AgentsRoom Agent permissions: let your AI coding agent finish without stopping to ask
A real session, filmed in the app: Claude Code, started by hand with the flag that skips permissions, still stops on "Yes, I trust this folder" and waits. In Settings > AI providers, Claude tab, Agent permissions is on Autonomous, with a switch for the launch warning, folder trust, deletions and the email check. An agent launched by AgentsRoom in another new folder then writes its README in 14 seconds, without a single question.
Read the full feature pageTranscript
You skipped permissions, and Claude Code still stopped to ask. This is AgentsRoom, the desktop app where your AI coding agents work side by side. Here I start Claude Code by hand in a new folder, with the flag that skips permissions.
It still asks if I trust this folder. Nobody answers, so nothing happens. AgentsRoom answers those questions for you.
In Settings, AI providers, Claude: Agent permissions on Autonomous. Your agents launch with the bypass flag, and the launch warning, deletions and the email check are handled. So is the folder trust question.
Each one has its own switch. Prefer an agent that asks first? Pick Manual.
Now an agent that AgentsRoom launched in another new folder. I give it a task. No trust question, no warning.
It reads the folder and writes the README on its own. Fourteen seconds later, the README is there, and it never stopped to ask. Agent permissions is in AgentsRoom.
It is not a sandbox, so keep it away from production secrets. The link is in the description.