screenshot_mcp_server

This MCP server provides a screenshot capability for AI tooling. It captures the user’s screen, compresses the image to JPEG, and delivers the data as a base64-encoded payload suitable for processing by AI assistants. Two transport modes are supported: stdio for command-line usage and SSE for web-based integrations. You can test the server using the provided CLI client or integrate programmatically via the MCP client interfaces to request a screenshot through the take_screenshot tool.

Prerequisites:

• Python 3.10 or newer
• Git
• uv (the Python package manager for UV runtimes) or a standard Python environment

From Source:

1. Clone the repository git clone https://github.com/codingthefuturewithai/screenshot_mcp_server.git cd screenshot_mcp_server

2. Install the package in editable mode (recommended via UV) uv pip install -e .

3. Alternatively, install with pip (editable) pip install -e .

Usage after installation:

• Run in stdio mode (default): use the provided CLI or integrate via MCP clients as shown in the usage example.

Notes and tips:

• The server supports two transport modes: stdio (default) and SSE (for web-based apps). SSE typically runs on port 3001 by default.
• Image data is JPEG-compressed and base64-encoded for efficient transfer.
• Ensure you have the necessary screen capture permissions on your OS (especially on macOS and Linux with Wayland/X11).
• Required dependencies include mcp, pyautogui (for screen capture), and Pillow (image processing).
• If you encounter permission or display issues on headless environments, consider using a virtual display or enabling an X server/Wayland session compatible with pyautogui.
• Environment variables are not strictly required for basic operation, but you can configure logging, ports, and transport options as needed in your deployment environment.