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

# Agents and fleets

> Learn how SuperEmber uses persistent agents and explicit fleets to divide work.

Agents are the workers in SuperEmber. Teams define the shared people-and-access boundary; Fleet shows the persistent Team Agent and workforce hierarchy inside those boundaries.

## Agents

An agent is a persistent worker with a role, instructions, model setting, tools, memory/runtime context, workspace files, and identity.

Use an agent for work that needs interpretation or judgment, such as:

* researching a customer account
* drafting a reply
* qualifying a lead
* summarizing a document
* checking campaign performance
* preparing a report

Agents can be used directly from chat or as workflow steps.

Agents are not just generic model calls. They are named workers that can be configured, connected to tools, and inspected through dashboard surfaces.

Each agent can follow the platform default model or store its own model choice. Personal agents use that setting like a per-user preference; shared agents use it for everyone who talks to that agent.

## Team Agents and workforce agents

A Team Agent is the primary shared AI identity for one team. Persistent specialists beneath it are called workforce agents in the UI. They may be called subagents generically, but their Fleet reporting edges determine who can delegate directly to whom.

A workforce agent is different from a subteam. The first is an AI role in the reporting hierarchy; the second is an organizational team boundary.

## Fleet

A fleet is a group of agents that can work together.

A fleet is useful when a task benefits from specialization. For example, one agent can research, another can draft, and a reviewer can synthesize the final response.

Fleet coordination should be visible. SuperEmber models collaboration through explicit delegation records, operation threads, and workflow joins instead of hiding child-agent work inside one opaque chat turn.

## Common fleet patterns

| Pattern              | What happens                                                |
| -------------------- | ----------------------------------------------------------- |
| Manager delegates    | One manager decides which specialist should handle the task |
| Parallel specialists | Multiple agents work at the same time                       |
| Reviewer synthesis   | A reviewer agent combines or selects outputs                |
| First success        | The workflow continues when one usable result is ready      |
| All complete         | The workflow waits for every assigned agent                 |

## Relationship to workflows

Agents and fleets do not need workflows for every task. A direct Slack or Teams delegation can be answered by SuperEmber when the work is simple.

Workflows become useful when the task needs structured inputs, branching, retries, waits, approvals, connector calls, replay, or run inspection.

## Related Pages

* [Agents](/product/agents)
* [Teams and Team Agents](/product/teams-and-team-agents)
* [Memory](/product/memory)
* [Fleets](/product/fleets)
* [Manage a fleet](/guides/manage-a-fleet)