What is a role-based access?
Role-based access means each person or tool gets only the permissions its job needs, so it can do some things and not others.
Also called: Permissions.
In plain words
The idea is the same as in a shop. The person at the till can take payments, but they may not have the keys to the stockroom. Giving each role only what it needs keeps mistakes small and makes it clear who can do what.
It matters even more with AI agents. An agent that only needs to read messages should not be able to send them. An agent that drafts replies should not be able to connect new accounts.
Limiting what each one may do is a simple, sensible habit.
A good way to start is with the smallest set of permissions that gets the job done, then add more when you trust the setup. You can always change it later, and you can switch one agent off without affecting the others.
Permissions are one layer. Limits, pacing and approvals sit alongside them, so even an agent with permission to send is still held to a safe pace.
How it works in Spaces
Every AI agent or key has its own permissions: read, reply, post and comment, connect accounts and ads. Each also has its own accounts, its own daily share and its own pause, set by the owner under Your agent.
A key can be read-only, or draft-only, so its messages always wait for approval. Account limits always apply on top, whichever permissions an agent has.
Related terms
- API key: An API key is a long secret code that proves to a service that a request comes from you, so it works like a password for software rather than for people.
- Approval: An approval is you checking and saying yes to a message or post your AI agent wrote before it is sent.
- Team member: A team member is a person on your own team, such as a colleague who answers customers, who can take over a conversation from your AI agent.
- AI agent: An AI agent is an AI model, such as ChatGPT, Claude or Grok, that can take actions with tools you give it, like reading a message and sending a reply, rather than only chatting.
- Workspace: A workspace is your home in Spaces: the one place that holds your connected accounts, API keys, AI agents, settings and billing.
FAQ
Can I let an agent read but not send?
Yes. A key can be read-only, so it can look at conversations without sending anything.
Can different agents have different permissions?
Yes. Each agent has its own permissions, accounts and daily share.
Can I pause just one agent?
Yes. A paused agent can read and send nothing until you resume it.
Do permissions override the limits?
No. The account limits always apply on top of what an agent is allowed to do.
Useful next: Account safety, Security, Docs. Every term: the glossary.