execute_python

Execute Python code inside the running Unreal Editor. DO NOT GUESS A FIELD NAME: most of this tool's errors are "object has no attribute" -- and the field usually EXISTS, just not where you reached for it: HitResult().impact_point raises, get_editor_property cannot find it, and only to_dict() returns it (Transform.translation is the reverse). The engine's own stub and Epic's web docs list it either way, so neither settles it -- ask the object (dir(x), x.to_dict(), x.get_editor_property(...)). PIE TRAP -- this tool is the ONE HOLE in the plugin's PIE guard: the commands that compile a Blueprint or write to disk are REFUSED while Play-in-Editor is live (McpPieSafety.gen.h), but arbitrary Python cannot be classified statically, so a script that reaches compile/save through unreal.* (BlueprintEditorLibrary.compile_blueprint, EditorAssetLibrary.save_asset, save_loaded_asset, anything that dirties a Blueprint) RUNS ANYWAY. With PIE live that ends either in a DEAD EDITOR -- the compiler's UREINST_* trampoline retains a world and CheckForWorldGCLeaks is Fatal, taking every unsaved edit with it -- or in a save that reports success without writing a byte (Editor.AllowSavingAssetsDuringPIE=0). So: call is_play_in_editor_active FIRST whenever your script compiles, saves, or mutates a Blueprint, and stop PIE before running it. READING the live world during PIE is safe and is exactly what this tool is for -- do not avoid it for that. BEFORE writing a loop, check what already answers it in ONE call: scripts that re-implement something a tool already does fail far more often than scripts doing something genuinely new -- hand-rolled scaffolding is where the errors are, and a bug in the scaffold reads exactly like a finding about the game. enumerate actors of a class -> get_actors_by_class. enumerate them AND read their properties, filtered by class or editor label, drilling into one component, with the answering world STAMPED on the result (this tool cannot tell you which world answered) -> get_actor_properties_by_class. read properties off ONE known object -> get_properties / get_property_value. list an actor's components -> get_actor_components. one line trace -> raycast_scene. sample a property across a whole PIE session ARMED BEFORE BeginPlay -- a window a post-tick callback cannot reach at all -> watch_pie_property. But do NOT read this as "always prefer the specific tool": only a minority of scripts are replaceable end-to-end. Python remains the right answer when the script does what those cannot: several classes in one round trip, walking components AND calling a method, filtering by name substring, correlating two values in the same tick, or anything that mutates. Do not trade one call for twelve. WARNING: log_output IS NOT THE LOG OF THE CALL -- it is Python's own prints only (UE fills it from a delegate broadcast solely by unreal.log/log_warning/log_error), so every UE_LOG the ENGINE emits during your script is absent from it. Read engine_log instead: it is the GLog slice bracketed to this call, Warning+Error by default. Example: build_meta_human can run, return None, write zero assets and report a clean log_output while "LogMetaHumanCoreTechLib: Error: Texture Synthesis has not been initialized" sits in GLog. When a call reports success and NOTHING HAPPENED, engine_log is the first place to look; widen with engine_log_level=Log, or pass log_cursor_before to get_output_log_lines(since_sequence_id=...) for the unfiltered rest.

Returns: { command, python_success, execution_mode, command_result, output, log_output, log_line_count, engine_log, engine_log_count, engine_log_truncated, log_cursor_before }