MCP Automation Bridge turns Unreal Editor into an AI-controllable editor automation endpoint using the Model Context Protocol (MCP).
It lets MCP-compatible AI clients inspect and control Unreal Engine through a native editor plugin, without constantly copying logs, screenshots, asset paths, or scene context by hand. The plugin exposes editor automation actions for assets, actors, levels, Blueprints, materials, Niagara, PCG, Sequencer, AI tools, audio, and more.
Key Features:
- Native HTTP/SSE MCP transport built directly into the Unreal Editor plugin
- Optional WebSocket bridge support for external MCP server workflows
- 200+ editor automation actions across 23 tool groups
- Asset browsing, creation, import, duplication, rename, and deletion helpers
- Actor spawning, transforms, tags, components, physics, and editor viewport control
- Level load/save, streaming, lighting, environment, foliage, landscape, and volume tools
- Blueprint, graph, widget, material, Niagara, PCG, AI, Sequencer, audio, and gameplay-system authoring helper tools
- Dynamic MCP tool loading so clients can enable only the tools they need
- Security-focused defaults: loopback binding, optional capability token auth, TLS settings, and rate limits
- Native MCP mode does not require Node.js
- Editor-only plugin; not intended for packaged game runtime builds
Requirements:
- Unreal Engine 5.7 for this Fab package version
- Required engine plugins: Editor Scripting Utilities, Niagara, and PythonScriptPlugin
- Standard Fab installation should not require Node.js for Native MCP mode
- Source rebuilds, custom engine builds, or manual source installs may require a C++ toolchain
- Optional integrations are enabled when available, including Sequencer, Control Rig, Geometry Scripting, PCG, Gameplay Abilities, MetaSounds, StateTree, Smart Objects, and Online Subsystem tools
Documentation:
https://github.com/ChiR24/Unreal_mcp/tree/main/plugins/McpAutomationBridge
Support:
https://github.com/ChiR24/Unreal_mcp/discussions
Native MCP client configuration example:
After enabling Native MCP in Project Settings and restarting Unreal Editor, connect your MCP client to:
{
"mcpServers": {
"unreal-engine": {
"url": "http://localhost:3000/mcp"
}
}
}
Beta / Active Development Notice:
This tool is in active development. You may encounter bugs, and the overall performance depends heavily on the specific AI model and MCP client you are using. Please use with caution and always back up your projects.