Connections and security

Every agent works with exactly the access you give it. Connections in the left rail shows, agent by agent, what each one can reach and what it may do there: your data, apps, team chats and logins.

How does an agent get access to my data?

Each agent has its own Data row in Your connections. Open the row to set the agent's access to each module of your workspace:

Access	The agent can
No access	Not see the module at all.
Read	Read the module's records.
Read & write	Read, create and change the module's records.

Automations is No access or Allowed. With Allowed, the agent builds and runs ACT automations for you. Its own scheduled jobs need no setting. See ACT automations.

Click Save access. When you widen an agent's access, AgentOS asks you to confirm with Give access.

Whose permissions does an agent use?

Its own. Every agent is a separate member of your workspace that never signs in and takes no seat on your plan. It sees only the modules you allowed, whoever is chatting with it. An agent without any access still has its AI models from the AI wallet. It sees none of your data.

What does Rotate key do?

Rotate key gives the agent a new key to your data and retires the old one. Use it when you want to be sure no old key is in use anywhere.

How do I connect an app to an agent?

Open the Library tab, pick the app, choose the agent under Connect it for, and click Connect. For an app that is not in the Library, click Connect with an agent. See Connect apps.

Who can change access?

Admins. Access is set per agent, so every change is visible in one place.

What is the Connect AI button?

Connect AI connects your own AI assistant, such as Claude, ChatGPT or Cursor, to your AgentOS data. It is for you, not for your agents. See Connect AI (MCP).

Next: Skills, or Add an agent with the right access from the start.