Tools Reference

What the server can do, and how it's organized. You never need this page to use it, because search finds capabilities from plain words. It's here to show what exists and where.

The catalog holds 389 capabilities (at the time of writing), organized three ways:

	Layer	Example	Used for
🧰	Internal tools: 23 of them	manage_asset, control_actor	Routing requests to the plugin. Old {tool, action} names still resolve through them.
🗂️	Domains and families: 40 domains	actor, widget, material	How describe browses the catalog
🔹	Capabilities: 389	control_actor.spawn, texture.create_texture	What you search for, describe and execute

Tip

The full list, with each capability's effect, required scope and consent, is the generated Action Reference. It is rebuilt from the same records the gateway validates against, so it is never out of date.

mindmap
  root((unreal))
    Core
      manage_asset
      manage_blueprint
      control_actor
      control_editor
      manage_level
      system_control
      inspect
      manage_tools
    World
      build_environment
      manage_level_structure
      manage_geometry
      manage_pcg
    Gameplay
      animation_physics
      manage_character
      manage_combat
      manage_effect
      manage_gas
      manage_ai
      manage_inventory
      manage_interaction
    Utility
      manage_audio
      manage_sequence
      manage_networking

Loading

🧱 Core

Tool	#	Covers	Examples
manage_asset	46	Assets and folders: import, move, rename, duplicate, delete. Materials, material graphs and instances; procedural textures and render targets; data tables, structs, enums; metadata, dependencies, source control, Fab.	import move create_material add_material_node create_material_instance create_data_table source_control_checkout
manage_blueprint	28	Blueprints, components, variables, functions, event graphs (node by node or a whole graph in one batch), compiling. UMG widgets: layout, fonts, bindings, animations, previews.	create add_variable add_scs_component build_graph edit_graph compile create_widget_blueprint bind_text preview_widget
control_actor	24	Spawning (single or batch), transforms, attachment, components, materials, tags, collision, visibility, calling actor functions, placement audits	spawn set_transform attach add_component set_material find_by_tag audit_placement call_actor_function
control_editor	21	Play-In-Editor, synthetic input, game speed, screenshots, viewport camera and view modes, bookmarks, opening assets, saving, undo and redo, editor preferences, restarting the editor	play simulate_input set_game_speed screenshot set_camera open_asset save_all undo restart_editor
manage_level	17	Create, load, save, duplicate, import and export levels; sublevels and streaming; world settings; lighting builds; validation	load save create_level add_sublevel stream get_summary validate_level
system_control	25	Console commands, logs, project settings, plugins, profiling and Insights, scalability, builds and packaging, tests, asset validation, Python	console_command read_log set_project_setting profile_performance run_build package_project execute_python
inspect	16	Read and write any UObject's properties; components, class info, Blueprint details, editor state, runtime reports	inspect_object get_property set_property inspect_class get_components get_editor_state
manage_tools	8	Which internal tools are enabled. Reached through configure, not execute.	get_status enable_tools disable_category reset

🌍 World

Tool	#	Covers	Examples
build_environment	40	Landscapes (create, sculpt, paint, heightmaps), foliage, lights and sky, atmosphere, shadows and Lightmass, water, weather, splines, procedural terrain, scene captures	create_landscape sculpt_landscape add_foliage create_light configure_sky_atmosphere create_spline_actor create_water_body_ocean
manage_level_structure	8	Levels and sublevels, World Partition, streaming, data layers, HLOD, level instances; volumes (trigger, blocking, physics, audio, post process, nav bounds)	create_level configure_level_streaming create_volume set_volume_properties
manage_geometry	15	Geometry Script meshes: primitives, booleans, deformers, UVs, simplification, collision, LODs, Nanite and static-mesh conversion	create_box boolean_union extrude bend auto_uv generate_collision convert_to_static_mesh
manage_pcg	3	PCG graphs: create, add and connect nodes, configure, execute	create_pcg_graph add_pcg_node execute_pcg_graph

🕹️ Gameplay

Tool	#	Covers	Examples
animation_physics	26	Animation Blueprints and state machines, blend spaces, montages, skeletons and sockets, retargeting, Control Rig and IK, physics assets, ragdolls, cloth, vehicles	create_animation_blueprint add_blend_sample add_montage_notify create_control_rig setup_ik setup_ragdoll configure_vehicle
manage_character	10	Character Blueprints, movement and locomotion; MetaHuman creation and export where the MetaHuman plugin is available	create_character_blueprint setup_movement configure_movement_speeds create_metahuman
manage_combat	5	Weapon Blueprints, meshes, projectiles, hit detection	create_weapon_blueprint configure_projectile configure_hit_detection
manage_effect	13	Niagara systems, emitters and modules; spawning and simulating effects; debug shapes	create_niagara_system add_emitter_to_system add_niagara_module spawn_niagara debug_shape
manage_gas	6	Gameplay Ability System: abilities, attribute sets, effects, tags, ability system components	create_gameplay_ability add_attribute set_effect_duration configure_asc
manage_ai	16	AI controllers, Behavior Trees and Blackboards, EQS, State Trees, Smart Objects, perception, NavMesh and nav links, Mass entity configs	create_behavior_tree add_blackboard_key create_eqs_query create_state_tree setup_perception
manage_inventory	6	Item data assets, loot tables, crafting recipes, inventory replication	create_item_data_asset add_loot_entry add_recipe_ingredient
manage_interaction	3	Doors and other interactable actors	create_door_actor configure_door_properties

🎛️ Utility

Tool	#	Covers	Examples
manage_audio	13	Playing, stopping and fading sounds; audio actors and components; Sound Cues and MetaSounds (read and edit their graphs); attenuation, sound classes, mixes	play_sound stop_sound create_audio_asset edit_metasound edit_sound_cue configure_sound_attenuation
manage_sequence	19	Level Sequences, bindings, tracks and keys, cameras; Movie Render Queue; media playback; Take Recorder; replays	create edit_sequence_bindings edit_sequence_tracks play
manage_networking	21	Replication, RPCs, prediction, sessions and LAN hosting, voice, split-screen players, game framework classes (game mode, player controller, player start), Enhanced Input actions and mappings	configure_replication configure_rpc host_lan_server create_framework_class configure_game_mode configure_input get_input_info

Note

Many capabilities need an engine plugin (Sequencer, Control Rig, Geometry Scripting, PCG, MetaSound, …). 📦 Installation lists which, and each capability says so under availability.requiredPlugins.

Folded families

Most capabilities are families: one capability whose selector parameter picks the variant. That keeps the catalog small without losing any operation.

Capability	Selector	Values
control_actor.spawn	spawnKind	class (default) · blueprint · batch
texture.create_texture	kind	gradient · noise · pattern · ao_from_mesh · normal_from_height · resized · channel_pack · channel_extract · combined

• describe shows the selector with its allowed values. A parameter only some variants read carries variants, naming them.
• Old names keep working. Before folding, create_noise_texture was its own action. { "tool": "manage_asset", "action": "create_noise_texture" } still describes and executes, and it fills in kind: "noise" for you. A value you pass yourself is never overridden.
• Former capability ids are aliases of the family, so a consent grant that names the old id is accepted too.

Effects, scopes and consent

Every capability declares three things:

Property	Values	Meaning
effect	read · write · destructive	What it does to your project
scope	read · write · destructive · admin	What the caller's token must hold
consent	none · explicit · elevated	Whether each call needs a consent grant

62 capabilities need consent: every destructive one, plus 33 writes. How scopes and consent work is on 🔐 Security.

Ids, aliases and legacy names

A capability answers to several names:

Name	Example	Accepted by
{tool, action} pair: what search rows and nextCall use	{ "tool": "control_actor", "action": "spawn" }	Both routes
Legacy pair: a pre-0.6 or pre-fold action name	{ "tool": "manage_asset", "action": "create_noise_texture" }	Both routes
Canonical id	"capability": "control_actor.spawn"	The stdio route
Alias: lets a caller's verb find the right capability when the action name uses another word	declared per capability	Search on both routes; as an id on stdio

If an alias ever matched two capabilities, the call would be refused with ALIAS_CONFLICT rather than guessed.

---

← 🧭 Using the Gateway  ·  🏠 Home  ·  📚 Resources and Prompts →