screenpipe

screenpipe is an open-source Application that creates a searchable memory of a user's computer activity. It continuously captures screen content and audio to allow users to recall information, transcribe meetings, and automate workflows based on their digital history. The software is deployed as a desktop application for macOS and Windows, or via a CLI for Linux, ensuring that the recording process remains local and private.

It operates as a local-first system where data is stored on the user's machine rather than in the cloud, supporting offline functionality and optional encryption at rest. By recording everything the user sees, says, or hears, it transforms the computer into a personal AI assistant that provides a complete audit trail of work history. This approach allows users to find specific details from past conversations or documents using natural language queries.

Key features

• Event-driven screen capture using the accessibility tree with OCR fallback
• Local audio transcription of system and microphone input via Whisper
• Natural language search across screen text and audio transcriptions
• Visual timeline view for scrolling through screen and audio history
• AI agents called Pipes that trigger actions based on work activity
• MCP server integration for providing context to AI assistants
• Deterministic data permissions for controlling agent access to specific apps
• Speaker identification and diarization for recorded conversations
• PII model for filtering sensitive data from recordings

The system is designed for knowledge workers, researchers, and developers who require a high-fidelity record of their activity without sacrificing privacy. It integrates with AI coding assistants by providing them with real-time context of the user's current task. The architecture minimizes resource impact, typically utilizing 5-10% CPU and 0.5-3GB of RAM through event-driven capture rather than constant video recording. This ensures that the software can run in the background without interrupting the primary workflow of the user.

It is a source-available, privacy-centric tool for personal AI memory and productivity automation.