ScreenMind captures your screen when content actually changes, analyzes every screenshot using Gemma 4's vision model running locally via llama.cpp, and builds a searchable AI memory of your entire computer activity.
Search & Chat Ask questions about anything you've seen on your screen. "What was that error message?" or "What did that Slack message say?" — ScreenMind finds the right screenshot and answers from your actual screen history using semantic search and conversational AI.
Meeting Transcription Automatically detects when you join Zoom, Teams, Meet, or Discord. Records audio, transcribes it using Gemma 4's native audio encoder, and generates structured meeting summaries with action items. No Whisper dependency needed.
Agents in Plain English Write a markdown file describing what you want — "every 6 hours, summarize my coding activity and send to Obsidian" — drop it in a folder, and it runs automatically on your screen data. Python plugins also supported for developers.
Works with Your AI Tools Built-in MCP server lets Claude Desktop, Cursor, and VS Code query your screen history directly. Ask Claude "what was I working on this morning?" and it pulls from your real data.
Privacy by Design Everything runs on your machine. No cloud. No accounts. No telemetry. Sensitive data like credit cards and API keys are auto-redacted before storage. Screenshots can be AES encrypted at rest. Works on a 4GB GPU.
