.