Turn on agent orchestration
What an admin turns on, which permissions people need, and how to give an AI connection access to agent work.
This feature is in beta. If you do not see it, ask your admin to turn it on for your organization.
Agent orchestration is off until an admin turns it on for your organization. When it is off, the Agents page does not appear and agents cannot claim anything.
Ask your admin to turn it on
Section titled “Ask your admin to turn it on”Ask your ProjectBrain admin to turn on agent orchestration for your organization. It is a beta feature, so it may change.
Turning it off again works as a stop switch. Agents stop getting work, and nothing about agent work is sent to people while it is off.
Permissions
Section titled “Permissions”| Permission | Who has it by default | What it allows |
|---|---|---|
agents.view |
Owner, executive, team lead, individual contributor | See the Agents page and what agents are working on. |
agents.manage |
Owner, executive, team lead | Approve plans, answer and decide, accept or send back work, pause agents, take over a task. |
Clients and external users hold neither. An individual contributor sees the page but has no buttons that change anything. Owners can change who holds these at Settings > Organization > Roles. A permission in one organization grants nothing in another.
Give an AI connection access to agent work
Section titled “Give an AI connection access to agent work”An AI client such as Claude Code connects to ProjectBrain through an MCP connection. To claim tasks, that connection needs the agents.work permission. It is only offered to people who hold agents.view.
- New Claude Code and Codex connections get it by default.
- An existing connection shows that a new permission is available in your MCP connection settings. Add it there.