Vibe Check: Agents don't need your password. They need your permission.

​AI agents are doing real work for us now. They read calendars, update records, send messages, and kick off jobs that keep running long after we have closed the tab. The models are used more and more as autonomous or semi-autonomous agents, which is why it matters to understand what is the agent actually allowed to touch.

​That is the hard part. An agent acts on your behalf, so it has to act under some identity, with some set of permissions. The easy path is to hand it your credentials and turn it loose, which is also the path that should worry you. The better one is scoped, revocable access it can use without ever holding your keys.

​Jessica Temporal, Senior Developer Advocate at Auth0 and co-founder of the Pizza de Dados podcast, is joining the stream to show how to think about and approach this problem. Melanie is co-hosting with Ian Douglas, who spends his days building with AI agents. We will explore what happens to that permission and the agent's state especially when the are disruptions. Where it goes depends on what Jessica brings, so come with questions.