Desktop Automation
Automate desktop actions and interact with your local environment using LLM applications.
Desktop Automation MCP Server
A Model Context Protocol (MCP) server that exposes desktop automation capabilities, allowing LLM applications to interact with your desktop environment through standardized tools.
Overview
This MCP server provides a bridge between LLM applications and desktop automation functionality. It exposes mouse, keyboard, and screen automation capabilities through the MCP protocol, enabling AI assistants to:
- Click, right-click, and double-click at specific coordinates
- Move the mouse cursor with smooth or instant movement
- Type text with optional character delays
- Press individual keys or key combinations
- Get current mouse cursor position
- Capture screenshots of the desktop
- Get screen dimensions
Features
Mouse Automation
- Click: Left click at specified coordinates
- Right Click: Right click at specified coordinates
- Double Click: Double click at specified coordinates
- Move Mouse: Move cursor with optional smooth animation
- Get Position: Retrieve current mouse coordinates
Keyboard Automation
- Type Text: Type text at current cursor position
- Press Key: Press individual keys or key combinations
- Delayed Typing: Type with configurable delays between characters
Screen Automation
- Take Screenshot: Capture the full screen and save to file
- Get Screen Size: Retrieve screen dimensions
Installation
Prerequisites
- Go 1.23.0 or later
- Task runner (optional, for using Taskfile commands)
# Install Task (optional)
go install github.com/go-task/task/v3/cmd/task@latest
Build from Source
# Clone the repository (if not already available)
git clone <repository-url>
cd desktop-automation-mcp
# Download dependencies
go mod download
go mod tidy
# Build the server
go build -o mcp-server ./cmd/mcp-server
# Or using Task
task build
Usage
Running the Server
The server uses stdio transport for communication:
# Run directly
./mcp-server
# Or using Task
task run
Integration with LLM Applications
Configure your LLM application (Claude Desktop, etc.) to connect to this MCP server:
{
"mcpServers": {
"desktop-automation": {
"command": "/path/to/mcp-server"
}
}
}
Available Tools
click
Click at specified screen coordinates.
Parameters:
x(number, required): X coordinatey(number, required): Y coordinate
right_click
Right click at specified screen coordinates.
Parameters:
x(number, required): X coordinatey(number, required): Y coordinate
double_click
Double click at specified screen coordinates.
Parameters:
x(number, required): X coordinatey(number, required): Y coordinate
move_mouse
Move mouse cursor to specified coordinates.
Parameters:
x(number, required): X coordinatey(number, required): Y coordinatesmooth(boolean, optional): Use smooth movement animationduration(number, optional): Duration for smooth movement in seconds (default: 1.0)
get_mouse_position
Get current mouse cursor position.
Parameters: None
type_text
Type text at current cursor position.
Parameters:
text(string, required): Text to typedelay(number, optional): Delay between characters in milliseconds
press_key
Press a key or key combination.
Parameters:
key(string, required): Key to press (e.g., 'enter', 'space', 'ctrl')modifiers(array, optional): Modifier keys (e.g., ['ctrl', 'shift'])
take_screenshot
Capture a screenshot of the screen.
Parameters:
path(string, optional): Path to save the screenshot (if not provided, saves to temp directory)
get_screen_size
Get the screen dimensions.
Parameters: None
Architecture
desktop-automation-mcp/
├── cmd/
│ └── mcp-server/ # MCP server entry point
│ └── main.go
├── internal/
│ └── automation/ # Desktop automation logic (copied from desktop-automation)
│ ├── keyboard.go
│ ├── mouse.go
│ └── screen.go
├── go.mod # Go module definition
├── Taskfile.yml # Task runner configuration
├── .gitignore # Git ignore rules
└── README.md # This file
Development
Task Commands
# Build the server
task build
# Run the server
task run
# Clean build artifacts
task clean
# Download and tidy dependencies
task deps
# Run tests
task test
# Install to GOPATH/bin
task install
Manual Commands
# Build
go build -o mcp-server ./cmd/mcp-server
# Run
./mcp-server
# Test
go test ./...
# Clean
go clean
rm -f mcp-server
Dependencies
- github.com/mark3labs/mcp-go: MCP protocol implementation for Go
- github.com/go-vgo/robotgo: Cross-platform desktop automation library
Safety Considerations
- Screen Bounds: All coordinate inputs are validated against screen dimensions
- Input Validation: Negative coordinates and invalid parameters are rejected
- Error Handling: Comprehensive error handling with descriptive messages
- Recovery: Built-in panic recovery for robust operation
Platform Support
This server supports the same platforms as robotgo:
- Windows
- macOS
- Linux
Contributing
- Fork the repository
- Create a feature branch
- Make your changes
- Add tests if applicable
- Submit a pull request
License
This project follows the same license as the parent desktop-automation project.
Related Projects
- desktop-automation: The original CLI-based desktop automation tool
- mcp-go: Go implementation of the Model Context Protocol
Máy chủ liên quan
Kone.vc
nhà tài trợMonetize your AI agent with contextual product recommendations
Squad AI
Your AI Product Manager. Surface insights, build roadmaps, and plan strategy with 30+ tools.
Fraud Detection Engine
中英双语 AI 欺诈文本检测引擎,可识别诈骗、钓鱼等风险,返回风险评分、判定和等级。
Resume Generator
Generates professional PDF resumes using LaTeX templates.
Tellers.AI - Prompt Based Video Editing
Give video editing skills to your agent
Tmux
Interact with your Tmux sessions, windows and pane, execute commands in tmux panes and retrieve result.
ShipBoss
An intelligent shipping assistant for managing shipments, requiring a ShipBoss API token.
Counsel
Multi LLM Council for deep counsel on your topic or idea
docx-mcp
Read and edit Word (.docx) documents with track changes, comments, footnotes, and structural validation. 18 tools with OOXML-level paraId validation and document auditing.
llmconveyors-mcp
39 tools for the LLM Conveyors AI agent platform. Run Job Hunter, B2B Sales, ATS scoring, resume rendering, and more from any MCP client.
Anytype MCP Server
Interact with the Anytype API using natural language.