> ## Documentation Index
> Fetch the complete documentation index at: https://docs.reeva.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Monitor agents

> Review what your agents produced, sort by confidence, and approve flagged items.

The **Monitor** page is the shared view of every item your agents have processed. It's where you spend your time after agents are running — reviewing flagged results, exporting published data, and spotting agents whose confidence is dropping.

## What you see

Each row is a single item produced by an agent. Columns:

| Column         | Meaning                                                                    |
| -------------- | -------------------------------------------------------------------------- |
| **Item**       | The ID and name of the record (e.g., a SKU, a drawing number, a material). |
| **Agent**      | Which agent produced this item.                                            |
| **Status**     | Where this item is in the workflow. See statuses below.                    |
| **Confidence** | How confident the agent was in its result, as a percentage.                |
| **Updated**    | When the item last changed.                                                |

## Statuses

<CardGroup cols={2}>
  <Card title="Published" icon="check">
    The agent produced a result with high confidence and wrote it to the destination system. No action needed.
  </Card>

  <Card title="Flagged" icon="flag">
    The result fell below the confidence threshold or hit a guardrail. Open the item, review, and approve or correct it.
  </Card>

  <Card title="Linked" icon="link">
    The item was successfully matched to its counterpart in another system (e.g., a Teamcenter drawing linked to a SAP material).
  </Card>

  <Card title="Queued" icon="hourglass">
    The item is waiting to be processed — usually because an upstream step is still running or an integration is offline.
  </Card>
</CardGroup>

## Filter and triage

Use the status pills above the table (**All / Published / Flagged / Linked / Queued**) to narrow down to what needs attention. Triage flagged items first — they're the ones blocking the workflow.

## Review a flagged item

Click any flagged row to open its review panel. From there you can:

* See the source data the agent worked from
* See the AI-extracted result and confidence
* Approve, edit, or reject the result
* See which step in the agent produced the result

Approving a flagged item commits the result and moves it to **Published**. Rejecting it sends it back into the queue or marks it for the agent to ignore — depending on how the agent's decision step is configured.

## Export

Click **Export** in the top-right to download the current filtered view as a CSV. Useful for sharing with people who don't have Reeva accounts, or for offline review.

## Tips

* **Watch confidence trends.** If an agent's average confidence drops, something upstream has changed — a new document format, a renamed field, a new product line. Open a few flagged items to see the pattern.
* **Don't auto-approve flagged items in bulk.** They're flagged for a reason. If too many are flagged, lower the agent's confidence threshold deliberately rather than rubber-stamping the results.

## Next steps

* Iterate on the agent in [Build an agent](/agents/build).
* Check the [audit trail](/administration/users-and-audit) to see who approved what.