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

# What is an Agent?

> Understanding Dialog's agent model and how agents work

An agent in Dialog is a specialized AI assistant with its own dedicated workspace, tool configuration, and personality. Each agent operates independently, maintaining its own memory and context across every interaction. Think of agents as team members you deploy for different jobs.

## Agent architecture

Every agent you create gets its own persistent workspace environment -- a dedicated runtime that persists across sessions. This workspace stores your agent's [identity files](/core-concepts/memory/identity-files), saved content, research notes, task lists, databases, and custom skills.

An agent bundles four key components:

* **Workspace** -- A persistent environment with up to 50MB of storage for files, notes, research, bookmarks, and databases
* **Tool configuration** -- A customizable set of [enabled data sources](/core-concepts/tools/tool-preferences) that determine what the agent can access (74+ tools available)
* **Session history** -- A full record of every conversation, research chat, and follow-up
* **Scheduled tasks** -- Any automated research, monitoring, or briefing jobs the agent runs on your behalf

These components work together to give each agent a distinct capability and growing knowledge base.

## Named personas

Agents come with configurable [personas](/core-concepts/agents/agent-personas) optimized for different workflows:

* **Chief of Staff** -- Your default agent. Strategic assistant for managing tasks, saving links, organizing research, and daily briefings.
* **Social Media Manager** -- A preset you can add for social-first workflows: content ideation, drafting posts, managing feeds.
* **Product Analytics Agent** -- A preset focused on product metrics, funnels, and user behavior analysis.
* **Custom Agent** -- Build your own persona with a name and instructions tailored to your needs.

Each persona shapes how the agent approaches tasks, structures its output, and communicates with you. See [Agent Personas](/core-concepts/agents/agent-personas) for details.

## One default agent

Every Dialog account comes with a **Chief of Staff** agent that's ready to use right away. This is the agent you interact with when you first sign in, and the one remaining after any others are removed.

Chief of Staff works like any other agent -- it has its own workspace, tool configuration, and session history. The only difference is that it can't be deleted. Think of it as your home base.

You can [create additional agents](/core-concepts/agents/creating-agents) whenever you need a specialized assistant for a different domain or workflow.

## Agent isolation

Agents are fully independent from each other. They don't share workspaces, session histories, tool configurations, or any other data.

This isolation is intentional. It means you can:

* Run a Chief of Staff for day-to-day operations and a Product Analytics Agent for metrics work without any bleed-through
* Give different agents access to different tools based on their purpose
* Keep research contexts completely separate across projects

If you need to share information between agents, you can reference findings from one agent in a conversation with another, or use connected apps to store findings in shared tools like Notion or Google Docs.

<Warning>
  Deleting an agent permanently removes its workspace, session history, and all associated data. This action cannot be undone.
</Warning>

## When to use multiple agents

| Scenario                       | Approach                                                                     |
| ------------------------------ | ---------------------------------------------------------------------------- |
| One person, multiple workflows | Create agents for each major workflow (research, content, operations)        |
| Different projects             | One agent per project to keep contexts clean                                 |
| Team members                   | Each person gets their own agent(s) with personalized workspaces             |
| Experimentation                | Spin up a new agent to test a workflow without affecting your main workspace |

## Lazy provisioning

Agent workspaces are created on first interaction, not when the agent is initially set up. This keeps account creation fast and avoids allocating resources for agents you haven't started using yet.

Agent caps depend on your plan -- see [Billing & Credits](/guides/billing-and-credits) for the per-tier limits. Each agent provisions its workspace independently when you first engage with it.