
Screeny
@rohanrav
About Screeny
Screeny is a privacy-focused macOS MCP server enabling AI agents to capture screenshots of user-approved windows
Config
Add this server to your MCP-compatible client using the configuration below.
{
"mcpServers": {
"screeny": {
"command": "uvx",
"args": [
"mcp-server-screeny"
]
}
}
}Tools
No tools detected
Fetch the live tool list by running this server in a temporary sandbox using the button above.
Overview
What is Screeny?
Screeny is a privacy-first MCP server that enables AI agents (such as Claude Desktop or Cursor) to capture macOS screenshots. It runs entirely locally, requires explicit user approval for each window, and deletes screenshots immediately after use.
How to use Screeny?
Install via pipx install mcp-server-screeny or run with uvx mcp-server-screeny. Then configure your MCP host (Claude Desktop or Cursor) with the command mcp-server-screeny (or full path to the executable). After configuration, run mcp-server-screeny --setup (or uvx mcp-server-screeny --setup) to approve which windows can be captured. Grant Screen Capture permission to both your terminal and MCP host application before running setup.
Key features of Screeny
- Only captures user‑approved windows
- All processing stays local on your machine
- Screenshots are temporary and deleted immediately after use
- Works with Claude Desktop and Cursor
- Setup includes interactive or auto‑approval of windows
- Debug mode (
--debug) to test permissions and window detection
Use cases of Screeny
- AI agents describing UI issues from live screenshots of an iOS simulator
- Developers iterating on design by having Cursor capture and analyze screenshots
- Privacy‑conscious users who want an AI agent to access only specific windows
- Automating screenshot‑based reporting without leaving sensitive data on cloud servers
FAQ from Screeny
What are the system requirements?
Python 3.10+, macOS with Screen Capture permission, and your MCP host (Claude Desktop or Cursor) must also have Screen Capture permission.
How does Screeny protect my privacy?
Only windows you explicitly approve can be captured. Screenshots are processed locally and deleted immediately after the AI agent uses them.
What if I get a “spawn uvx ENOENT” error?
Replace "uvx" in your MCP config with the full path to uvx (find it with which uvx).
Why do I see “No approved windows found”?
Run the setup command (mcp-server-screeny --setup or uvx mcp-server-screeny --setup) to approve windows. Also ensure Screen Recording permission is granted to your MCP host application.
Frequently asked questions
What are the system requirements?
Python 3.10+, macOS with Screen Capture permission, and your MCP host (Claude Desktop or Cursor) must also have Screen Capture permission.
How does Screeny protect my privacy?
Only windows you explicitly approve can be captured. Screenshots are processed locally and deleted immediately after the AI agent uses them.
What if I get a “spawn uvx ENOENT” error?
Replace `"uvx"` in your MCP config with the full path to uvx (find it with `which uvx`).
Why do I see “No approved windows found”?
Run the setup command (`mcp-server-screeny --setup` or `uvx mcp-server-screeny --setup`) to approve windows. Also ensure Screen Recording permission is granted to your MCP host application.
Basic information
More Other MCP servers
Blender
ahujasidOpen-source MCP to use Blender with any LLM
🚀 Model Context Protocol (MCP) Curriculum for Beginners
microsoftThis open-source curriculum introduces the fundamentals of Model Context Protocol (MCP) through real-world, cross-language examples in .NET, Java, TypeScript, JavaScript, Rust and Python. Designed for developers, it focuses on practical techniques for building modular, scalable,
FastMCP v2 🚀
jlowin🚀 The fast, Pythonic way to build MCP servers and clients.
Reactive Resume
amruthpillaiA one-of-a-kind resume builder that keeps your privacy in mind. Completely secure, customizable, portable, open-source and free forever. Try it out today!
IDA Pro MCP
mrexodiaAI-powered reverse engineering assistant that bridges IDA Pro with language models through MCP.
Comments