UnrealMCP — AI Bridge for Unreal Engine 5

Add a node to a Blueprint graph. Node types by category: FLOW: Branch, Comparison, Switch, SwitchEnum, SwitchInteger, ExecutionSequence DATA: VariableGet, VariableSet, MakeArray CASTING: DynamicCast, ClassDynamicCast, CastByteToEnum UTILITY: Print, CallFunction, Select, SpawnActor SPECIAL: Timeline, GetDataTableRow, AddComponentByClass, Self, Knot EVENT: Event (specify event_type: BeginPlay, Tick, Destroyed, etc.) Extra args are only needed for specific node types (message for Print, variable_name for Variable nodes, target_function for CallFunction, etc.).

Add a component to a Blueprint's default components. Args: blueprint_path: Full path to blueprint (e.g. "/Game/Blueprints/BP_MyActor") component_class: e.g. "StaticMeshComponent", "PointLightComponent" component_name: Optional custom name for the component

Add a new row, optionally setting initial field values. data is a JSON string mapping field names to values, e.g. '{"FieldA": 1}'.

Add an event node (ReceiveBeginPlay, ReceiveTick, ReceiveDestroyed, etc.).

Add an input parameter to a Blueprint function.

Add an output parameter to a Blueprint function.

Add a modifier to an Input Action. Args: asset_path: Path to the InputAction asset modifier_type: Class name — InputModifierDeadZone | InputModifierScalar | InputModifierNegate | InputModifierSwizzleAxis | InputModifierSmooth | InputModifierSmoothDelta | InputModifierScaleByDeltaTime | InputModifierResponseCurveExponential | InputModifierResponseCurveUser | InputModifierFOVScaling | InputModifierToWorldSpace properties: Optional modifier properties (e.g. {"LowerThreshold": 0.2, "Type": "Radial"})

Add a trigger to an Input Action. Args: asset_path: Path to the InputAction asset trigger_type: Class name — InputTriggerDown | InputTriggerPressed | InputTriggerReleased | InputTriggerHold | InputTriggerHoldAndRelease | InputTriggerTap | InputTriggerRepeatedTap | InputTriggerPulse | InputTriggerChordAction | InputTriggerCombo properties: Optional trigger properties (e.g. {"HoldTimeThreshold": 0.5, "bIsOneShot": true})

Add a key-to-action mapping to an Input Mapping Context. Args: context_path: Path to the IMC asset action_path: Path to the InputAction asset key: Key name (e.g. "SpaceBar", "W", "Gamepad_LeftY", "LeftMouseButton") triggers: Optional list of triggers, each {"type": "InputTriggerHold", "properties": {...}} modifiers: Optional list of modifiers, each {"type": "InputModifierNegate", "properties": {...}}

Add a modifier to a specific mapping in an Input Mapping Context. Args: context_path: Path to the IMC asset mapping_index: Index of the mapping (from get_input_mapping_context) modifier_type: Modifier class name (e.g. "InputModifierNegate") properties: Optional modifier properties

Add a trigger to a specific mapping in an Input Mapping Context. Args: context_path: Path to the IMC asset mapping_index: Index of the mapping (from get_input_mapping_context) trigger_type: Trigger class name (e.g. "InputTriggerPressed") properties: Optional trigger properties

Add a SINGLE trait to a Mass Entity Config without replacing existing traits. Safe for configs with complex replicator settings — only appends, never replaces. Returns error if the trait already exists on the config. Args: asset_path: Path to MassEntityConfigAsset (e.g. "/Game/.../DA_MinerMassEntityConfig_Mk1") trait_class: Full class path (e.g. "/Script/Jiggify.MassBuildingHealthTrait") properties: Optional dict of property values to set on the new trait. Supports floats, ints, bools, strings. Example: {"MaxHealth": 150, "Resistance_Fire": 0.2}

Add comment boxes to a material graph for organization. comments is a JSON array. Each entry: {text, pos_x, pos_y, size_x, size_y, font_size, color: [r,g,b], show_bubble, color_bubble, group_mode}.

Add a single expression node without clearing other nodes. Returns the new node_index. Wire it up with connect_material_expressions. node is a JSON object: {"type": "ScalarParameter", "pos_x": -1200, "pos_y": -400, "properties": {"parameter_name": "Speed", "default_value": 0.5}} For Custom HLSL nodes, include top-level "code", "description", "output_type", "inputs", and "outputs" fields.

Add an input pin to a Material Function. Args: function_path: Full path to existing material function input_name: Pin name (e.g. "BaseColor", "Strength", "UV") input_type: Scalar | Vector2 | Vector3 | Vector4 | Texture2D | TextureCube | Texture2DArray | VolumeTexture | StaticBool | MaterialAttributes | TextureExternal | Bool | Substrate description: Tooltip text sort_priority: Controls display order (lower = higher) use_preview_as_default: Use preview_value as the default when input is unconnected preview_value: JSON array for default value (e.g. "[0.5]", "[1.0, 0.0, 0.0, 1.0]") pos_x: Graph X position (default -600) pos_y: Graph Y position (default 0)

Add an output pin to a Material Function. Args: function_path: Full path to existing material function output_name: Pin name (e.g. "Result", "Normal", "Mask") description: Tooltip text sort_priority: Controls display order (lower = higher) pos_x: Graph X position (default 200) pos_y: Graph Y position (default 0)

Add a NiagaraComponent to an existing actor in the level. Args: actor_name: Name/label of the target actor system_path: Path to the Niagara System asset to assign component_name: Optional name for the component relative_location: Relative offset as "(X,Y,Z)" auto_activate: If true, auto-activate on begin play

Add a typed input pin to a scratch pad module's Custom HLSL node. Types: float, int, bool, vec2/3/4, color, quat, matrix, position, ParameterMap, or any registered Niagara type. The pin's name is usable inside the HLSL source.

Add a typed output pin to a scratch pad module's Custom HLSL node.

Add an emitter to a Niagara System. Provide either emitter_path (existing emitter asset) or template (built-in template name). Args: system_path: Path to the Niagara System asset emitter_path: Path to an existing NiagaraEmitter asset to add emitter_name: Optional display name for the emitter in this system template: Built-in template name (e.g. "Fountain", "Sprite", "Mesh")

Add an event handler to an emitter that responds to events from a source emitter. Args: system_path: Path to the Niagara System asset emitter_name: Name of the receiving emitter source_emitter: Name of the emitter producing the event event_name: Event name (e.g. "CollisionEvent", "DeathEvent", "LocationEvent") execution_mode: How events are handled — "every_particle", "spawn_particles", "single_particle" spawn_number: Number of particles to spawn per event (for spawn_particles mode) max_events_per_frame: Limit on events processed per frame (0 = unlimited)

Create a new node inside a Niagara scratch-pad / dynamic-input / module graph. Mirrors the node-creation pattern used by the Niagara editor schema actions (FGraphNodeCreator + AllocateDefaultPins + NotifyGraphChanged). Spawns on both asset and edit-copy graphs when a system is open. Args: node_type: One of "Op", "FunctionCall", "DataInterfaceFunction", "ParameterMapGet", "ParameterMapSet", "Reroute", "Input" pos_x, pos_y: Graph position op_name: (node_type=Op) — discover via get_niagara_schema_actions (e.g. "Numeric::Add", "Numeric::Mul", "Numeric::Length") function_script: (node_type=FunctionCall) UNiagaraScript asset path input_name + input_type: (node_type=Input) variable name + Niagara type di_class + function_name: (node_type=DataInterfaceFunction) — wraps a data-interface member function. Discover via list_niagara_data_interface_functions. Examples: di_class="NiagaraDataInterfaceArrayPosition", function_name="Length" Use list_niagara_script_parameters after to verify the pin layout.

Add a typed output pin to the first ParameterMapGet node in a scratch pad graph. Mirrors clicking "+" on a Map Get node in the editor and picking a variable. The pin name becomes the parameter handle (e.g. "Engine.DeltaTime", "User.MyParam", "Particles.Velocity"). Args: system_path: Niagara System asset path module_name: Scratch pad module name parameter_name: Variable name including namespace (e.g. "Engine.DeltaTime") parameter_type: Niagara type (float, int, bool, vec2/3/4, color, quat, matrix, position, ...)

Add a typed input pin to the first ParameterMapSet node in a scratch pad graph. Mirrors clicking "+" on a Map Set node in the editor. Args: system_path: Niagara System asset path module_name: Scratch pad module name parameter_name: Variable name including namespace parameter_type: Niagara type (see add_niagara_map_get_pin)

Add a module to an emitter's script usage stack. Args: system_path: Path to the Niagara System asset emitter_name: Name of the target emitter module_path: Path to the module script asset (e.g. "/Niagara/Modules/Update/SolveForcesAndVelocity") script_usage: Target stack — EmitterSpawn, EmitterUpdate, ParticleSpawn, ParticleUpdate, SystemSpawn, SystemUpdate index: Position in the stack (-1 = append at end)

Generic: add a typed dynamic pin to any node in a scratch pad graph. Works with Custom HLSL, Map Get, Map Set, or any node inheriting from UNiagaraNodeWithDynamicPins. Identify the target node via exactly one of node_class / node_index / node_id. Args: system_path: Niagara System asset path module_name: Scratch pad module name pin_name: New pin name pin_type: Niagara type string (float, vec3, color, etc.) direction: "input" or "output" node_class: Class short name (e.g. "NiagaraNodeParameterMapGet", "ParameterMapGet") node_index: Raw Graph->Nodes index node_id: UEdGraphNode::NodeGuid string

Add a renderer to an emitter. Args: system_path: Path to the Niagara System asset emitter_name: Name of the target emitter renderer_type: Renderer class — SpriteRenderer, MeshRenderer, RibbonRenderer, LightRenderer, ComponentRenderer material_path: Optional material asset path for sprite/mesh/ribbon renderers mesh_path: Optional static mesh path for MeshRenderer

Add an input or output parameter to a Niagara script. For a dynamic input, "output" adds a new entry to the Output Dynamic Input node (screenshot 5's "Add Parameter" dropdown). Type must be a name registered in FNiagaraTypeRegistry — use list_niagara_parameter_types to discover valid options (e.g. "float", "Vector", "LinearColor", "Quat", "Position", "NiagaraID", "Matrix"). Args: name: Parameter name (input names without namespace prefix become Module.<name>) type: Niagara type name direction: "output" (default — appends to output node) | "input"

Add a simulation stage to an emitter. Simulation stages allow additional processing passes over particles or data interfaces. Args: system_path: Path to the Niagara System asset emitter_name: Name of the target emitter stage_name: Display name for the simulation stage iteration_source: What to iterate over — "particles" or a data interface name num_iterations: Number of times to execute this stage per frame

Add a user parameter to a Niagara System. Supports every type the editor's User Parameters "+" menu can add — primitives, structs, enums, and Data Interfaces. Use list_niagara_parameter_types to discover valid type names before calling. Args: system_path: Path to the Niagara System asset parameter_name: Name for the new parameter (auto-prefixed with "User.") parameter_type: Type name. Resolved via NiagaraTypeRegistry, so anything Niagara registered is valid. Examples: - Primitives: "float", "int", "bool", "Vector2", "Vector", "Vector4", "LinearColor", "Quat", "Position", "Matrix" - Niagara structs: "NiagaraID", "NiagaraRandInfo", "NiagaraSpawnInfo" - Enums: "ENiagaraExecutionState", any registered UEnum path - Data Interfaces (bare or prefixed): "ArrayFloat", "ArrayVector", "ArrayVector4", "ArrayInt32", "Curve", "Vector2DCurve", "VectorCurve", "Vector4Curve", "CurlNoise", "Grid2D", "Grid3D", "SkeletalMesh", "StaticMesh", "Texture", "TextureCube", "VolumeTexture", "Texture2DArray", "RenderTarget2D", "SplineComponent", "Camera", "Collision", "PhysicsAsset", "Spline", and any other NiagaraDataInterface subclass including plugin-provided ones. default_value: Optional default. For scalars pass as string or number; for vector/color/struct pass UE-standard text form like "(X=1,Y=2,Z=3)" or "(R=1,G=0.5,B=0,A=1)". Data interfaces use the CDO defaults — use set_niagara_user_parameter afterwards for richer initialization. Returns the created parameter with its resolved type name, kind (primitive / struct / enum / data_interface / object), size_bytes, and (for DIs) class_path.

Add a property binding connecting a source node's output to a target node's input. Args: asset_path: Content path to the StateTree asset source_node_guid: GUID of the source node source_property: Property path on source (e.g. "TargetLocation") target_node_guid: GUID of the target node target_property: Property path on target (e.g. "TargetLocation")

Add a condition to a state's enter conditions or a transition's conditions. Args: asset_path: Content path to the StateTree asset condition_class: Condition struct name (e.g. "StateTreeCompareIntCondition") state: State index path state_name: State name state_guid: State GUID target: "enter_conditions" or "transition" transition_index: Required when target="transition" operand: "And" or "Or" (how to combine with previous conditions) properties: JSON object for condition properties instance_properties: JSON object for condition instance data

Add a global evaluator to the StateTree. Args: asset_path: Content path to the StateTree asset evaluator_class: Evaluator struct name (e.g. "MassEnemyHasTargetEvaluator") properties: JSON object for evaluator node properties instance_properties: JSON object for evaluator instance data properties

Add a global task to the StateTree. Args: asset_path: Content path to the StateTree asset task_class: Task struct name properties: JSON object for task node properties instance_properties: JSON object for task instance data properties

Add a tree-level parameter. Args: asset_path: Content path to the StateTree asset name: Parameter name type: Property type ("bool", "int32", "float", "FVector", "FName", "FString", etc.)

Add a state to the tree. Omit parent params for a root-level state. Args: asset_path: Content path to the StateTree asset name: Name for the new state parent: Parent state index path (e.g. "0/1") parent_name: Parent state name parent_guid: Parent state GUID state_type: "State", "Group", "Linked", "LinkedAsset", or "Subtree" selection_behavior: "None", "TryEnterState", "TrySelectChildrenInOrder", "TrySelectChildrenAtRandom", "TrySelectChildrenWithHighestUtility", "TryFollowTransitions" position: Insert position (-1 = append)

Add a task to a state. Args: asset_path: Content path to the StateTree asset task_class: Task struct name (e.g. "MassEnemyAttackTask", "MassNavMeshPathFollowTask") or full path state: State index path state_name: State name state_guid: State GUID properties: JSON object for task node properties instance_properties: JSON object for task instance data properties

Add a transition to a state. Args: asset_path: Content path to the StateTree asset trigger: "OnStateCompleted", "OnStateFailed", "OnStateSucceeded", "OnTick", "OnEvent" state: Source state index path state_name: Source state name state_guid: Source state GUID target_state: Target state index path, or special: "TreeSucceeded", "TreeFailed", "TreeStopped", "NextSibling" target_state_name: Target state name target_state_guid: Target state GUID priority: "Low", "Normal", "Medium", "High", "Critical" delay: Delay in seconds before transitioning delay_variance: Random variance (+/-) on delay event_tag: Gameplay tag for OnEvent trigger

Add a new widget to a Widget Blueprint's Designer tree. Args: widget_class: Short name like "TextBlock", "Image", "Button", "CanvasPanel", "VerticalBox", "HorizontalBox", "Overlay", "SizeBox", "Border", "ScrollBox", "ProgressBar", etc. parent_widget_name: Name of the parent panel (e.g. "CanvasPanel_0") widget_name: Optional name (auto-generated if empty) index: Insert position among siblings (-1 = append) slot_properties: Canvas anchors, offsets, etc. widget_properties: Widget values like {"Text": "Hello"}

Analyze a specific graph — nodes, connections, and execution flow.

Apply a scratch pad module's edit-copy AND save the asset (Apply & Save button). Mirrors FNiagaraScratchPadScriptViewModel::ApplyChangesAndSave.

Apply a material to a level actor.

Apply a material to a Blueprint component.

Commit a scratch pad module's edit-copy to the original script (Apply button). Mirrors FNiagaraScratchPadScriptViewModel::ApplyChanges. Requires the Niagara System to be open in the editor so the view model exists.

Build the internal node graph of a Material Function in one atomic call. Same interface as build_material_graph but for Material Functions. FunctionInput nodes use type="FunctionInput" with extra fields: input_name, input_type (Scalar/Vector2/Vector3/Vector4/Texture2D/TextureCube/ Texture2DArray/VolumeTexture/StaticBool/MaterialAttributes/TextureExternal/Bool/Substrate), description, sort_priority, use_preview_as_default, preview_value ([x,y,z,w]) FunctionOutput nodes use type="FunctionOutput" with extra fields: output_name, description, sort_priority Connections between nodes use the same format as build_material_graph: from_node (int), from_pin (str), to_node (int), to_pin (str) To connect a node to a FunctionOutput's input, use to_pin="" (the output has a single unnamed input pin labeled "A" internally). Args: function_path: Full path to existing material function nodes: JSON array of node definitions connections: JSON array of connections clear_existing: Remove existing expressions before building (default True)

Build a complete material node graph in one atomic operation. Creates expression nodes and wires them together in a single call. By default clears existing expressions first (safe rebuild — all external references stay intact). Args: material_path: Full path to existing material (e.g. "/Game/Materials/M_Portal") nodes: JSON array of node definitions. Each node has: - type: Short class name (e.g. "TextureCoordinate", "Custom", "ScalarParameter", "Constant3Vector", "TextureSample", "Multiply", "Add", "Panner", "Time", "VectorParameter", "LinearInterpolate", "ComponentMask") Auto-prefixed with "MaterialExpression" if needed. - pos_x, pos_y: Graph position (default -300, 0) - properties: Dict of editor properties to set. Common ones: - parameter_name: str — name shown in material instances - default_value: number or [r,g,b,a] - slider_min, slider_max: float — slider range - group: str — parameter group name - Values starting with "/" are loaded as assets - Lists of 3-4 become LinearColor, lists of 2 become Vector2D For Custom HLSL nodes (type="Custom"), also supports: - code: HLSL source code string - description: Node title in the graph - output_type: "float" | "float2" | "float3" | "float4" - inputs: List of input pin names (e.g. ["UV", "Speed"]) - outputs: List of {"name": str, "type": str} for additional outputs connections: JSON array of connections. Each connection has: - from_node: Source node index (int, 0-based into nodes array) - from_pin: Output pin name ("" for default output) - to_node: Target node index (int) 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 (FrontMaterial is for Substrate BSDF nodes — use get_available_material_pins to see which pins are active for a specific material) clear_existing: Remove existing expressions before building (default True)

Remove unconnected/orphaned nodes from a Material Function. Deletes: - FunctionOutput nodes with no input wired (duplicate output pins) - FunctionInput nodes whose output nobody consumes (duplicate input pins) - Orphaned internal nodes WARNING: Only call when the user explicitly asks to clean up. Args: dry_run: If true, reports what would be deleted without deleting.

Delete orphaned and/or dead-end nodes from a material graph. WARNING: NEVER call this tool unless the user explicitly asks to clean up or delete unconnected nodes. Users may intentionally keep disconnected nodes as references, scratchpads, or work-in-progress. Deleting them without permission will cause data loss and frustration. Args: mode: What to delete: "orphaned" — only nodes with NO connections at all (safest) "dead_ends" — nodes whose output goes nowhere (has inputs but no consumers) "all" — both orphaned and dead-ends dry_run: If true, reports what WOULD be deleted without actually deleting. Always use dry_run=true first to preview before deleting.

Reset a module input to its default — equivalent to "Reset to Default" in the stack UI. Finds the override pin, breaks its connections, removes any orphan nodes that were exclusively feeding it (dynamic input node, map get, custom HLSL, etc.), then removes the override pin itself from the ParameterMapSet so the input reverts cleanly. Supports nested path syntax: input_name="Spawn Count.Position Array" clears the Position Array binding INSIDE the dynamic input attached to Spawn Count — without touching Spawn Count's own binding. Mirrors FNiagaraStackGraphUtilities::RemoveNodesForStackFunctionInputOverridePin (non-exported — replicated here).

Compile a Blueprint.