PeepIt

PeepIt is an MCP server that provides macOS screen capture and visual analysis capabilities to AI assistants and development environments. Built specifically for macOS 14.0 or later, PeepIt connects MCP clients like Cursor and Claude Desktop to Apple ScreenCaptureKit, allowing agents to capture full displays, targeted applications, or specific application windows without altering window focus or interrupting user workflows. Software engineers, UI designers, and automated testers use PeepIt to provide contextual visual input directly to their AI assistants. Beyond basic screen capture, PeepIt integrates with both local and cloud-based vision models, including OpenAI GPT-4o and local Ollama models like LLaVA and Qwen2-VL. This architecture allows the server to inspect active user interfaces, identify UI bugs, answer questions about visible application states, and process visual screen context. Users can isolate windows by title, process ID, or active focus while filtering out background browser helper processes. By handling screen capture at the system level and passing visual representations to AI models, PeepIt enables agents to inspect running software and verify graphical output entirely through standard MCP tool calls.

Category: Design, Media & Creative

Tags: image-analysis, macos, screenshots, vision

Visit PeepIt

How to install and configure PeepIt

1. Ensure your system runs macOS 14.0+ and Node.js 20.0+. 2. Grant Screen Recording permission to your MCP client (such as Claude Desktop or Cursor) in macOS System Settings under Privacy & Security > Screen & System Audio Recording. 3. Add the server to your client configuration (e.g., ~/Library/Application Support/Claude/claude_desktop_config.json): json { "mcpServers": { "peepit": { "command": "npx", "args": ["-y", "@mantisware/peepit-mcp"], "env": { "PEEPIT_AI_PROVIDERS": "openai/gpt-4o,ollama/llava:latest", "OPENAI_API_KEY": "your-openai-api-key-here" } } } } 4. If using local analysis, run ollama serve and pull a model with ollama pull llava:latest. 5. Restart your MCP client to load the tools.

What you can do with PeepIt

• Debugging frontend user interfaces by capturing specific application windows without stealing focus and passing the image to AI models for visual inspection. * Auditing open browser web pages and styling elements by asking vision models targeted layout questions directly from an MCP client prompt. * Inspecting system processes and active desktop screens by identifying target windows through fuzzy title matching or application process IDs. * Analyzing graphical software state locally without sending confidential visual data off the machine by routing captures to Ollama vision models. * Capturing active error dialogues and UI bug states during software testing runs to automatically document visual software discrepancies.

Key facts

• https://github.com/MantisWare/peepit
• Design, Media & Creative, Files, Documents & PDFs
• image-analysis, macos, screenshots, vision

Part of MCP Servers

What can PeepIt do?

PeepIt allows AI agents to capture and analyze macOS screens, application windows, or specific processes without disrupting window focus. It uses Apple ScreenCaptureKit for image retrieval and integrates with vision models like OpenAI GPT-4o or Ollama-hosted models to interpret visual layouts and UI elements.

Which operating systems support PeepIt?

PeepIt is strictly a macOS-only tool. It requires macOS 14.0 Sonoma or later because it relies directly on Apple ScreenCaptureKit APIs and Swift CLI tooling. It also requires Node.js version 20.0 or higher to run the MCP server package.

Which MCP clients work with PeepIt?

PeepIt works with any Model Context Protocol compliant client that runs on macOS and supports stdio transport. Common tested clients include Claude Desktop and Cursor IDE. You can also test the server directly using the Model Context Protocol Inspector tool.

How does PeepIt handle local AI vision models?

PeepIt supports local vision analysis through Ollama integration. By setting the PEEPIT_AI_PROVIDERS environment variable to models such as ollama/llava:latest or ollama/qwen2-vl:7b, PeepIt routes screen captures to your local Ollama instance, ensuring visual data remains entirely on your machine.

What permissions does PeepIt require on macOS?

PeepIt requires Screen Recording permissions within macOS Privacy and Security settings for the host application, such as Claude Desktop, Cursor, or Terminal. Accessibility permissions are also optionally recommended to assist with full window discovery and targeting across desktop spaces.