Screenpipe

Screenpipe is an open-source (source-available) desktop application that continuously captures your screen and audio, creating a searchable, AI-powered memory of everything you do on your computer. It extracts text via OS accessibility APIs with OCR as a fallback and transcribes audio using local Whisper models, storing all data in a local SQLite database. Designed for individuals and teams who want private, local-first work memory without cloud recorders, it integrates with 100+ apps and lets AI agents search, summarize, and automate workflows based on real activity. Screenpipe runs on macOS, Windows, and Linux, recording multiple monitors and system/mic audio. Recent updates (v2.5.86–v2.5.90) added per-process audio capture for meetings, per-monitor pause control from the tray, optional echo cancellation, and improved notification routing with live generation progress. The built-in redaction engine scrubs PII (credit cards, SSNs, API keys) using a custom on-device AI model that runs on Apple Neural Engine or Windows GPU—0 bytes leave your machine. Key features include natural language chat search across your history, Pipes for automated workflows and SOPs triggered by meetings or activities, MCP server support for agent integration, and a CLI for advanced querying. You can exclude specific apps, windows, or URLs at the source, and exports support meeting summaries and MP4 clips. The vendor also offers SOC 2 materials on request. Compared to cloud-based tools like Rewind.ai, Microsoft Recall, Granola, or Otter.ai, Screenpipe is fully local-first by default and fully auditable (19k+ GitHub stars, Rust core). It’s a strong choice for privacy-conscious power users and teams building AI agents that need real-time desktop context. However, it requires more setup than turnkey cloud recorders and isn’t HIPAA-compliant out of the box (SOC 2 available on request).