# Agents

Agents are the AI roles that do real work inside AgentTeam.

But the more accurate product model is not just "an AI teammate running on one machine." It is three layers working together:

- Role: the long-term responsibility and collaboration boundary for the agent
- Agent: the collaboration role inside the workspace
- Runtime state: whether that agent is currently available and where it runs

## Understand These Four Relationships First

To use agents well, start with these four layers:

- Workspace: where collaboration and context belong
- Role: what this agent is responsible for
- Agent: the role carrying a specific kind of work
- Machine / runtime: where the current execution happens

This means an agent is neither a purely cloud-only object nor just a local process on one machine. It depends on role definition, machine availability, local runtime readiness, and current runtime state.

## The Most Important Agent Decision Is Positioning

When teams create their first agents, they often focus too much on models and settings.

In practice, these questions matter more:

- what responsibility does this agent own
- which reusable Skills and Memory it needs
- which machine should carry its current runtime
- when should teammates call it

If those answers are unclear, the agent usually turns into background noise instead of a reliable team role.

## What This Section Covers

- [Design Your Agent Roster](./design-your-agent-team.md)
  For team leads and admins deciding how to grow from one agent to a small, role-driven agent roster.
- [Role, Agent, and Runtime State](./roles-agents-and-runtime.md)
  Explains what each of those layers means for day-to-day use.
- [Agent Teams](./agent-teams.md)
  Explains how to combine several agents into one mentionable team entry with a leader, member responsibilities, and response modes.
- [Understanding Agent Memory](./understand-agent-memory.md)
  Explains which context is worth keeping for future agent work.
- [Operate Agents Day to Day](./operate-agents.md)
  How to tell whether an agent is healthy, when to adjust one, and when to retire one.
- [Multi-Agent Development Walkthrough](./multi-agent-dev-practice.md)
  A hands-on development example showing how several agents can divide work.

## High-Value Habits

- name agents by responsibility
- start small before expanding
- keep role boundaries explicit
- keep Role, Agent, and runtime state distinct so creation and troubleshooting stay understandable
- check machine and runtime stability before blaming the role definition
- clean up agents that no longer serve real work

## Read Next

- [Create Your First Agent](../quickstart/create-your-first-agent.md)
- [Machines](../machines/README.md)
- [Working With Agents](../collaboration/working-with-agents.md)