Unreal Engine Assistant MCP
Connects to Epic's Unreal AI Assistant to provide accurate, up-to-date documentation, C++ code snippets, and Blueprint guidance through the Model Context Protocol.
README
<!-- SPDX-FileCopyrightText: 2026 Daniel Eder
SPDX-License-Identifier: CC0-1.0
-->
Unreal Engine Assistant MCP
A Model Context Protocol (MCP) server that connects to the Epic Games Unreal AI Assistant.
This server allows AI assistants (like Claude, Cursor, or any MCP-compatible client) to query Epic's Unreal AI and retrieve highly accurate, up-to-date documentation, C++ code snippets, and Blueprint guidance. It respects Epic's allowed contents (/allowed endpoint) and rate limits (/check_limit endpoint). Puppeteer is used to avoid reverse-engineering their API.
Prerequisites
- Node.js (v18+)
- Yarn (
yarn set version stable)
Installation
Clone this repository and install dependencies:
git clone https://github.com/daniel-eder/unreal-engine-assistant-mcp.git
cd unreal-engine-assistant-mcp
yarn install
Usage
This server supports multiple MCP transport methods:
1. Stdio (Default)
Standard communication over stdout/stderr. This is the most common integration method for tools like Claude Desktop and Cursor.
Note: Because this project uses Yarn Zero Installs (PnP), you must use
yarn nodeinstead of standardnodeto resolve dependencies.
yarn node index.js
2. Standard Server-Sent Events (SSE)
Starts a local web server to handle context protocol requests via plain SSE.
yarn node index.js --sse
By default it listens at: http://127.0.0.1:3000/sse
(You can override the port/host by passing --port XXXX and --host XXXX, e.g., yarn node index.js --sse --port 4000 --host localhost)
3. Streamable HTTP
Starts a local web server with full MCP Streamable HTTP session support.
yarn node index.js --streamable-http
By default it listens at: http://127.0.0.1:3000/mcp
(You can override the port/host by passing --port XXXX and --host XXXX, e.g., yarn node index.js --streamable-http --port 4000 --host localhost)
Development
If you want to run the server in development mode, simply use the scripts provided in package.json:
yarn start- Starts the MCP server via standard stdio transport.yarn start:sse- Starts the plain HTTP SSE server.yarn start:streamable-http- Starts the MCP Streamable HTTP server.yarn start:streamable-http --port 8080 --host localhost- Starts the HTTP server with a custom port and host.
Command-Line Arguments
The server accepts several command-line flags to customize its behavior:
| Argument | Description | Default |
|---|---|---|
--sse |
Runs the server using the plain Server-Sent Events (SSE) transport. Suitable for simple HTTP integration. | (Disabled) |
--streamable-http |
Runs the server using the advanced MCP Streamable HTTP transport with session support. | (Disabled) |
--port <number> |
Overrides the port used when binding the HTTP server (for --sse and --streamable-http). |
3000 |
--host <string> |
Overrides the host used when binding the HTTP server. Use 0.0.0.0 to expose to the local network. |
127.0.0.1 |
--delay <number> |
Adjusts the sleep duration (in milliseconds) used by the headless browser to wait for Cloudflare validation before querying Epic's APIs. If queries fail, try increasing this. | 2000 |
(If no transport is defined via --sse or --streamable-http, the server defaults to Stdio transport)
MCP Client Configuration
Add the following to your mcp config json:
{
"mcpServers": {
"unreal-ai-assistant": {
"command": "yarn",
"args": [
"node",
"/path/to/your/unreal-engine-assistant-mcp/index.js"
]
}
}
}
Available Tools
ask_unreal_engine_assistant- Sends your question to the Unreal Engine Assistant and returns the structured markdown/HTML response. Takes a single string parameter:question.
Troubleshooting
- Puppeteer crashing or failing to launch: Sometimes the stealth plugin requires system libraries. Ensure you have standard local chromium dependencies installed if running on Linux. On Windows/Mac, it usually runs out of the box.
- ARM / Raspberry Pi Support: Puppeteer's bundled Chromium may not work on ARM architectures out of the box. To run this project on a Raspberry Pi or other ARM devices:
- Install the native Chromium browser:
sudo apt-get install chromium-browser - The server will automatically attempt to use
/usr/bin/chromium-browser. - If your Chromium is installed in a different location, set the
PUPPETEER_EXECUTABLE_PATHenvironment variable before running the server (e.g.,export PUPPETEER_EXECUTABLE_PATH=/custom/path/to/chromium).
- Install the native Chromium browser:
- Initial Request Delay: The very first question may take a few seconds as the headless browser sets up the site https://dev.epicgames.com/community/assistant/unreal-engine. Subsequent requests in the same session will be much faster. You can use
--delayto control the artificial delay while waiting for cloudflare verification.
Licensing
Copyright (c) 2026 Daniel Eder
All content in this repository is licensed under at least one of the licenses found in ./LICENSES; you may not use this file, or any other file in this repository, except in compliance with the Licenses. You may obtain a copy of the Licenses by reviewing the files found in the ./LICENSES folder.
Unless required by applicable law or agreed to in writing, software distributed under the Licenses is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See in the ./LICENSES folder for the specific language governing permissions and limitations under the Licenses.
This project follows the REUSE standard for software licensing. Each file contains copyright and license information, and license texts can be found in the ./LICENSES folder. For more information visit https://reuse.software/. You can find a guide for developers at https://telekom.github.io/reuse-template/.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。
mcp-server-qdrant
这个仓库展示了如何为向量搜索引擎 Qdrant 创建一个 MCP (Managed Control Plane) 服务器的示例。