Your agents work within rules

Giving an AI agent access to your content is a question of trust. How much can it do? What happens if it gets something wrong? What if it is compromised?

Forge answers these questions with the same mechanism it uses for your team: roles and tokens.

Access by token

Every agent -- and every team member -- gets a token. The token determines what that agent is allowed to do: draft content, publish it, archive it, delete it. You decide what each token can do when you issue it.

An agent that can draft cannot publish. An agent that can publish cannot necessarily delete. If you want an agent that creates content for a human to review before it goes live, that boundary is enforced by the system -- not by a process people have to remember to follow.

When something goes wrong

If an agent produces bad output, the lifecycle rules contain the damage. A draft is not visible to anyone outside your team until someone with publishing rights approves it. Visitors, search engines, and AI indexes see nothing.

If a token is compromised, you revoke it. Access stops immediately.

One rule you cannot override

There is one guard built into the system regardless of how you configure access: you cannot remove the last administrator. If you are the only person with full access, the system will not let you revoke it -- accidentally or otherwise.

This is not a setting. It is not something to remember to configure. It is always there.

What this means in practice

Your agents operate under the same access rules as your team. The rules are enforced by the platform -- not by the people following them. You grant exactly what is needed, and the system holds that boundary.

Read about content lifecycle · Get started with Forge