An MCP server for interacting with the UmgMcp plugin in Unreal Engine 5, providing tools for sensing, understanding, and manipulating UMG assets programmatically with context-aware workflow management.
This MCP server exhibits significant quality gaps across naming, descriptions, and schemas. Of 36 tools, most lack sufficient parameter descriptions and comprehensive input schemas. Tools like `bluecode_read_function` and `bluecode_apply` have complex, partially documented parameter structures. Many compatibility tools (tools 29-36) are marked deprecated with empty input schemas. Descriptions range from adequate (bluecode_* tools) to minimal (animation tools). The server shows domain expertise in UMG/Blueprint concepts but fails to apply agent-design principles: tool names are not consistently verb-led, many parameters are undescribed, and output schemas are not visible in the source code. Tools like `set_animation_scope` and `set_widget_scope` have no input parameters documented despite likely accepting configuration. The codebase shows testing infrastructure but does not evidence LLM-optimized definitions.
Append time slice to animation
Append widget tracks to animation
Delete widget keyframes from animation
Get overview of animations in the current scope
Get time properties for animation
Get widget properties for animation
Apply blueprint code with support for node creation, property modification, and connection hints. Supports 'append' and 'union' modes with optional dry-run validation.
Four compatibility tools (set_property_keys, set_animation_data, remove_property_track, remove_keys) have empty input schemas and minimal descriptions marked 'deprecated'. These tools are non-functional from an LLM perspective and should either be removed or fully documented.
Tools `set_animation_scope` and `set_widget_scope` claim to set scope with no input parameters documented. It is unclear what configuration they accept or require. Are these no-op context setters? Do they accept target names?
The `bluecode_apply` tool accepts 13 parameters with complex nested structures (source_map, member_classes, action_handles, action_hints, expression_hints, action_hints_by_line, node_aliases, node_properties). These are inadequately documented. Each object parameter needs field-level schema and usage guidance. LLMs cannot infer the structure or purpose of `source_map` or the difference between `action_handles` and `action_hints`.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 54 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 49 | 1.4.1+ | v1 |
Apply variable modifications to the blueprint function
Compile the current blueprint function
Connect blueprint pins using source->target or shorthand syntax. Supports node aliases for flexible node referencing.
Delete blueprint elements (nodes, variables, connections) with confirmation requirement.
Read blueprint function events
Read the current blueprint function code with semantic analysis. Supports 'debug' and 'roundtrip' detail levels with optional connection information.
Read blueprint function variables
Search for blueprint nodes with filtering by query, category, class, and pins. Supports context-sensitive and exact matching.
Set the target blueprint function
Connect to a specific UmgMcp server endpoint
Create a new animation
Delete an animation with confirmation
Get all animations in the current scope
Get animated widgets from animation (compatibility tool, deprecated)
Get full animation data (compatibility tool, deprecated)
Get keyframes from an animation (compatibility tool, deprecated)
Get the current target UMG asset
Get the current target widget
Get widget animation data (compatibility tool, deprecated)
List available UmgMcp servers and their endpoints
Remove keyframes (compatibility tool, deprecated)
Remove property track (compatibility tool, deprecated)
Save the current target asset
Set animation data (compatibility tool, deprecated)
Set the scope for animation operations
Set property keyframes (compatibility tool, deprecated)
Set the target UMG asset for subsequent operations
Set the target widget for subsequent operations
Set the widget scope for animation operations
No output schemas are visible in the source code for any tool. LLMs need to know what fields and types are returned to plan downstream calls and extract data. For example, does `bluecode_read_function` return `action_hints` as a dict keyed by statement string? Does `get_all_animations` return an array or object structure?
Descriptions for simple animation tools (animation_overview, animation_widget_properties, animation_time_properties, get_animation_keyframes, get_animated_widgets) are under 50 characters and lack context for when an LLM should select them. 'Get overview of animations in the current scope' does not explain what 'overview' contains or why to call this before other queries.
Parameters in complex tools like `bluecode_connect` are under-described. The `connects` parameter is described as 'List of connections to create' but provides no schema for what each connection object must contain. The `source` and `target` parameters exist alongside a `connects` array, are these parameters mutually exclusive? Are they alternatives to each other?
Animation tools that accept array parameters (animation_append_widget_tracks, animation_append_time_slice) lack documentation of the expected structure of array elements. What fields does a widget track object require? What is a time_slice and what properties does it have?
No error handling guidance is visible. Tools like `bluecode_apply` with dry_run support and `bluecode_delete` with confirm_delete flags suggest mutation hazards, but there is no documentation of failure modes, recovery steps, or actionable error messages. What does the LLM do if `bluecode_compile` fails? Can it retry? Should it call a diagnostic tool?