MCP server that bridges Unreal Engine editor with language models via socket-based communication, enabling programmatic control of actors, materials, blueprints, and Niagara particle systems
This MCP server exhibits significant structural quality issues across naming, descriptions, and schema completeness. While tool names follow verb_noun convention and all 11 tools have basic descriptions, the descriptions are generic and lack LLM-optimization guidance. Parameter descriptions are minimal or absent in many tools. No output schemas are documented. Error handling exists but is minimal. The server targets Unreal Engine editor automation, a narrow domain, but lacks the polish and clarity expected of production-grade tools. Average tool quality: 48/100.
Delete all static mesh actors in the scene
Search for assets by name (case-insensitive substring) across both /Game (project content) and /Engine (engine assets). Returns a JSON list of matches: [{"asset_class": ..., "asset_path": ...}, ...]. Common substrings (e.g. "Material") can return many results — prefer a more specific fragment when possible. Project (/Game) hits are listed first.
Search for basic shapes for building
Get details for a specific actor by name.
List all actors in the current level
Get the content directory
No output schemas documented for any tool. LLMs cannot predict response structure or plan downstream tool calls. All 11 tools lack documented return types.
Tool descriptions are generic and lack context. 'List all actors in the current level' (get_actors) does not explain what fields are returned, when to call this vs get_selected_actors, or what 'actor' means to an unfamiliar user. Descriptions should be 50-200 chars and answer WHAT, WHEN, and WHAT IT RETURNS.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 51 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 29 | - | v1 |
Get the top level project directory
Get the currently selected actors in the editor
Modify a property of an existing actor.
Apply a material to a static mesh actor.
Spawn an actor in the current level.
spawn_actor has 9 numeric parameters (location_x, location_y, location_z, rotation_x, rotation_y, rotation_z, scale_x, scale_y, scale_z) with no range constraints, no defaults, and no explanation of coordinate systems or units. LLMs will guess arbitrary values, causing invalid spawns.
delete_all_static_mesh_actors is a destructive operation with no confirmation step, dry-run option, or permission gate. An LLM can irreversibly delete all mesh actors in one call with no safeguard.
modify_actor accepts a generic property_name string with no validation or enumeration. LLMs will pass invalid property names. Code shows special handling for 'actorlabel', 'hidden', 'location', 'rotation', 'scale', but these are not documented as the only supported properties, leaving LLMs to guess.
Error messages in source code (e.g. 'Actor not found', 'Failed to create actor') do not guide recovery. No mention of what the LLM should do next (retry, call a discovery tool, ask the user).
find_basic_shapes has no input parameters. Description says 'Search for basic shapes' but does not explain what search criteria apply (by name? by category?). Result structure is unknown.
No tool chains documented. If spawn_actor creates an actor, does its response include the actor name/ID so the LLM can immediately call modify_actor or set_material? Source shows partial JSON responses but no field-level documentation.