An MCP server that speaks Unreal's native dialect

What it does

Unreal_mcp exposes Unreal Engine’s editor internals to AI assistants through the Model Context Protocol. It combines a native C++ plugin with an optional TypeScript server, letting tools like Claude or Cursor manipulate assets, control the viewport, edit Blueprint and Niagara graphs, manage Sequencer timelines, and execute console commands through structured tool calls instead of ad-hoc Python snippets.

The interesting bit

The plugin includes a built-in HTTP/SSE server that speaks MCP natively, so AI clients can connect directly without the TypeScript bridge or any Node.js runtime. The server also starts independently of Unreal and retries the automation handshake with exponential backoff until the editor appears.

Key highlights

• Native C++ Automation Bridge for UE 5.0–5.8 with an optional TypeScript WebSocket fallback
• Direct HTTP/SSE transport means AI clients can connect straight to the plugin without npm or Node.js
• Broad surface coverage: asset management, actor control, Blueprint/Niagara/Material graph editing, Sequencer cinematics, audio, physics, and console commands
• Runtime dynamic type discovery for lights, debug shapes, and sequencer tracks
• Safety features include pattern-based blocking of dangerous console commands, optional capability-token auth, and per-IP rate limiting on metrics endpoints

Caveats

• Blueprint-only projects cannot compile the plugin directly and must either convert to a C++ target or use version-locked pre-built binaries
• Advanced tool categories (Sequencer, Control Rig, Niagara authoring) require optional engine plugins, though the bridge auto-enables them when present

Verdict

Technical artists and developers who want AI agents to actually manipulate the Unreal editor rather than just generate code will find this invaluable; pure Blueprint creators who avoid C++ compilation may find the setup friction annoying.