# UE MCP Strategy — Index

## Scope

Learn the landscape around one question — **should Unreal Engine open an MCP (an agent-facing tool surface) rather than build an in-house AI "harness," and does that strengthen the runtime's gravity?** — well enough to reason critically. See [[research/ue-mcp-strategy/WORKSPACE|WORKSPACE]] for full scope and the working thesis.

## Start Here

- [[research/ue-mcp-strategy/outputs/first-pass-synthesis|First-Pass Synthesis]] — the revised thesis: not "MCP vs harness," but a layer/data/output-constraint decision pointing at Verse/UEFN.
- [[research/ue-mcp-strategy/outputs/critical-open-questions|Critical Open Questions]] — the five questions that decide the answer, ranked, plus verification debt.

## Concepts

- [[research/ue-mcp-strategy/concepts/ue-runtime-architecture|UE Runtime Architecture]] — how the runtime/editor work; the reflection system is the substrate an MCP plugs into (but reflection is *curated*, and Verse may be the real strategic surface).
- [[research/ue-mcp-strategy/concepts/mcp-as-tooling-surface|MCP as a Tooling Surface]] — what MCP is, what an Unreal MCP wraps, and the 8+ existing community proto-MCPs (plus UE 5.7's official AI Assistant).
- [[research/ue-mcp-strategy/concepts/harness-vs-mcp|Harness vs MCP]] — own-the-intelligence vs rent-the-frontier's; how Figma/Unity/Adobe/Notion/Roblox/Blender chose, sorted by value-capture layer.
- [[research/ue-mcp-strategy/concepts/ue-value-capture-map|Epic / UE Value-Capture Map]] — the money map; "the runtime" = Fortnite (~$3.5-6.2B), not the UE engine (~$275M); which channel captures the AI-content wave.
- [[research/ue-mcp-strategy/concepts/ai-content-generation-frontier|AI + PCG Frontier]] — rule-based PCG vs AI generation; state of the art; why the flywheel is training-time, not exposure-time.
- [[research/ue-mcp-strategy/concepts/runtime-gravity-and-lock-in|Runtime Gravity & Lock-In]] — what the moat mechanically is, and how USD/glTF portability (Epic co-founded AOUSD) cuts against content lock-in.
- [[research/ue-mcp-strategy/concepts/the-bear-case|The Bear Case]] — the strongest arguments *against* opening the MCP, ranked; engine-agnostic commoditization is the live falsifier.

## Sources (cited digests)

- [[research/ue-mcp-strategy/sources/ue-reflection-and-scripting-apis|UE reflection + scripting APIs]]
- [[research/ue-mcp-strategy/sources/mcp-spec|MCP spec / announcement]]
- [[research/ue-mcp-strategy/sources/existing-unreal-mcp-servers|Existing Unreal MCP servers]]
- [[research/ue-mcp-strategy/sources/platform-ai-tooling-strategies|Platform AI tooling strategies]]
- [[research/ue-mcp-strategy/sources/epic-monetization-model|Epic monetization model]]
- [[research/ue-mcp-strategy/sources/ai-3d-and-world-gen-tools|AI 3D + world-gen tools]]
- [[research/ue-mcp-strategy/sources/interchange-and-lock-in|Interchange standards & lock-in (USD/glTF)]]

## Open Questions (full list in Critical Open Questions)

1. Does more AI-generated content grow Fortnite's total wallet, or just reshuffle the fixed 40% UEFN pool? *(decides everything)*
2. Can an MCP be forced to emit UE-advantaged, non-portable content? *(the linchpin)*
3. Is the flywheel real for Epic — does the MCP capture a private training loop?
4. What exactly is UE 5.7's AI Assistant, and does it speak MCP? *(resets the frame; verify from primary source)*
5. Was Unity's Muse retirement strategy or capability failure?

## Promotion Candidates

- Value-capture layer determines who can safely open an MCP
- Rent-the-frontier's-intelligence vs own-the-intelligence
- Content gravity (not the runtime) is the real lock-in — and Epic is dissolving it via USD
- UE-advantaged vs engine-neutral generation (the unenforced linchpin)

## Verification Debt

- Fortnite revenue is analyst estimate only (undisclosed since 2022).
- UE 5.7 AI Assistant: single secondary source; confirm from Epic primary docs.
- UE MCP tool counts are self-reported; most capable server is closed.
- Engine-agnostic-MCP threat is a live risk, not a realized instance.