> ## 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.

# Many jobs at once: AI agents that work in parallel

> One Workser AI agent can answer customers in seconds while it works on a website or a cloud computer. Every request is its own job, and nothing waits for a computer it doesn't need.

**A Workser AI agent is not one tab doing one thing.** It answers a customer on LINE in seconds while it works a website on the cloud computer, and while a teammate builds something else. Every request is its own job, every job gets what it needs, and nothing waits for anything it doesn't need.

## Every request is its own job

* **A customer writes in a chat app** (WhatsApp, Telegram or LINE): a new job starts as soon as their messages pause, whatever else the AI agent is doing. A reply that needs no computer goes out right away.
* **You write while a job is running:** Workser reads whether you're asking how it's going, changing the running job, or asking for something new. New work starts beside the running job, never queued behind it.
* **An automation fires, or a teammate hands work over:** each starts its own job.
* **A job that is waiting** (for your answer, or for a teammate) saves itself and frees its place, so waiting costs nothing.

## Three ways an AI agent works

| Where | Good for |
| - | - |
| **No computer at all** | Answering customers, writing, research, your business data, connected apps, email and phone. Starts in seconds. |
| **The cloud computer** | Websites, desktop apps, files and downloads, with your saved logins. |
| **A private sandbox** | Documents, design or code that need their own space, removed when the job ends. |

On a computer an AI agent picks the fastest way that works: a connected app first, then reading and filling pages as text, then looking at the screen and clicking like a person.

## Several jobs on one computer

Every project has a shared cloud computer, signed in once for every AI agent. When a second job needs it while another is working on it:

1. **It takes another screen of the same computer**, with the same files and logins and its own mouse and keyboard, up to three screens.
2. **When those are taken, it works on a copy of the computer** made from its latest restore point. The AI agent is told it is on a copy. Files it downloads or saves in Downloads, Desktop, Documents or its work folder are saved to your project's files (in **From computer copies**) when the job ends, then the copy is removed.
3. **A step that needs the live computer itself** (a file just saved there, or you signing in) waits its turn instead of working on an older copy.

## How many at once

Your plan sets how many cloud computers and sandboxes work at the same time (see [Logins and computers](https://docs.workser.ai/ai-workforce/logins-and-computers.md)). Jobs that need no computer don't count against it, and a paused computer frees its place. You can add dedicated computers for \$5 a month each.

When everything your plan runs at once is busy, a new job **waits in line and starts by itself** when a place frees up:

* **People first.** A job you or a customer asked for goes ahead of automations and hand-offs between AI agents.
* **Told where it was asked.** The task card shows "Starting soon" with its place in line. A customer who wrote in a chat app gets one short message that their request is being handled, and if you sent the task from a chat app, you get the same line there.

## Common questions

> [accordion] Can one AI agent do several things at once?
>
> Yes. Every request is its own job, so one AI agent can answer customers while it works on the cloud computer, and new requests start beside the running ones instead of waiting.

> [accordion] Will my customers wait because an AI agent is busy?
>
> No. Customer messages start their own job right away, and a reply that needs no computer goes out in seconds. If your plan's places are all busy, the customer gets a short message that their request is being handled, and their job goes ahead of automations in the line.

> [accordion] Do AI agents fight over the cloud computer?
>
> No. A second job gets another screen of the same computer, then a copy of it, and the work done on a copy is saved to your project's files.

> [accordion] Do I set up logins for each AI agent?
>
> No. Sign in once on the cloud computer and every AI agent you allow can use it.

## Related

* [AI model routing](https://docs.workser.ai/ai-workforce/model-routing.md): the best model for each job
* [Logins and computers](https://docs.workser.ai/ai-workforce/logins-and-computers.md)
* [Chat apps](https://docs.workser.ai/ai-workforce/chat-apps.md)


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