Agents

An agent is the part of Burrak that actually does the work. When you assign a task, an agent breaks it into steps, decides which tools to use, takes the necessary actions, and reviews the result before handing it back to you.

What makes an agent autonomous

• Planning — it works out the steps needed to reach your goal instead of waiting for instructions at each turn.
• Tool use — it can read and write files, search, run commands, and call the connectors you have enabled.
• Self-review — it checks its own output and iterates before declaring a task done.
• Memory — it remembers context across tasks so you do not have to re-explain yourself.

Creating and managing agents

You can run more than one agent, each with its own name, chat history, and setup. Create agents from the Overview page of any app, and open a chat with any of them whenever you like.

Roles & personalities

Give an agent a role from the marketplace — a ready-made persona and area of expertise, from copywriter to data analyst. The role shapes how that agent approaches every task and conversation, so you can keep a specialist agent for each kind of work.

Asking for input

A good agent knows when not to guess. If a decision genuinely needs you — an ambiguous requirement, an irreversible action — Burrak pauses and asks rather than charging ahead.

Permissions

You control what an agent may do on your machine. For example, you choose whether shell commands are always refused, always allowed, or require your approval each time — and agents ask before touching files outside their workspace.