Screenpipe turns your computer into a personal AI memory system. Built in Rust with Tauri for the desktop app, it captures screen activity and audio using an event-driven architecture that only stores frames when something actually changes; storage and CPU use depend on capture and retention settings, and Screenpipe's pricing page lists about 30 GB per month with configurable retention. Text extraction primarily uses the OS accessibility tree for structured data (buttons, labels, text fields), falling back to OCR when needed. Captured history is stored in a local SQLite database with FTS5 full-text search.
The platform's plugin system called 'Pipes' enables AI agents defined as simple markdown files to run on schedules, querying your captured data via the screenpipe API to perform automated tasks. Pipes available in Screenpipe's pipe store cover meeting notes, CRM updates, time tracking, and expense tracking. Screenpipe also runs as an MCP server, allowing Claude Desktop, Cursor, and other AI assistants to directly query your screen history using natural language.
With 21K+ GitHub stars, Screenpipe has positioned itself as a local-first alternative to Rewind.ai. Its source code is public under the Screenpipe Commercial License, which makes it source-available, not OSI open source: it is free for personal non-commercial, non-profit, educational, and research use, and commercial use requires a paid license; versions released under MIT before the June 2026 license change remain MIT. The official app has a free plan, Basic at $25/month or $250/year, Business at $50 per seat per month or $500 per seat per year, and custom Enterprise pricing, and existing lifetime licenses keep working. It supports macOS, Windows, and Linux with multi-monitor capture, speaker identification, PII redaction, and compatibility with local, cloud, or self-hosted AI models.
Data flow: captured screen, audio, and transcript history is stored on the device by default. Optional cloud AI, cloud transcription, sync, connectors, exports, and team workflows send the data those features need off-device, and AI clients connected through MCP or the local API receive the context they request under their own data policies. Product analytics (PostHog) and crash reporting (Sentry) are on by default; analytics use a stable install ID and are linked to your account, including your email, once you sign in. Both can be turned off in Settings → Privacy → Analytics, and the MCP server's crash reporting can be disabled with an environment variable. Remote support logs are also enabled by default, but nothing is uploaded unless Screenpipe support sends a short-lived request.
