📋 Development Roadmap - All 53 Phases (Updated with Context Optimization)

[ChiR24]

Unreal MCP Development Roadmap

Target: ~2,855 actions covering all Unreal Engine subsystems and major plugin integrations.

Current Status

Category	Phases	Status
Foundation	1-5	✅ Complete (4/5)
Core Systems	6-25	✅ Complete (20/20)
Advanced Systems	26-35	🔄 In Progress (1/10)
Plugin Integration	36-47	⏳ Pending (0/12)
Context Optimization	48-53	🆕 NEW (0/6)

---

✅ Completed Phases (1-25)

Foundation

• Phase 1: Architecture & Foundation
• Phase 2: Graph & Logic Automation
• Phase 3: Cinematic & Visual Automation
• Phase 4: System & Developer Experience
• Phase 5: Infrastructure Improvements 🔄

Core Systems

• Phase 6: Geometry & Mesh Creation (manage_geometry)
• Phase 7: Skeletal Mesh & Rigging (manage_skeleton)
• Phase 8: Advanced Material Authoring (manage_material_authoring)
• Phase 9: Texture Generation & Processing (manage_texture)
• Phase 10: Complete Animation System (manage_animation_authoring)
• Phase 11: Complete Audio System (manage_audio_authoring)
• Phase 12: Complete Niagara VFX (manage_niagara_authoring)
• Phase 13: Gameplay Ability System (manage_gas)
• Phase 14: Character & Movement (manage_character)
• Phase 15: Combat & Weapons (manage_combat)
• Phase 16: Complete AI System (manage_ai)
• Phase 17: Inventory & Items (manage_inventory)
• Phase 18: Interaction System (manage_interaction)
• Phase 19: Complete UI/UX System (manage_widget_authoring)
• Phase 20: Networking & Multiplayer (manage_networking)
• Phase 21: Game Framework (manage_game_framework)
• Phase 22: Sessions & Local Multiplayer (manage_sessions)
• Phase 23: World & Level Structure (manage_level_structure)
• Phase 24: Volumes & Zones (manage_volumes)
• Phase 25: Navigation System (manage_navigation)

---

⏳ Pending Phases (26-47)

Advanced Systems (26-35)

• Phase 26: Spline System 🔄 In Progress
• Phase 27: PCG Framework
• Phase 28: Environment Systems
• Phase 29: Advanced Lighting & Rendering
• Phase 30: Cinematics & Media
• Phase 31: Data & Persistence
• Phase 32: Build & Deployment
• Phase 33: Testing & Quality
• Phase 34: Editor Utilities
• Phase 35: Additional Gameplay Systems

Plugin Integration (36-47)

• Phase 36: Character & Avatar Plugins (MetaHuman, RPM, Mutable)
• Phase 37: Asset & Content Plugins (USD, Megascans, Houdini, Substance)
• Phase 38: Audio Middleware (Wwise, FMOD)
• Phase 39: Motion Capture & Live Link
• Phase 40: Virtual Production (nDisplay, Composure, DMX)
• Phase 41: XR Plugins (OpenXR, Quest, ARKit)
• Phase 42: AI & NPC Plugins (Convai, Inworld)
• Phase 43: Online Services (EOS, Steam)
• Phase 44: Streaming & Distribution (Pixel Streaming)
• Phase 45: Utility Plugins (Python, Paper2D)
• Phase 46: Physics & Destruction (Chaos)
• Phase 47: Accessibility System

---

🆕 Context Optimization Phases (48-53)

Why? As of v0.5.9, the MCP server has grown to 38 tools consuming ~78,000 tokens (~60k context). These phases address token reduction while maintaining full functionality.

Expected Impact

Metric	Before	After	Reduction
Tool Definitions	~78k tokens	~15k tokens	81%
Default Context	~60k used	~15k used	75%
Available for Chat	~40%	~85%	+45%

Phases

Phase	Name	Priority	Est. Reduction
48	Schema Pruning	🔴 Critical	~23k tokens
49	Common Schema Extraction	🟠 High	~8k tokens
50	Dynamic Tool Loading (Lazy)	🔴 Critical	~50k tokens
51	Stub Tools System	🟠 High	~15k tokens
52	Tool Help System	🟡 Medium	~5k tokens
53	Strategic Tool Merging	🟡 Medium	~10k tokens

Phase Details

Phase 48: Schema Pruning (Immediate Win)

• Remove redundant "Supported actions:" lists from descriptions
• Condense all descriptions to 1-2 sentences
• Remove verbose "Use it when you need to:" bullet points

Phase 49: Common Schema Extraction

• Move repeated parameters (assetPath, location, save) to shared schemas
• Eliminate ~30+ duplicate parameter definitions

Phase 50: Dynamic Tool Loading (Biggest Impact)

• Implement MCP notifications/tools/list_changed for runtime tool updates
• Categorize tools: core (always on), vfx, ui, audio, gameplay, advanced
• Only load non-core tools when explicitly enabled via enable_tools
• No server restart required - fully dynamic

Phase 51: Stub Tools System

• Show minimal placeholders for disabled tools
• AI can still discover tools exist without full schema cost
• Guides AI to call enable_tools when needed

Phase 52: Tool Help System

• Create get_tool_help for on-demand documentation
• Move examples and detailed usage out of schemas

Phase 53: Strategic Tool Merging

• Merge related tool pairs (e.g., manage_audio + manage_audio_authoring)
• Reduces total tool count from 38 to ~34

---

📊 Track Progress

👉 View Full Project Board

Each phase has detailed action lists, implementation status, and priorities.

💡 Want to Contribute?

Pick a pending phase and help us implement it! Check the project board for details on each phase.

[xztyle]

Amazing work on this. I'm a close follower of your MCP and I use it actively.

As of ver. 0.5.9, due to the increase in the number of tools, the context window is being eaten up rather quickly, with like 60k tokens used by the MCPs, being this one the heaviest in that sense

Is there a way to reduce token usage, or you plan to address that in future releases? Thanks

[ChiR24]

Hey @xztyle! 👋 Thanks for the feedback and for actively using the MCP!

You're absolutely right - this is a critical issue I been aware of. The tool definitions have grown to ~78,000 tokens (~312k characters), which is eating up a massive chunk of the context window well when more and more tool added in future it will become a massive context so I plan to reduce the context for current tools as some will also for future addition tool.

Great news: I just added 6 new phases specifically to address this! 🎉

Check out Phases 48-53: Context Optimization in the updated roadmap above.

TL;DR - Expected Impact

Metric	Before	After	Reduction
Tool Definitions	~78k tokens	~15k tokens	81%
Default Context	~60k used	~15k used	75%
Available for Chat	~40%	~85%	+45%

Key Strategies

1. Phase 48: Schema Pruning (Immediate) - Remove verbose descriptions that repeat enum values
2. Phase 49: Common Schemas - Extract repeated parameters like assetPath, location
3. Phase 50: Dynamic Tool Loading - Only load tools when needed using MCP's notifications/tools/list_changed
4. Phase 51: Stub Tools - Show minimal placeholders for disabled tools so AI knows they exist
5. Phase 52: Help System - Move detailed docs to on-demand get_tool_help
6. Phase 53: Tool Merging - Combine related tools (e.g., manage_audio + manage_audio_authoring)

Timeline

• Phase 48 (Schema Pruning) is marked as Critical priority and will be the first addressed
• Phase 50 (Dynamic Loading) will have the biggest impact but needs MCP SDK integration

These are now tracked on the Project Board under the Foundation category.

Thanks for raising this - it's genuinely one of the most important scalability issues to solve! 🚀

While I was working on it you can use previous version v0.5.8 till some context reduction is done.

[ukie-552]

Thank you so much for your hard work, this MCP is amazing!