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

# Tasks and Requests

> What a Task's states mean, the four kinds of ask, how to answer or retry one, and where everything waiting for you is collected in Requests.

**A Task is one piece of work: it starts when you Send, moves through steps, may ask you something, and ends as Done or
Stuck.** You follow it in the chat where you asked. Everything waiting for you is collected in **Requests** (the page at
`/needs-you`, titled Requests), so you do not have to hunt through chats.

## The words a Task uses

| Word | What it means |
| - | - |
| **Queued** | Waiting for a free place in the plan's "Tasks at once". |
| **Starting…** | Its computer or workspace is being prepared. |
| **Working** | It is doing the work now. |
| **Request** | It asked you something and is waiting. |
| **Ready for you** | It finished a step that needs a person before it carries on. |
| **Waiting** | It is waiting on something else — a free computer, or a step outside it. The Task's own line says which. |
| **Paused** | Stopped for now; you can let it carry on. |
| **Done** | Finished. A final check says "Checked: done", or "Not sure it's complete" when it could not confirm. |
| **Didn't finish** | It ended without doing the work. The reason is in the chat. The task badge and Requests say **Stuck** for the same state. |
| **Put away** | You stopped it, or it was cancelled. |

## The four kinds of ask

Every ask looks the same and its head says what kind it is:

| Head | It asks | Your answers |
| - | - | - |
| **Question for you** | A question, with choices or free text | A chip (its words are the answer), or "Or say it in your own words" |
| **Needs your OK** | What it will do if you agree | **Yes, go ahead** / **No**, or **Allow once** / **Don't send** |
| **Needs your permission** | A step on a computer or in an app | **Allow** / **Deny**, and sometimes **Always allow** |
| **Stuck and waiting for you** | What happened, what it tried, what it needs | **Carry on**, or tell it what to do next and **Send**, or **Dismiss** |

Asks are docked at the foot of the conversation, just above the message box, under the newest reply. When one Task asks
several yes/no questions, you answer them one page at a time ("Reply 2 of 3") and nothing is sent until you review.

Two answers have their own buttons inside the same card: **money** opens the card service's passkey page ("Approve with your
passkey"), and **Save your sign-in for \<site>** asks for a username, password and optional 2FA key, saves it to the Logins
vault for that AI agent, and never puts those values in the chat.

**Always allow** is for an owner or admin and is offered only on a permission ask. It lets that AI agent do that kind of step
without asking again ("Always allow for \<name>"); you can undo it in Autonomy.

Replying to a Stuck Task does not resume the old one: the card says so. **Your reply starts the next Task for this AI agent,
with your words.**

## Requests

**Requests** is the page with every question, approval, suggestion, payment and sign-in waiting on you, grouped by project
and "Most urgent first". Filters are **All**, **Approvals**, **Questions**, **Suggestions**, **Payments** and **Sign-ins**,
and **From** narrows to one AI agent. Home shows the first few under **Requests**, with **See all**.

* If two identical approvals are waiting, it offers **Approve all N** or "I'll read them one by one".
* Answer one and the card leaves the list. "That one is already answered." means someone else got there first.
* **Nothing is waiting on you.** is the empty state; your team is still working.

## Watch, take over, or step in

* **Watch**: a Task's live view shows the screen it is working on. The Watch tab becomes a replay after it ends.
* **Take over**: on the project's cloud computer or browser, choose **Take over the computer** or **Take over the browser**.
  In the computer view it is **Take control**, and **Give back to \<agent>** returns it. While you hold it, the Task waits;
  you can keep it for 30 more minutes.
* **Stop**: the send button becomes **Stop** when the box is empty. It asks "Stop this task? What it has done so far is
  kept." with **Stop task** and **Keep going**.
* **Retry**: a failed Task's own conversation offers **Retry**, which sends the same starting message again. The AI agent's
  message in a chat does not: it says what got stuck and why.

## When it says Done

The chat shows each step, so you can check the evidence. If the result is a document, a sheet, a deck, an image or a page, it
is in the Task's **Result**; if it changed an app, the **Changes** tab shows the diff with Preview, Publish and Undo.

The final check has three outcomes: **Checked: done**, **Not sure it's complete** (finished, but it could not confirm), and
**Didn't finish** with the reason.

## Cost and the 10% pause

Work is charged in credits as it runs. A single paid Task pauses and asks after it uses **10% of the month's credits**
(Starter: 200; Pro: 1,050; Team: 10% of the team's shared credits, 440 for two people), so one Task cannot quietly use the
month. A Task that passes its limit pauses and asks whether to keep going or stop. With no answer in 4 hours, the Task stops. Nothing is billed past
your credits: work stops until you add credits or change plan.

A Task with its own **Credit limit per task** (Manage → Advanced) stops at that limit without asking.

## Related

* [Chats](https://docs.workser.ai/using/chats.md): giving the Task and reading the answer
* [Approvals and Stuck](https://docs.workser.ai/ai-workforce/approvals-and-stuck.md): the rules behind what asks
* [Autonomy and probation](https://docs.workser.ai/using/autonomy-and-probation.md): reducing the asks over time
* [Troubleshooting](https://docs.workser.ai/using/troubleshooting.md): when a Task will not move


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.