Background Agents

Background Agents

Background Agents are long-running AI tasks that continue processing after the initial request has started.

They are useful for work that may take longer than a normal chat turn, such as:

• multi-step reasoning
• web research
• document-heavy processing
• tool-driven workflows
• jobs that continue after a page response has already returned

What appears in Running Agents

The Running Agents section in the Usage Dashboard is designed for background work only.

It shows background runs that are currently:

• Queued
• Running

Normal foreground chat activity is not shown in this list, even if a user is actively chatting with an agent.

What the status means

• Queued means the background run was accepted and is waiting to begin or has just been handed off.
• Running means the background run is actively processing.
• Completed, Cancelled, and Failed are terminal states and do not remain in the Running Agents list.

How monitoring works

The Running Agents list gives you a quick operational view of active background work.

Each row shows:

• the agent
• the current status
• when the run started
• the most recent update time

This helps an administrator or business owner quickly spot a job that has been running longer than expected.

How to stop a background agent

To stop a long-running background agent:

1. Open the Usage Dashboard.
2. Go to the Agents tab.
3. Find the row in Running Agents.
4. Click Stop.
5. Confirm the stop action.

The system sends a cancellation request to the active background run and then refreshes the Running Agents list.

What to expect after you click Stop

• If the run is still active, it should move to Cancelled and disappear from the Running Agents list after refresh.
• If the run already finished just before the stop request, you may see a message that it was already completed or already cancelled.
• Stopping a run prevents additional work from continuing, but it does not undo work that has already finished.

Billing and credit consumption

Background agents consume credits while they work.

Depending on the job, a background run may use:

• message credits
• skill credits

In general:

• longer processing can consume more credits than a short chat turn
• tool-heavy workflows may consume both message and skill credits
• stopping a run helps prevent further usage from continuing

Important billing note

Stopping a background agent does not refund credits that were already consumed before the stop request was processed.

The stop action is intended to limit additional usage, not reverse work that has already been completed.

When to stop a run immediately

Consider stopping a background agent right away if:

• the wrong agent was started
• the request was submitted by mistake
• the task is clearly going in the wrong direction
• the run appears unusually long for the type of work requested
• you are concerned about unnecessary credit consumption

Good operating practice

For organizations that use background agents regularly:

• review the Running Agents section during heavy usage periods
• stop unexpected jobs quickly
• compare active runs with monthly usage trends in the Usage Dashboard
• train admins to distinguish normal chat activity from background processing

Need help

If a background run appears stuck, keeps returning unexpectedly, or usage looks higher than expected, contact iDialogue Support with:

• the organization name
• the approximate time of the run
• the agent name
• any visible status shown in the dashboard