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

# Message Queue

> Send follow-up prompts while the AI works

You do not have to wait for the AI to finish before sending your next request. Prompts sent while a turn is running join a queue and run one at a time, in the order you sent them. Nothing is dropped, and nothing interrupts the work already in progress.

## Queueing a prompt

Type your message and press **Enter** as usual. If the AI is busy, or other prompts are already waiting, your message is queued instead of being sent immediately. An amber queue strip appears just above the message input showing how many prompts are waiting and a preview of the one that runs next.

<Note>
  The queue lives on the server, not in your browser tab. Queued prompts keep running in order even if you close the tab or navigate away, and the chat catches up when you return.
</Note>

## Managing the queue

Click the queue strip to expand it. From the expanded view you can:

* **Reorder** -- drag any waiting prompt by its handle to change the run order
* **Inspect** -- click a prompt to see its full text, attached images, and the model tier and mode it will run with
* **Remove** -- delete a single prompt from the queue, either from its row or from the inspector with **Remove from queue**
* **Clear** -- remove every waiting prompt at once. Clearing asks you to confirm before anything is deleted

While a prompt is running, its row stays pinned at the top of the expanded view so you can see what the queue is currently working on.

### Attachments in queued prompts

Images and files attached to a queued message travel with it and reach the AI when the prompt runs. Up to 6 images per queued prompt are sent to the model; if you attach more, the inspector marks the extras so you know they will not be included.

## Interrupting and pausing

* **Interrupt & run next** -- shown on the queue strip while a turn is running with prompts waiting. It stops the current turn and starts the next queued prompt right away.
* **Stop** -- stopping the current turn also pauses the queue. The strip switches to a paused state showing how many prompts are waiting, and nothing runs until you click **Resume**.

<Note>
  When you stop or interrupt a turn mid-edit, any unsaved changes from that turn are discarded rather than half-applied. The chat tells you when this happens. A cancelled prompt never leaves a partial edit in your project.
</Note>

## When a queued prompt fails to start

If a queued prompt cannot start (for example, the AI service is briefly unreachable), it is tried again automatically. After 3 failed attempts it is marked **failed** and stays visible in the queue with the error.

Because the queue runs strictly in order, a failed prompt blocks the prompts behind it. Click **Retry** to try it again, or **Remove** to skip it and let the rest of the queue continue. Failed prompts are never silently deleted.

## Tasks sent to chat

Using **Send to AI chat** on a task works through the same queue. If the AI is busy, the task is queued like any other prompt, and its screenshots and attachments are included when it runs.

## Queued prompts and cost

A queued prompt is charged when it runs, exactly like a prompt you send directly, based on the model tier it was queued with. Removing a prompt before it runs costs nothing. See [AI Credits](/settings/ai-credits) for how credits are consumed.