> ## 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.

# Chats: give work and follow it

> How to give a Task in Workser chat, choose whether Workser or one AI agent takes it, pick a model, watch the work, stop it, and find what it made.

**Chat is where you give work in plain words.** Type in the box at the bottom ("Give your team a task, or ask anything"), and
Workser picks the right AI agent — or you pick one yourself, and the box says "Message \<name>". A Task starts when you press
Send, and the chat shows what the AI agent is doing until it is Done.

## Send a Task

**1. Say what you want, as you would to a new hire**

"Match last month's payments to our invoices", "Draft replies to this week's reviews for my OK", "Turn this sheet into a
short report".

**2. Choose who takes it (optional)**

Leave it on Workser and the hint applies: **"The right AI agent picks it up"**. Or name one AI agent and the box becomes
**Message \<name>**.

**3. Add what it needs**

Attach up to 10 files, or type **@** to add a tab, app, AI agent, computer, file, connector, record or earlier chat. The
model chip above the box picks the model. See [AI models](https://docs.workser.ai/using/models.md).

**4. Send**

The Task is charged in credits as it runs. A start screen shows an estimate when enough similar Tasks have finished, and
the Task's own sheet shows the credits used so far.

While it runs, the send button becomes **Stop** when the box is empty. It asks "Stop this task? What it has done so far is
kept." — **Stop task** or **Keep going**. Work already done stays.

## While it runs

* **A live line** under the AI agent's name says **Working**, **Starting · …** or **Request · …** with the Task's title.
* **Every step** appears in the chat as it happens, with what it used and, when a step fails, what it said.
* **The AI agent's screen** is one click away: open its panel → **Today**, or the Task's **Watch** view.
* **It asks when it is risky.** The ask is docked above the message box; the Task waits. See [Tasks](https://docs.workser.ai/using/tasks.md).
* **Credits used so far** are on the Task's own sheet (the credit meter), not in the chat header.

## One Task, one chat, many chats

The chats list keeps every conversation. **Search chats**, or use the jump strip at the top of a chat to go back to an
earlier one. A chat remembers the workspace tabs you opened in it, so returning to it returns to that work.

Tasks at once by plan: Starter 5, Pro 10, Team 5 per person. When the plan or the computer is busy, the Task says **Waiting
for a computer** or **Starting soon (2 in line)** and begins on its own; it is queued, not refused.

## When it ends

A finished Task is checked against what you asked before it says **Done**. You see:

* **Its answer** in the chat.
* **What it made**: a result card (documents, sheets, slides, images, pages), each opening in its own tab.
* **What it changed**, if it changed an app: a **Changes** tab with a diff and Preview, Publish or Undo.
* **What it wants next**, when the AI agent suggests a follow-up Task. You start it or dismiss it.

If it cannot finish, it says **Stuck** and what it needs. The ask waits in [Requests](https://docs.workser.ai/using/tasks.md).

## What a chat is not

* It is not a general chatbot: the AI agent has your business's rules, logins and files, and it works in your apps.
* It is not a queue of prompts: **Send** starts a Task with a start, steps, a result and a cost.
* It is not private from your team: people in the business see the same projects and Tasks, according to their role. See
  [Settings](https://docs.workser.ai/using/settings.md) for who can change what.

## Related

* [Tasks and Requests](https://docs.workser.ai/using/tasks.md): what happens while it runs
* [The AI agent panel](https://docs.workser.ai/using/agent-panel.md): the AI agent's screen, memory and rules
* [Automations](https://docs.workser.ai/ai-workforce/automations.md): work that starts without a chat


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