🖼️ MCP Screenshot Server
About
A lightweight MCP-compatible Python server for capturing Windows screenshots via REST API. Supports full screen, region-based, or window-specific captures. Ideal for AI agent integrations and automation workflows.
Details
- Author
- margusmartsepp
- Downloads
- 166
- Categories
- Automation
Jump to
- Capture full-screen screenshots
- Capture specific windows by title
- Capture custom screen regions
- MCP-compliant REST API
- Returns images as PNG or base64
- Built with FastAPI for production readiness
Setting up with Highlight
This MCP is not yet compatible with Highlight’s one-click setup. However, you can still use it with Highlight by following these steps:
- Download and install Highlight from highlightai.com/download
- Navigate to the plugins tab and select "Add Custom Plugin"
-
Configure the plugin with the settings below
Plugin Name
🖼️ MCP Screenshot ServerCommand (node, npx, python, etc.)Please refer to the README for specific instructions on how to obtain API keys or other required environment variables.
- Enable "Start Automatically" if you want the plugin to start when Highlight launches
From the repository
Clone the repository, create a Python virtual environment, install dependencies from requirements.txt, and run the server with uvicorn main:app --reload. Send a POST request to the /screenshot endpoint with an optional JSON body containing region, window_title, or base64 parameters.
Claude Desktop / Cursor
Paste into your MCP client config file to install this server.
{
"mcpServers": {
"\ud83d\uddbc\ufe0f mcp screenshot server": {
"mcp-screenshot-server-margusmartsepp": {
"command": "python",
"args": [
"-m",
"venv",
".venv"
]
}
}
}
}McpServers
{
"mcp-screenshot-server-margusmartsepp": {
"command": "python",
"args": [
"-m",
"venv",
".venv"
]
}
}🖼️ MCP Screenshot Server
A lightweight, MCP-compatible screenshot microservice built with FastAPI for Windows.
It allows AI agents and automation tools to capture full-screen, region-based, or window-specific screenshots via simple HTTP calls.
---
🔧 Features
- 📸 Capture full-screen screenshots
- 🪟 Capture specific window by title
- 🔲 Capture custom regions [x, y, width, height]
- 🧠 MCP-compliant REST API
- 🖼️ Returns images as PNG or base64
- 🚀 Built with FastAPI, ready for production or LLM use
---
🧠 Use Cases
- Integrating with LLMs using Model Context Protocol (MCP)
- QA test automation pipelines
- Monitoring and remote capture tools
- Visual logging/debugging tools for agents
---
📦 Installation
git clone https://github.com/yourusername/mcp-screenshot-server.git
cd mcp-screenshot-server
python -m venv .venv
source .venv/bin/activate # or .venv\Scripts\activate on Windows
pip install -r requirements.txt
uvicorn main:app --reload
---
🔌 API Usage
POST /screenshot
Request JSON body:
{
"region": [0, 0, 1280, 720], // optional
"window_title": "Untitled - Notepad", // optional
"base64": true // optional (default: false)
}
Response (base64 mode):
{
"status": "ok",
"mode": "region",
"image_format": "base64",
"image": "<base64-encoded-image>"
}
---
🛠️ Tech Stack
- Python 3.11+
- FastAPI
- mss or pyautogui for screenshot
- pillow for image processing
- pygetwindow for window matching (optional)
---
📄 License
MIT License.
Feel free to use, fork, and integrate — commercial or personal.
See LICENSE for details.
---
📬 Contributing
Pull requests and issues welcome!
Open a PR to add features or improve compatibility across platforms (e.g., Mac/Linux support).
---
🙋 FAQ
- Does it work on Linux/macOS?
Not yet. This version is Windows-focused, but you’re welcome to extend it.
- Is it MCP-certified?
This project aims to follow the MCP spec as closely as possible for maximum compatibility with LLM agents.
---
🧠 Inspired By
- Anthropic’s Model Context Protocol
- Real-world automation use cases powered by LLMs and Python
---
```
Would you like me to tailor a specific section to emphasize AI agent use (e.g., “how to use with o1 or GPT-4o via plugin”)?
Sign in to leave a review
Use Google, GitHub, or an email account so ratings stay tied to real people.
No reviews posted yet.
