Approvals and autonomy, keeping risky actions gated

MyAgens agents are real agents that can run commands and edit files on your machine, whichever AI backend they run on, so control matters. MyAgens gives you two things: an approval prompt on anything risky, and an autonomy level you set.

Approve, deny, or always allow

Every action that could change something, running a shell command, writing or editing a file, using a connector that writes, pauses and asks you: Approve, Deny, or Always allow. "Always allow" whitelists that tool for the rest of the session. Read-only actions like reading or searching files run without prompting. You answer right in Telegram, or in the panel's Approvals view.

Three autonomy levels

Set how much MyAgens does on its own with /mode:

• Supervised: every tool prompts, even read-only ones. Maximum control.
• Standard (default): read-only runs freely, anything risky prompts. The sensible middle.
• Full: no prompts, the agent runs on its own. Use it when you trust the task.

You can set the level per agent or per chat, so a trusted background Lead can run in full while your main chat stays supervised.