ScreenMind is a privacy-first AI screen memory that runs 100% locally. It captures your screen periodically, analyzes what you're doing using Gemma 4 via llama.cpp, and builds a searchable timeline of your entire digital day — all without sending a single byte to the cloud.

Ask natural language questions like "what was that API endpoint I was looking at yesterday?" or "how much time did I spend on VS Code this week?" and get instant answers from your own screen history.

Key features:

• Local AI analysis with Gemma 4 (4B) — no API keys, no cloud

• Natural language search across your screen history

• Smart timeline with activity categories and summaries

• Screenshot encryption with OS keyring integration

• Meeting transcription (local whisper)

• Obsidian and Notion sync

• MCP server for Claude Desktop / Cursor / VS Code

• Cross-platform: Windows, macOS, Linux

Install in one command: pip install screenmind