Rohlik MCP Proxy
Enables interaction with the Rohlik grocery delivery service through MCP by proxying requests with automatic authentication header injection.
README
Rohlik MCP Proxy Server
A Docker-based proxy MCP (Model Context Protocol) server for the Rohlik MCP service that forwards requests to https://mcp.rohlik.cz/mcp while automatically injecting authentication headers from a base64url token in the URL path (or optional env fallback).
Features
- Transparent Proxying: Forwards MCP protocol messages between clients and the Rohlik MCP server
- Authentication Injection: Automatically injects
rhl-emailandrhl-passheaders from a base64url path token (env fallback optional) - SSE/Streaming Support: Streams GET responses for real-time MCP sessions
- Docker-based: Easy deployment with Docker and Docker Compose
- Health Monitoring: Built-in health check endpoint
- CORS Enabled: Allows browser-based and desktop MCP clients to connect
Prerequisites
- Docker and Docker Compose installed
- Rohlik account credentials (email and password)
Quick Start
1. Generate a base64url token
Create a base64url-encoded token for email:password:
python - <<'PY'
import base64
print(base64.urlsafe_b64encode(b"email@gmail.com:password").decode())
PY
2. Build and Run with Docker Compose
Note: docker-compose.yml expects an external network named cloudflare. Create it with
docker network create cloudflare or remove the networks section if you don't need it.
# Build and start the proxy server
docker-compose up -d
# View logs
docker-compose logs -f rohlik-mcp-proxy
# Stop the server
docker-compose down
The proxy server will be available at:
- MCP Endpoint:
http://localhost:8000/mcp/<base64url-token> - Health Check:
http://localhost:8000/health - Info:
http://localhost:8000/
3. Alternative: Run with Docker
# Build the image
docker build -t rohlik-mcp-proxy .
# Run the container
docker run -d \
--name rohlik-mcp-proxy \
-p 8000:8000 \
rohlik-mcp-proxy
# View logs
docker logs -f rohlik-mcp-proxy
# Stop the container
docker stop rohlik-mcp-proxy
docker rm rohlik-mcp-proxy
Configuration
Environment Variables
| Variable | Required | Default | Description |
|---|---|---|---|
RHL_EMAIL |
No | - | Optional fallback email (used when no path token is provided) |
RHL_PASS |
No | - | Optional fallback password (used when no path token is provided) |
ROHLIK_MCP_URL |
No | https://mcp.rohlik.cz/mcp |
Rohlik MCP backend endpoint |
Base64url Path Token
The proxy expects a URL path like /mcp/<base64url(email:password)>. Padding (=) is optional.
If you set RHL_EMAIL and RHL_PASS, you can also use /mcp without a token.
MCP Client Configuration
Configure your MCP client (e.g., Claude Desktop, ChatGPT) to connect to the proxy. Use the base64url token in the URL path:
{
"mcpServers": {
"rohlik": {
"url": "http://localhost:8000/mcp/<base64url-token>",
"transport": "streamable-http"
}
}
}
Or for fastmcp client:
Or using MCP Inspector:
npx @modelcontextprotocol/inspector http://localhost:8000/mcp/<base64url-token>
Development
Local Development (Without Docker)
# Install uv package manager
pip install uv
# Install dependencies
uv pip install -e .
# Run the proxy server
python proxy_server.py
Project Structure
rohlik-mcp-proxy/
├── proxy_server.py # Main proxy server implementation
├── pyproject.toml # Python project configuration
├── Dockerfile # Docker image definition
├── docker-compose.yml # Docker Compose configuration
├── .env.example # Environment variables template
├── .gitignore # Git ignore rules
└── README.md # This file
How It Works
- Client Connection: MCP client connects to the proxy at
/mcp/<base64url-token>(GET for streaming, POST for calls) - Request Forwarding: Proxy receives MCP requests from the client
- Header Injection: Proxy decodes the path token and adds authentication headers (
rhl-email,rhl-pass) - Backend Communication: Proxy forwards the authenticated request to Rohlik MCP server
- Response Handling: Proxy streams GET responses and returns POST/other responses
- Bidirectional Flow: Process repeats for all subsequent requests
┌─────────────┐ ┌──────────────────┐ ┌─────────────────┐
│ │ │ │ │ │
│ MCP Client │ <-----> │ Rohlik Proxy │ <-----> │ Rohlik MCP │
│ │ HTTP │ (adds headers) │ HTTPS │ Backend Server │
│ │ │ │ │ │
└─────────────┘ └──────────────────┘ └─────────────────┘
Port 8000 mcp.rohlik.cz
Health Check
The proxy includes a built-in health check endpoint:
curl http://localhost:8000/health
# Returns: OK
Docker Compose automatically monitors this endpoint and restarts the container if it becomes unhealthy.
Security Considerations
- Credentials Storage: Never commit your
.envfile with real credentials to version control - Network Security: For production use, consider running the proxy behind a reverse proxy with HTTPS
- Access Control: No additional authentication is enforced by the proxy; restrict access at the network or reverse proxy level
- Password Safety: Credentials are passed via URL path and HTTP headers (ensure HTTPS and avoid logging full URLs in production)
Troubleshooting
Container won't start
# Check logs
docker-compose logs rohlik-mcp-proxy
# Common issues:
# - Port 8000 already in use
# - Invalid or missing base64 token in the URL path
Connection refused
# Verify the container is running
docker-compose ps
# Check if port is accessible
curl http://localhost:8000/health
Rohlik authentication errors
# Verify your base64url token decodes to email:password
python - <<'PY'
import base64
print(base64.urlsafe_b64decode(b"<base64url-token>").decode())
PY
License
This project is provided as-is for integration with the Rohlik MCP service.
Related Documentation
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。
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。
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。
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。
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。
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。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
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。