KarianaUMCP / UMCP_TEST_PLAN.md
barlowski's picture
Upload folder using huggingface_hub
5ccfb25 verified

A newer version of the Gradio SDK is available: 6.13.0

Upgrade

UMCP Test Plan for Claude Desktop

Connection: localhost:9877 | PIN: 4336 Server: KarianaUMCP v1.0.0 Test Duration: 15-20 minutes


Pre-Test Checklist

  • Unreal Engine is running
  • KarianaUMCP plugin is loaded
  • Socket server is active on port 9877
  • Claude Desktop is configured to connect to localhost:9877
  • Current level has some space for spawning actors

Test Suite 1: Basic Connectivity (2 min)

Test 1.1: Ping Server

Prompt for Claude:

Test the UMCP server connection with a ping command. Show me the server information.

Expected Response:

  • Server responds with "pong"
  • Server name: KarianaUMCP
  • Version: 1.0.0

Pass Criteria: โœ“ Connection successful, server info returned


Test 1.2: List Available Tools

Prompt for Claude:

List all available UMCP tools and categorize them by functionality.

Expected Response:

  • 29 tools listed
  • Categories: Actors, Assets, Blueprints, Editor, Skills, Python, Screenshot, System

Pass Criteria: โœ“ All 29 tools discovered and categorized


Test Suite 2: Actor Lifecycle (5 min)

Test 2.1: Spawn Actor

Prompt for Claude:

Spawn a cube actor at location (0, 0, 100) and name it "TestCube1".

Expected Response:

  • Actor spawned successfully
  • Name: TestCube1
  • Location confirmed at (0, 0, 100)

Pass Criteria: โœ“ Actor appears in World Outliner


Test 2.2: List Actors

Prompt for Claude:

List all actors in the current level.

Expected Response:

  • Actor list includes "TestCube1"
  • Other actors: DirectionalLight, SkyLight, PlayerStart, etc.

Pass Criteria: โœ“ TestCube1 is in the list


Test 2.3: Move Actor

Prompt for Claude:

Move TestCube1 to location (500, 500, 100).

Expected Response:

  • Location updated successfully
  • New position confirmed

Pass Criteria: โœ“ Actor moved in viewport (verify visually)


Test 2.4: Rotate Actor

Prompt for Claude:

Rotate TestCube1 45 degrees around the Z axis (yaw).

Expected Response:

  • Rotation updated successfully
  • New rotation: [0, 45, 0]

Pass Criteria: โœ“ Actor rotated visually


Test 2.5: Scale Actor

Prompt for Claude:

Scale TestCube1 to 2x its current size in all dimensions.

Expected Response:

  • Scale updated successfully
  • New scale: [2, 2, 2]

Pass Criteria: โœ“ Actor is larger (visual confirmation)


Test 2.6: Delete Actor

Prompt for Claude:

Delete TestCube1.

Expected Response:

  • Actor deleted successfully

Pass Criteria: โœ“ TestCube1 removed from World Outliner


Test Suite 3: Scene Discovery (3 min)

Test 3.1: Current Level Info

Prompt for Claude:

What level am I currently working in? How many actors does it have?

Expected Response:

  • Level name (e.g., "Untitled_1")
  • Actor count (e.g., 9 actors)
  • Level path

Pass Criteria: โœ“ Accurate level information returned


Test 3.2: Asset Discovery

Prompt for Claude:

List all assets in the /Game folder. How many are there?

Expected Response:

  • Asset list with names, paths, and classes
  • Count: ~61 assets
  • Includes: Materials, Meshes, Blueprints, Animations

Pass Criteria: โœ“ Asset list matches Content Browser


Test 3.3: Capture Screenshot

Prompt for Claude:

Capture a screenshot of the current viewport at 1920x1080 resolution.

Expected Response:

  • Screenshot captured
  • Base64 encoded image data returned (or saved to file)

Pass Criteria: โœ“ Screenshot file created or data returned


Test Suite 4: Python Execution (3 min)

Test 4.1: Simple Python

Prompt for Claude:

Execute this Python code in Unreal Engine:
print("Hello from Claude Desktop!")

Expected Response:

  • Code executed successfully
  • Output: "Hello from Claude Desktop!"

Pass Criteria: โœ“ Python output returned


Test 4.2: Query Unreal API

Prompt for Claude:

Use Python to get the Unreal Engine version.

Expected Response:

  • Engine version string (e.g., "5.6.0")

Pass Criteria: โœ“ Version matches running UE version


Test 4.3: Create Actor via Python

Prompt for Claude:

Use Python to spawn a sphere at (1000, 0, 100) and name it "PythonSphere".

Expected Response:

  • Actor spawned via Python
  • Name: PythonSphere
  • Location: (1000, 0, 100)

Pass Criteria: โœ“ PythonSphere appears in World Outliner


Test Suite 5: Skill System (3 min)

Test 5.1: List Skills

Prompt for Claude:

What skills are available in the UMCP system?

Expected Response:

  • 6 skills listed:
    • screenshot
    • blueprint-helper
    • spawn-actor
    • level-organizer
    • organize_level
    • create_lighting_setup

Pass Criteria: โœ“ All 6 skills discovered


Test 5.2: Execute Screenshot Skill

Prompt for Claude:

Use the screenshot skill to capture a high-quality render of the viewport.

Expected Response:

  • Skill executed successfully
  • High-quality screenshot generated

Pass Criteria: โœ“ Screenshot created with higher quality than basic capture


Test 5.3: Execute Level Organizer Skill

Prompt for Claude:

Use the level-organizer skill to organize the World Outliner.

Expected Response:

  • Actors organized into folders
  • Folders created by type (Lights, Geometry, etc.)

Pass Criteria: โœ“ World Outliner shows folder hierarchy


Test Suite 6: Editor Control (2 min)

Test 6.1: Play in Editor

Prompt for Claude:

Start Play in Editor mode.

Expected Response:

  • PIE started successfully
  • Editor enters play mode

Pass Criteria: โœ“ Game starts playing in viewport


Test 6.2: Stop Play in Editor

Prompt for Claude:

Stop Play in Editor.

Expected Response:

  • PIE stopped successfully
  • Editor returns to edit mode

Pass Criteria: โœ“ Game stops, editor returns to normal


Test Suite 7: Complex Workflows (5 min)

Test 7.1: Create Scene with Multiple Actors

Prompt for Claude:

Create a simple scene with:
- A floor plane at (0, 0, 0) scaled to (10, 10, 1)
- Three cubes at (0, 0, 100), (200, 0, 100), (400, 0, 100)
- Name them Floor, Cube1, Cube2, Cube3

Expected Response:

  • All 4 actors spawned
  • Positioned correctly
  • Named correctly

Pass Criteria: โœ“ All actors visible in viewport at correct positions


Test 7.2: Batch Transformation

Prompt for Claude:

Move all three cubes up by 50 units (Z+50).

Expected Response:

  • All cubes moved up
  • New Z positions: 150, 150, 150

Pass Criteria: โœ“ All cubes elevated visually


Test 7.3: Scene Cleanup

Prompt for Claude:

Delete all actors I just created (Floor, Cube1, Cube2, Cube3).

Expected Response:

  • All actors deleted

Pass Criteria: โœ“ Actors removed from World Outliner


Test 7.4: Asset Query and Material Creation

Prompt for Claude:

Find all materials in the project, then create a new material called "M_TestMaterial".

Expected Response:

  • Material list returned
  • New material created
  • Path: /Game/Materials/M_TestMaterial (or similar)

Pass Criteria: โœ“ New material appears in Content Browser


Test Suite 8: Error Handling (2 min)

Test 8.1: Invalid Actor Name

Prompt for Claude:

Try to delete an actor named "NonExistentActor".

Expected Response:

  • Error message: Actor not found
  • Graceful error handling

Pass Criteria: โœ“ Error message returned, no crash


Test 8.2: Invalid Command

Prompt for Claude:

Try to execute a non-existent command called "do_magic_things".

Expected Response:

  • Error message: Unknown command
  • List of available commands

Pass Criteria: โœ“ Helpful error message with alternatives


Test 8.3: Malformed Parameters

Prompt for Claude:

Try to spawn an actor without specifying the actor_class parameter.

Expected Response:

  • Error message: Missing required parameter
  • Parameter documentation

Pass Criteria: โœ“ Clear error about missing parameter


Test Suite 9: Performance & Stress (Optional, 3 min)

Test 9.1: Batch Spawn

Prompt for Claude:

Spawn 100 cubes in a 10x10 grid pattern on the ground plane.

Expected Response:

  • 100 actors spawned
  • Grid pattern visible

Pass Criteria: โœ“ All actors spawned, no timeout


Test 9.2: Large Asset List

Prompt for Claude:

List all assets in the entire Content folder, including all subfolders.

Expected Response:

  • Complete asset list
  • Response time <5 seconds

Pass Criteria: โœ“ All assets returned without timeout


Test Results Summary

Scorecard

Test Suite Tests Passed Tests Failed Pass Rate
1. Basic Connectivity __ / 2 __ __%
2. Actor Lifecycle __ / 6 __ __%
3. Scene Discovery __ / 3 __ __%
4. Python Execution __ / 3 __ __%
5. Skill System __ / 3 __ __%
6. Editor Control __ / 2 __ __%
7. Complex Workflows __ / 4 __ __%
8. Error Handling __ / 3 __ __%
9. Performance (Optional) __ / 2 __ __%
TOTAL __ / 28 __ __%

Target Pass Rate: โ‰ฅ 90% (25/28 tests)


Issues Encountered

Issue 1:

  • Test: _________
  • Description: _________
  • Severity: Critical / Major / Minor
  • Workaround: _________

Issue 2:

  • Test: _________
  • Description: _________
  • Severity: Critical / Major / Minor
  • Workaround: _________

Issue 3:

  • Test: _________
  • Description: _________
  • Severity: Critical / Major / Minor
  • Workaround: _________

Observations

What Worked Well:




What Needs Improvement:




Unexpected Behaviors:





Performance Metrics

Metric Value
Average Response Time _____ ms
Longest Response Time _____ ms (Test: _____)
Failed Requests _____
Timeout Events _____
Server Uptime _____ seconds

Next Steps

Based on test results:

  1. If pass rate โ‰ฅ 90%:

    • System is production-ready
    • Begin real-world usage with Claude Desktop
    • Monitor for edge cases
  2. If pass rate 70-89%:

    • Investigate failed tests
    • Check parameter documentation
    • Review error logs in UE
  3. If pass rate < 70%:

    • Critical issues detected
    • Review server configuration
    • Check plugin installation
    • Verify Unreal Engine version compatibility

Recommended Follow-Up Tests

After completing this test plan, consider:

  1. Multi-Instance Testing

    • Test with multiple Unreal instances
    • Verify instance discovery
  2. Long-Running Operations

    • Test large-scale scene creation
    • Monitor memory usage
  3. Blueprint Automation

    • Test blueprint creation
    • Test auto-wiring
  4. Asset Import Pipeline

    • Test asset import
    • Test batch import
  5. Real-World Workflows

    • Level design tasks
    • Asset organization
    • Scene setup automation

Test Plan Version: 1.0.0 Last Updated: 2025-11-30 Prepared By: MCP Tool Analyzer Agent Estimated Duration: 15-20 minutes (core tests), 30+ minutes (with optional tests)