MCP Terminal Server
Provides cross-platform terminal access through MCP, enabling AI assistants to create and manage interactive terminal sessions, execute commands, and capture visual snapshots on Windows, Linux, and macOS.
README
🖥️ MCP Terminal Server
Interactive terminal sessions via Model Context Protocol (MCP)
MCP Terminal Server provides cross-platform PTY (pseudo-terminal) support for Windows, Linux, and macOS, exposing terminal sessions through both REST API and MCP protocol. Perfect for AI assistants, remote terminal access, and terminal automation.
✨ Features
- 🌍 Cross-platform: Works on Windows (cmd.exe/PowerShell), Linux, and macOS
- 🔄 Real PTY: Supports interactive commands (vim, nano, htop, etc.)
- 👁️ Visual Snapshots: Captures what a human would see on the terminal screen
- 🔤 UTF-8 Support: Handles special characters and emojis correctly
- 🚀 MCP Protocol: Auto-exposes all endpoints as MCP tools via fastapi-mcp
- 📡 REST API: Full HTTP/REST API with FastAPI
- 🎯 Multiple Sessions: Manage multiple terminal sessions simultaneously
📦 Installation
From PyPI
pip install mcp-terminal
From Source
git clone https://github.com/alejoair/mcp-terminal
cd mcp-terminal
pip install -e .
🚀 Quick Start
Start the Server
# Start on default port (8777)
mcp-terminal
# Start on custom port
mcp-terminal --port 9000
# Development mode with auto-reload
mcp-terminal --reload
# Custom host and port
mcp-terminal --host 0.0.0.0 --port 8888
Access the Server
Once running, you can access:
- API Documentation: http://localhost:8777/docs
- MCP Endpoint: http://localhost:8777/mcp
- Health Check: http://localhost:8777/health
📖 Usage
REST API
Create a Terminal
curl -X POST http://localhost:8777/terminals \
-H "Content-Type: application/json" \
-d '{"rows": 24, "cols": 80}'
Response:
{
"success": true,
"terminal_id": "550e8400-e29b-41d4-a716-446655440000",
"message": "Terminal created successfully"
}
Send Commands
curl -X POST http://localhost:8777/terminals/{terminal_id}/input \
-H "Content-Type: application/json" \
-d '{"data": "echo Hello World\n"}'
Get Visual Snapshot
curl http://localhost:8777/terminals/{terminal_id}/snapshot
Response:
{
"terminal_id": "550e8400-e29b-41d4-a716-446655440000",
"display": "C:\\Users\\...\nHello World\nC:\\Users\\...> ",
"lines": ["C:\\Users\\...", "Hello World", "C:\\Users\\...> "],
"cursor": {"row": 2, "col": 15},
"size": {"rows": 24, "cols": 80},
"is_alive": true,
"created_at": "2024-01-01T00:00:00"
}
List Terminals
curl http://localhost:8777/terminals
Resize Terminal
curl -X PUT http://localhost:8777/terminals/{terminal_id}/resize \
-H "Content-Type: application/json" \
-d '{"rows": 30, "cols": 100}'
Close Terminal
curl -X DELETE http://localhost:8777/terminals/{terminal_id}
MCP Tools
When using with MCP clients (like Claude Desktop), the following tools are automatically available:
create_terminal_terminals__post- Create new terminallist_terminals_terminals__get- List active terminalsget_terminal_snapshot_terminals__terminal_id__snapshot_get- Get visual snapshotsend_terminal_input_terminals__terminal_id__input_post- Send commandsresize_terminal_terminals__terminal_id__resize_put- Resize terminalclose_terminal_terminals__terminal_id__delete- Close terminal
Python API
from mcp_terminal import TerminalManager
# Create manager
manager = TerminalManager()
# Create terminal
terminal_id = await manager.create(rows=24, cols=80)
# Send input
await manager.send_input(terminal_id, "echo Hello\n")
# Get snapshot
snapshot = await manager.get_snapshot(terminal_id)
print(snapshot["display"])
# Close terminal
await manager.close(terminal_id)
🏗️ Architecture
src/mcp_terminal/
├── __init__.py # Package initialization
├── __main__.py # CLI entry point
├── server.py # FastAPI application with MCP integration
├── core/
│ └── terminal/
│ ├── session.py # TerminalSession - PTY management
│ ├── buffer.py # TerminalBuffer - Visual screen capture
│ └── manager.py # TerminalManager - Multi-session coordinator
└── models/
└── terminal.py # Pydantic models for API
🔧 Configuration
Command Line Options
--host HOST Host to bind to (default: 127.0.0.1)
--port PORT Port to bind to (default: 8777)
--reload Enable auto-reload for development
--log-level LEVEL Set log level (debug, info, warning, error, critical)
--version Show version and exit
Environment Variables
You can also configure via environment variables:
export MCP_TERMINAL_HOST=0.0.0.0
export MCP_TERMINAL_PORT=9000
export MCP_TERMINAL_LOG_LEVEL=debug
🧪 Testing
# Install dev dependencies
pip install -e ".[dev]"
# Run tests
pytest
# Run with coverage
pytest --cov=mcp_terminal
🤝 Contributing
Contributions are welcome! Please feel free to submit a Pull Request.
- Fork the repository
- Create your feature branch (
git checkout -b feature/amazing-feature) - Commit your changes (
git commit -m 'Add amazing feature') - Push to the branch (
git push origin feature/amazing-feature) - Open a Pull Request
📄 License
This project is licensed under the MIT License - see the LICENSE file for details.
🙏 Acknowledgments
- Built with FastAPI
- Uses terminado for cross-platform PTY support
- Uses pyte for terminal emulation
- MCP integration via fastapi-mcp
📚 Resources
🐛 Bug Reports
If you find a bug, please open an issue on GitHub.
💬 Support
For questions and support, please use GitHub Discussions.
Made with ❤️ by the MCP Terminal team
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。