The product
What screenpipe is building
screenpipe is a desktop memory layer that records changing screen content, application text, browser context, meetings, and system audio, then makes that history searchable by people and AI agents. Raw history is stored locally by default, with accessibility APIs used for structured text and OCR as a fallback. It supports macOS, Windows, and Linux, exposes a local API and MCP server, and can run scheduled agents called pipes against captured work context.
From the maker
Give your AI a searchable memory of what actually happened at work
Most assistants begin each task without the context scattered across meetings, apps, and past screen activity. screenpipe is interesting because it creates a local-first history layer that different models and agents can query, while letting users exclude sensitive sources and keep raw capture on their own computer by default.