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

# Working with agents

> How to choose agents and keep conversations clear.

# Working with agents

AgentDesk is designed around a team of agents. Each agent has a role, context, and session. Choosing the right agent helps keep work focused.

## Agent vs assignee

| Term         | Meaning                                                                                    |
| ------------ | ------------------------------------------------------------------------------------------ |
| **Agent**    | A named AI teammate you can talk to.                                                       |
| **Assignee** | The agent currently responsible for doing a task or work chunk.                            |
| **Lane**     | Internal UI/session mapping for one agent conversation. You usually do not need this term. |

## Choosing the right agent

| Work type                                | Suggested agent type        |
| ---------------------------------------- | --------------------------- |
| Requirements, scope, acceptance criteria | Product / PO assistant      |
| Architecture or cross-system design      | Solution architect          |
| Backend/API/data work                    | Backend engineer            |
| Frontend/UI work                         | Frontend engineer           |
| UX/UI design                             | Designer                    |
| Testing/review                           | QA/reviewer                 |
| Deployment/runtime operations            | DevOps                      |
| Positioning/marketing/PMF                | Marketing or business agent |

## Avoiding confusion

* Check which agent is selected before sending a message.
* Keep follow-up questions in the same relevant agent conversation when possible.
* If you want another agent to review or continue work, say that explicitly.
* If an agent seems to be missing context, reference the task id or attach/link the relevant file.

## What should be visible

When an agent is working on something, you should be able to see:

* The task or subtask being worked on.
* The assignee.
* The current next action.
* Whether the task is active, waiting, blocked, or done.
* Evidence or output when the task is complete.