MCP Selenium Grid
Enables browser automation by managing Selenium Grid browser instances (Chrome, Firefox, Edge) in Docker or Kubernetes environments. Provides secure, scalable browser control for AI agents and testing workflows.
README
🤖 MCP Selenium Grid
A Model Context Protocol (MCP) server for managing Selenium Grid browser instances. Useful for browser automation and testing workflows.
The MCP Selenium Grid provides a MCP Server for creating and managing browser instances in both Docker and Kubernetes environments. It's designed to work with AI agents and automation tools that need browser automation capabilities.
Key Features
- Multi-browser support: Chrome, Firefox and Edge
- Dual backend support: Docker and Kubernetes deployment modes
- Secure API: Token-based authentication for browser management
- Scalable architecture: Support for multiple browser instances
- MCP compliance: Follows Model Context Protocol standards
🚀 Quick Start
Prerequisites
- uv (Python dependency manager)
- Docker (for Docker deployment mode)
- K3s (for Kubernetes deployment mode, optional)
📖 Usage
The MCP Selenium Grid provides a Web API for creating and managing browser instances. The server runs on localhost:8000 and exposes MCP endpoints at /mcp (Http Transport) and /sse (Server Sent Events).
Note: All requests to the server root
http://localhost:8000will be redirected to either/mcpor/sseendpoints, depending on the request, but you can choose to use directly/mcp(Http Transport) or/sse(Server Sent Events) endpoints.
Known Issues & Limitations
-
When you use STDIO transport with the
--stdioflag, the MCP Servers do not shut down the same way they do in the terminal. To clean up resources and remove containers or pods, run:uv run mcp-selenium-grid cleanuv run mcp-selenium-grid clean -d kubernetes
MCP Client Configuration
🐳 Docker Deployment
For Docker-based deployment, ensure Docker is running and use the Docker configuration in your MCP client setup.
{
"mcpServers": {
"mcp-selenium-grid": {
"command": "uvx",
"args": ["mcp-selenium-grid", "server", "run",
"--host", "127.0.0.1",
"--port", "8000",
"--stdio",
],
"env": {
"AUTH_ENABLED": "false",
"ALLOWED_ORIGINS": "[\"http://127.0.0.1:8000\"]",
"DEPLOYMENT_MODE": "docker",
"SELENIUM_GRID__USERNAME": "USER",
"SELENIUM_GRID__PASSWORD": "CHANGE_ME",
"SELENIUM_GRID__VNC_PASSWORD": "CHANGE_ME",
"SELENIUM_GRID__VNC_VIEW_ONLY": "false",
"SELENIUM_GRID__MAX_BROWSER_INSTANCES": "4",
"SELENIUM_GRID__SE_NODE_MAX_SESSIONS": "1",
"FASTMCP_EXPERIMENTAL_ENABLE_NEW_OPENAPI_PARSER": "true"
}
}
}
}
The server will be available at
http://localhost:8000with interactive API documentation athttp://localhost:8000/docs.
☸️ Kubernetes Deployment
3. Kubernetes Setup (Optional)
This project supports Kubernetes deployment for scalable browser instance management. We use K3s for local development and testing.
Install K3s (https://docs.k3s.io/quick-start)
# Install K3s
curl -sfL https://get.k3s.io | sh -
# Verify installation
k3s --version
# Start if not running
sudo systemctl start k3s
Create K3s Kubernetes Context
After installing K3s, you might want to create a dedicated kubectl context for it:
# Copy K3s kubeconfig
mkdir -p ~/.kube
sudo cp /etc/rancher/k3s/k3s.yaml ~/.kube/config-local-k3s
sudo chown $USER:$USER ~/.kube/config-local-k3s
chmod 600 ~/.kube/config-local-k3s
# Create context
KUBECONFIG=~/.kube/config-local-k3s \
kubectl config set-context k3s-selenium-grid \
--cluster=default \
--user=default
Deploy Selenium Grid
Please run for help to get to know the available commands and parameters:
uvx mcp-selenium-grid helm --help
uvx mcp-selenium-grid helm deploy --help
uvx mcp-selenium-grid helm uninstall --help
Deploy using default parameters:
uvx mcp-selenium-grid helm deploy
Uninstall using default parameters:
# using default parameters
uvx mcp-selenium-grid helm uninstall --delete-namespace
{
"mcpServers": {
"mcp-selenium-grid": {
"command": "uvx",
"args": ["mcp-selenium-grid", "server", "run",
"--host", "127.0.0.1",
"--port", "8000",
"--stdio",
],
"env": {
"AUTH_ENABLED": "false",
"ALLOWED_ORIGINS": "[\"http://127.0.0.1:8000\"]",
"DEPLOYMENT_MODE": "kubernetes",
"SELENIUM_GRID__USERNAME": "USER",
"SELENIUM_GRID__PASSWORD": "CHANGE_ME",
"SELENIUM_GRID__VNC_PASSWORD": "CHANGE_ME",
"SELENIUM_GRID__VNC_VIEW_ONLY": "false",
"SELENIUM_GRID__MAX_BROWSER_INSTANCES": "4",
"SELENIUM_GRID__SE_NODE_MAX_SESSIONS": "1",
"KUBERNETES__KUBECONFIG": "~/.kube/config-local-k3s",
"KUBERNETES__CONTEXT": "k3s-selenium-grid",
"KUBERNETES__NAMESPACE": "selenium-grid-dev",
"KUBERNETES__SELENIUM_GRID_SERVICE_NAME": "selenium-grid",
"FASTMCP_EXPERIMENTAL_ENABLE_NEW_OPENAPI_PARSER": "true"
}
}
}
}
{
"mcpServers": {
"mcp-selenium-grid": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"--init",
"--network=host",
"-v", "/home/user/.kube/config-local-k3s:/.kube/config-local-k3s:ro", // path to your kubeconfig file
"-e", "AUTH_ENABLED=false",
"-e", "ALLOWED_ORIGINS=[\\\"http://127.0.0.1:8000\\\"]",
"-e", "DEPLOYMENT_MODE=kubernetes", // required for docker
"-e", "SELENIUM_GRID__USERNAME=USER",
"-e", "SELENIUM_GRID__PASSWORD=CHANGE_ME",
"-e", "SELENIUM_GRID__VNC_PASSWORD=CHANGE_ME",
"-e", "SELENIUM_GRID__VNC_VIEW_ONLY=false",
"-e", "SELENIUM_GRID__MAX_BROWSER_INSTANCES=4",
"-e", "SELENIUM_GRID__SE_NODE_MAX_SESSIONS=1",
"-e", "KUBERNETES__KUBECONFIG=/.kube/config-local-k3s", // path to your kubeconfig file
"-e", "KUBERNETES__USE_HOST_DOCKER_INTERNAL=true",
"-e", "KUBERNETES__CONTEXT=k3s-selenium-grid",
"-e", "KUBERNETES__NAMESPACE=selenium-grid-dev",
"-e", "KUBERNETES__SELENIUM_GRID_SERVICE_NAME=selenium-grid",
"-e", "FASTMCP_EXPERIMENTAL_ENABLE_NEW_OPENAPI_PARSER=true",
"ghcr.io/catchnip/mcp-selenium-grid:latest"
]
}
}
}
The server will be available at
http://localhost:8000with interactive API documentation athttp://localhost:8000/docs.
Server with auth enabled
UVX
Using default args
uvx mcp-selenium-grid server run
Custom args
API_TOKEN=CHANGE_ME uvx mcp-selenium-grid server run --host 127.0.0.1 --port 8000
Docker
Default args
docker run -i --rm --init --network=host \
-v ~/.kube/config-local-k3s:/kube/config-local-k3s:ro \
-e KUBERNETES__KUBECONFIG=/kube/config-local-k3s \
ghcr.io/catchnip/mcp-selenium-grid:latest
Custom args
docker run -i --rm --init \
--network=host \
-v ~/.kube/config-local-k3s:/kube/config-local-k3s:ro \
-e API_TOKEN=CHANGE_ME \
-e ALLOWED_ORIGINS='["http://127.0.0.1:8000"]' \
-e DEPLOYMENT_MODE=kubernetes \
-e SELENIUM_GRID__USERNAME=USER \
-e SELENIUM_GRID__PASSWORD=CHANGE_ME \
-e SELENIUM_GRID__VNC_PASSWORD=CHANGE_ME \
-e SELENIUM_GRID__VNC_VIEW_ONLY=false \
-e SELENIUM_GRID__MAX_BROWSER_INSTANCES=4 \
-e SELENIUM_GRID__SE_NODE_MAX_SESSIONS=1 \
-e KUBERNETES__KUBECONFIG=/kube/config-local-k3s \
--add-host=host.docker.inte.rnal:host-gateway \
-e KUBERNETES__USE_HOST_DOCKER_INTERNAL=true \
-e KUBERNETES__CONTEXT=k3s-selenium-grid \
-e KUBERNETES__NAMESPACE=selenium-grid-dev \
-e KUBERNETES__SELENIUM_GRID_SERVICE_NAME=selenium-grid \
ghcr.io/catchnip/mcp-selenium-grid:latest
Note: All environment variables have default values.
MCP Server configuration (mcp.json)
{
"mcpServers": {
"mcp-selenium-grid": {
"url": "http://localhost:8000",
"headers": {
"Authorization": "Bearer CHANGE_ME"
}
}
}
}
{
"mcpServers": {
"mcp-selenium-grid": {
"url": "http://localhost:8000/mcp",
"headers": {
"Authorization": "Bearer CHANGE_ME"
}
}
}
}
{
"mcpServers": {
"mcp-selenium-grid": {
"url": "http://localhost:8000/sse",
"headers": {
"Authorization": "Bearer CHANGE_ME"
}
}
}
}
🤝 Contributing
For development setup, testing, and contribution guidelines, please see CONTRIBUTING.md.
📄 License
MIT
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。