Permission modes

Every tool is tagged with a risk level, and the mode you pick decides which

tools your AI can actually call. You set the mode in the app; the MCP server

enforces it, so the model can't exceed it no matter what it asks for.

Read Only

Inspect only. The AI can read scenes, scripts, nodes, the project structure, and

live editor state, but cannot change anything. Great for reviews, planning, and

letting a model learn your project before it touches it.

Safe Edit

The recommended default. Adds creating and editing (files, nodes, scripts,

scene changes) while still blocking anything destructive. No deletes, no

shell/command execution. You can hand the reins over and not worry.

Full Control

Everything, including deletes and running commands. Destructive and execute-level

actions still ask for confirmation before they run, so nothing irreversible

happens silently. Use this when you trust the workflow and want the AI to move

fast.

Risk levels, briefly

• Read: never changes anything.
• Write: creates or edits.
• Destructive: deletes or overwrites; confirmation required.
• Execute: runs commands; confirmation required.

Switching modes takes effect immediately, with no restart. When in doubt, drop to a

lower mode; you can always raise it for a single task.