Python & Console Tools

Most of what you’ll want to do has a dedicated tool. For the rest, Direktor offers two escape hatches.

run_python

run_python executes arbitrary Python in the editor’s process using Unreal’s unreal module. It’s the universal fallback for multi-step logic, loops, or an API that no first-class tool covers yet.

Because it can do anything you could do in a Python script — including modifying or deleting assets — it is disabled by default. Enable it under:

Project Settings → Plugins → Direktor → Enable Python Tools

Once enabled, your AI client sees the run_python tool and can run code. When the code matches something a first-class tool does better, the result includes an agent_hint nudging the assistant toward that tool.

Caution

Treat an editor session with Python tools enabled like running an unreviewed script. Keep the project under version control, and turn the setting back off when you’re done if you’d rather not leave it on.

Note

Some console operations can crash the editor through native code paths that Python cannot catch — for example, passing a null world to a console command. Prefer run_console_command for console commands rather than driving them through run_python.

run_console_command

run_console_command passes a command straight to the Unreal console — the same commands you’d type into the editor’s console window (stat fps, r.ScreenPercentage, show flags, and so on).

This tool is enabled by default. Unreal blocks genuinely destructive console commands in the editor context, so it’s a low-risk way to toggle overlays, stat readouts, and rendering switches.

When to reach for each

• A dedicated tool exists → use it. Better ergonomics and clearer feedback.
• An Unreal API with no dedicated tool → the capability catalog (search → describe → invoke).
• Multi-step logic, loops, or something genuinely custom → run_python.
• A console command → run_console_command.