Unreal MCP
Unreal MCP connects Claude, Cursor, Copilot and other AI agents to the Unreal Editor. Learn how the plugin, bridge, CLI and app fit together.
Unreal MCP is the Unreal Engine plugin of AI Game Developer. It lets an AI agent such as Claude Code, Cursor, GitHub Copilot, Codex or Gemini work inside the Unreal Editor: query and change actors, components, assets and levels, author and compile Blueprints, edit and compile C++, run editor Python, and capture screenshots it can look at. You describe what you want in plain language and the agent calls the editor's tools through the Model Context Protocol (MCP).
This wiki explains how to install the plugin, connect your AI client, and get useful results from AI in the Unreal Editor. The plugin is in beta; the current release is 0.26.2.
How the pieces fit together
An Unreal MCP setup has four parts. Knowing them makes setup and troubleshooting much easier.
- The Unreal MCP plugin (
UnrealMCP). A C++ editor plugin installed in your project'sPlugins/folder. It registers the AI tools and adds the AI Game Developer window to the editor. - The bridge (
unreal-mcp-bridge). Unreal's editor is C++, so the .NET MCP host runs as a small helper process next to the editor. The plugin starts it automatically and talks to it over a local, token-protected connection. A packaged release bundles the bridge, so you do not install .NET. - The MCP server. By default the bridge connects to the hosted server at ai-game.dev (Cloud mode). You can instead run the shared local server
gamedev-mcp-serveron your own machine (Custom mode). - Your AI client. Any MCP client connects to the same server and reaches your editor through it.
Two tools help you set this up:
- AI Game Dev, the desktop app, installs the plugin into an Unreal project for you and can open the editor. No Node.js is needed. See the app wiki.
unreal-mcp-cli, a cross-platform command-line tool on npm. It installs and updates the plugin, signs you in, writes MCP client configs, opens the editor and runs tools from a terminal or CI.
What the AI agent can do
The plugin ships 27 tools in 7 families. Most editor data goes through RQL, one query language for the whole editor, instead of one tool per operation:
- RQL (
rql-query,rql-mutate,rql-schema): find and change actors, components, assets, levels, DataTables, gameplay tags, Project Settings, Blueprint graphs, Widget Blueprints and Level Sequences. - Blueprints: create, edit and compile Blueprints, with a structured error list the agent can fix.
- C++ source: read, scaffold, edit and compile project C++ (Live Coding or Unreal Build Tool).
- Python: run Python with the full
unrealeditor API. - Screenshots: viewport, game view, camera and isolated-actor captures.
- Console and tool list helpers.
Six optional extensions add Control Rig, Enhanced Input, Gameplay Ability System, Landscape and Water, Niagara and PCG support. You can also register your own tools from any Unreal plugin.
Wiki pages
| Page | What you will find |
|---|---|
| Installation | Requirements and every install path that exists today: the app, the CLI, and manual install |
| Getting started | First connection, first prompts, and checking that everything works |
| Connecting AI clients | Cloud vs Custom mode, device-code sign-in, project keys, and setup-mcp per agent |
| Configuration | The AI Game Developer window, the MCP Tools window, the config file and UNREAL_MCP_* variables |
| Tools | Every built-in tool and what to ask for |
| Extensions | The six optional extensions and how to install them |
| Custom tools | Add your own tools, prompts and resources from C++, including in-game (runtime) tools |
| CLI reference | Every unreal-mcp-cli command and flag |
| Troubleshooting | Fixes for connection, bridge, build and tool problems |
| FAQ | Short answers to common questions |
Where to go next
- New here? Start with Installation, then Getting started.
- Want the short version? The Unreal Engine install page and the quickstart cover the essentials.
- Looking for tool details? Browse the Unreal tools catalog and the Unreal best practices.
- Using another engine as well? See the Unity wiki and the Godot wiki. All three plugins share the same server, account and AI clients.
- InstallationInstall the Unreal MCP plugin for Unreal Engine 5.5+ with the AI Game Dev app or unreal-mcp-cli. Requirements, updates, removal and Fab status.
- Getting startedConnect Unreal MCP for the first time, sign in to ai-game.dev, hook up your AI client and run your first AI prompts in the Unreal Editor.
- Connecting AI clientsConnect Claude Code, Cursor, Copilot, Codex, Gemini and other MCP clients to Unreal MCP through the ai-game.dev cloud endpoint or a local server.
- ConfigurationConfigure Unreal MCP - the AI Game Developer window, MCP Tools visibility, the project config file, .env and every UNREAL_MCP_* environment variable.
- ToolsThe 27 built-in Unreal MCP tools - RQL, Blueprints, C++ source, Python, screenshots and console - and example prompts for AI in the Unreal Editor.
- ExtensionsOptional Unreal MCP extensions for Control Rig, Enhanced Input, GAS, Landscape and Water, Niagara and PCG. What each adds and how to install it.
- Custom toolsAdd your own AI tools, RQL modules, prompts and resources to Unreal MCP from any C++ plugin, and expose gameplay tools in a running game.
- CLI referenceComplete unreal-mcp-cli command reference - install the Unreal MCP plugin, sign in, configure AI agents, open the editor, run tools and Python.
- TroubleshootingFix common Unreal MCP problems - plugin not loading, bridge missing, connection stuck, AI agent without Unreal tools, stale builds and tool errors.
- FAQAnswers to common questions about Unreal MCP - supported engine versions, AI clients, Blueprint-only projects, Fab, offline use, cost and packaged games.