Screenpipe is an open-source local AI data layer that continuously captures and indexes information from a user's screen and audio so it can be searched or used by personal AI applications. The project turns day-to-day computer activity into a structured, queryable history while emphasizing local processing and user control.

Its pipeline can record screen content, perform OCR or related extraction, transcribe audio, and expose the resulting timeline through APIs and developer tools. Applications can build on this data for personal search, meeting recall, productivity analysis, automation, and context-aware assistants without each app implementing its own capture system.

Screenpipe is aimed at developers and power users building personal AI around real computer activity. Because the data can be highly sensitive, its local-first and self-hostable approach is an important part of the design, giving users more control over what is captured and how it is consumed.