UnrealMCP — AI Bridge for Unreal Engine 5

Compile a Niagara System and report any errors. Args: system_path: Path to the Niagara System asset wait_for_completion: If true, block until compilation finishes

Compile a StateTree and return errors/warnings. Args: asset_path: Content path to the StateTree asset

Connect two nodes in a Blueprint graph. Args: source_node_id: GUID of the source node source_pin_name: Output pin name on source target_node_id: GUID of the target node target_pin_name: Input pin name on target function_name: Function graph (empty = EventGraph)

Connect two expression nodes using their indices. Args: from_node: Source node index to_node: Target node index (e.g. "5") or "material" for material output to_pin: Input pin name, or material property when to_node="material": BaseColor, Metallic, Specular, Roughness, Anisotropy, EmissiveColor, Opacity, OpacityMask, Normal, Tangent, WorldPositionOffset, Displacement, SubsurfaceColor, CustomData0, CustomData1, AmbientOcclusion, Refraction, PixelDepthOffset, ShadingModel, SurfaceThickness, FrontMaterial from_pin: Output pin name ("" = primary output)

Wire one node's output pin to another node's input pin inside a scratch pad graph. Connection is validated via UEdGraphSchema_Niagara::TryCreateConnection so type mismatches are rejected cleanly. Identify each side via the (class | index | id) trio.

Control a Niagara effect actor in the level. Args: actor_name: Name/label of the Niagara actor action: Control action — "activate", "deactivate", "reset", "destroy"

Create a new Blueprint asset from any C++ or Blueprint parent class. Supports parent classes from ANY module (Engine, Game, plugins, etc.). Use search_parent_classes first to find the correct name. parent_class accepts short names ("MinerActor"), prefixed ("AMinerActor"), full paths ("/Script/Jiggify.AMinerActor"), or BP paths ("/Game/Blueprints/BP_Base").

Create a new function in a Blueprint.

Create a variable in a Blueprint. variable_type: bool, int, float, string, vector, rotator

Create a new data asset of any UDataAsset subclass. Args: class_name: Short name ("ItemData") or full path ("/Script/Jiggify.UItemData") asset_path: Full content path including name (e.g. "/Game/Data/Items/DA_IronOre") initial_properties: Optional dict of property values to set after creation. Supports primitives, structs, arrays, object paths, gameplay tags, etc.

Create a new UInputAction asset. Args: asset_path: Full content path including name (e.g. "/Game/Input/IA_Jump") value_type: Boolean | Axis1D | Axis2D | Axis3D properties: Optional dict of additional properties to set after creation

Create a new UInputMappingContext asset. Args: asset_path: Full content path including name (e.g. "/Game/Input/IMC_Default") description: Optional localized description

Create a new Material asset. Returns error if the asset already exists (avoids editor overwrite popup). Pass force=true to delete and recreate. Args: name: Material name (e.g. "M_MyMaterial") path: Content Browser path blend_mode: opaque | masked | translucent | additive | modulate | alpha_composite | alpha_holdout shading_model: default_lit | unlit | subsurface | clear_coat | subsurface_profile | two_sided_foliage | cloth | eye | thin_translucent two_sided: Render on both sides opacity_mask_clip_value: Clip value for masked blend mode (0.0-1.0) force: If true, delete existing asset and recreate (default false)

Create a new Material Function asset. Returns error if asset already exists. Pass force=true to delete and recreate. Args: name: Function name (e.g. "MF_CustomBlend") path: Content Browser path description: Function description shown in tooltips expose_to_library: Whether the function appears in the material editor's function library force: If true, delete existing asset and recreate (default false)

Create a Material Instance from a parent material. Returns error if asset already exists. Pass force=true to delete and recreate. Args: parent_path: Full path to parent (e.g. "/Game/Materials/M_Base") name: Instance name (e.g. "MI_Red") scalar_params: JSON dict of scalar overrides — {"Opacity": 0.5, "Metallic": 1.0} vector_params: JSON dict of vector overrides — {"BaseColor": [1.0, 0.0, 0.0, 1.0]} texture_params: JSON dict of texture overrides — {"Texture": "/Game/Textures/T_Wood"} force: If true, delete existing asset and recreate (default false)

Create a standalone Niagara script asset (module / dynamic input / function). Duplicates the editor's default template when available so the produced asset has the same initial graph as one created via the Niagara asset wizard. Args: asset_path: Full content path including name (e.g. "/Game/FX/Modules/MyModule") module_type: "module", "dynamic_input", or "function" description: Optional description text

Create a new scratch pad script on a Niagara System. Duplicates the editor's default template (DefaultModuleScript / DefaultDynamicInputScript / DefaultFunctionScript from NiagaraEditorSettings) so the resulting graph is pre-wired with Input / MapGet / MapSet / Output just like clicking "Create New Module" in the Scratch Script Manager. Falls back to a manually-built minimal graph if the template asset is unavailable. Args: system_path: Niagara System asset path module_name: Desired scratch script name (unique within the system's ScratchPadScripts) module_type: "module" (default), "dynamic_input", or "function"

Create a new Niagara System asset. Args: asset_path: Full content path including name (e.g. "/Game/FX/NS_Fire") template: Template to base the system on — "empty" or a path to an existing system/template

Create a new StateTree asset with the specified schema. Args: asset_path: Content path for the new asset (e.g. "/Game/AI/ST_MyTree") schema_class: Schema class name (e.g. "MassStateTreeSchema", "StateTreeComponentSchema", "StateTreeAIComponentSchema")

Delete an actor from the level by name.

Delete an asset or directory from the Content Browser. Checks for references first unless force=True. Paths without a dot are treated as directories.

Delete a function from a Blueprint.

Delete a node from a Blueprint graph by GUID.

Delete a row from a Data Table by name.

Delete a single expression node by index. Recompiles and saves automatically. Remaining node indices may shift — re-query with get_material_graph_nodes afterwards.

Delete a node from a Niagara graph (asset + edit-copy). Provide node_index (ordinal) OR node_id (FGuid string from get_niagara_graph_nodes).

Delete a scratch pad module from a Niagara System.

Delete a Niagara System asset. Args: system_path: Path to the Niagara System asset force: If true, skip reference checks and force deletion

Return the full schema of any Niagara type — primitive, enum, struct, or data interface. Built on a generic FProperty introspector, so the output handles every Unreal property kind recursively (bool / int{8,16,32,64} / uint{16,32,64} / float / double / string / name / text / enum / struct / array / map / set / object / class / soft_object / soft_class / interface / delegate). For enums, returns every entry with name + display_name + value + tooltip. For structs, walks every editable field. For data interfaces, returns the full CDO schema. Args: type: Type identifier — built-in name ("float", "vec3", "Color"), registered Niagara type, UEnum / UScriptStruct path, or UNiagaraDataInterface subclass name (with or without the U prefix).

Disconnect a specific input pin on a material expression node. Args: node_index: The target node whose input pin will be disconnected input_pin: Name of the input pin to disconnect

Break all connections on a specific pin of a node in a scratch pad graph.

Duplicate an asset to a new location.

Copy an existing row under a new name.

Duplicate a node (same type and properties, offset from original). Returns the new node_index. Connections are NOT copied.

Duplicate an emitter within a Niagara System. Args: system_path: Path to the Niagara System asset emitter_name: Name of the emitter to duplicate new_name: Optional name for the duplicated emitter

Duplicate an existing scratch pad module on a system. Args: system_path: Niagara System asset path module_name: Source scratch pad module name new_name: Optional new name (defaults to "<source>_Copy")

Deep-copy a widget including children. Auto-renames on conflicts.

Execute arbitrary Python code inside the running Unreal Editor. The `unreal` module is pre-imported. Set a variable named `result` to return structured data back to Claude Code.

Flexible actor search across the level — combine any number of filters. Use this to answer "is X in the world", "first 5 actors of class Y", "all actors starting with name Z and tagged W". All filters AND together. Returns total count scanned/matched even when results are truncated. Args: name_pattern: Case-insensitive substring on UObject name (e.g. "Light"). label_pattern: Case-insensitive substring on outliner display label. class_filter: Class name (short or full path). When it resolves to a real UClass, uses IsA() (or exact_class for ==). Otherwise falls back to substring match on the actor's class name/path. tag: Match actors with this FName in their Tags. exact_class: When class_filter resolves to a UClass, require exact match instead of subclass. max_results: Cap on returned entries (0 = unlimited; truncated flag set). include_transform: Include location/rotation/scale per entry.

Find actors whose name matches the given pattern.

Search the Asset Registry by class, path, and/or name pattern. Args: class_type: Shortcut ("material", "blueprint", "static_mesh", "texture", "data_table", "niagara_system", etc.) or "PackagePath.ClassName" path: Content path filter (e.g. "/Game/Materials") name_pattern: Substring match on asset name (case-insensitive) recursive: Search subdirectories (default True) max_results: Maximum results to return (default 200)

Reverse lookup: find where a scratch pad script is referenced in the stack. Scans every emitter/system script graph (system_spawn, system_update, emitter_spawn, emitter_update, particle_spawn, particle_update) for UNiagaraNodeFunctionCall nodes whose FunctionScript matches the named scratch pad. Returns each site as {emitter, script_usage, function_name, node_id, is_dynamic_input}. Use this to answer "what uses GetDataInterfaceLength?" without walking every emitter manually.

Find assets that reference or are referenced by the given asset. direction: "dependents" (who uses this), "dependencies" (what this uses), or "both"

Get material slot info for a level actor.

Get property values from a live actor instance placed in the world. Unlike get_blueprint_class_defaults (CDO), this reads the placed instance — per-instance overrides are returned correctly. Two output modes: - Nested (default, flat=False): top-level FProperty names, structs as nested JSON. Filter only matches top-level names — best for quickly inspecting one actor. - Flat (flat=True): every leaf returned as "Settings.BloomIntensity": value with a path-aware filter. Best for AI search ("find me everything containing 'bloom'"). Args: actor_label: Outliner label or UObject name. filter: Case-insensitive substring. Nested mode = top-level only; flat mode = matches full dotted path. include_components: Also return components' properties. flat: Return flat dotted-path dict instead of nested structure. max_depth: (flat mode) Max struct nesting depth. Default 3. include_metadata: (flat mode) Replace each value with full metadata object (type info, clamps, enum values, etc.). expand_arrays: (flat mode) Emit array elements as Field[N] entries. array_element_limit: (flat mode) Max elements emitted per array. Default 16.

Inspect type/clamp/enum metadata for properties on a placed actor. Returns a flat dict keyed by dotted path with per-property metadata (cpp_type, ue_type, category, display_name, tooltip, current_value, clamp_min/max, ui_min/max, valid_values for enums, inner/element/key/value types for containers, object_class / meta_class for refs, plus is_struct/is_array/is_enum/is_object /is_bool / editable / transient / readonly flags). ALSO returns a "_summary" header containing total_available, total_returned, truncated, next_cursor, own_count, inherited_count, class_chain, and a "categories" map (UPROPERTY Category -> count) — use this to navigate without flooding context. max_entries=0 returns ONLY the summary so you can plan first. Args: actor_label: Outliner label or UObject name. property_path: Optional dotted path. Empty = enumerate top-level properties. If the path is a struct, enumerates that struct's fields. If a leaf scalar/enum, returns metadata for just that property. If an OBJECT REFERENCE, returns single-property metadata for the ref + a hint (use component_name to inspect the target's properties), unless descend_into_objects=True is passed. filter: Case-insensitive substring on full dotted path. category: Case-insensitive exact match against UPROPERTY(Category="X"). Sub-categories use "|" — e.g. "Lens|Bloom". depth: How many struct levels to expand below the start point. 0 = flat. expand_enums: Include valid_values list for enum-typed fields. include_inherited: Include properties from super classes. False = only those declared on the most-derived class (use to scope down components). descend_into_objects: When property_path lands on an FObjectProperty, enumerate the target object's class instead of returning a single metadata blob. Disabled by default — prefer component_name. max_entries: Hard cap on emitted entries. Default 50. 0 = summary only.…

List all actors in the current level.

Get asset metadata: class, package, and key properties.

Get all editable properties of an asset.

Query all available material output pins for a specific material. Returns every pin that is currently visible (based on material settings), whether it's connected, what type it expects, and which node feeds it. Use this to discover valid to_pin values for connect_material_expressions with to_node="material". The available pins change based on blend mode, shading model, and whether Substrate is enabled. Also reports substrate_enabled and has_front_material_connected status.