Screenpipe

Screenpipe continuously records the visual and auditory output of a computer, storing the data locally and making it searchable with natural‑language queries. It captures the full accessibility tree, performs OCR when needed, transcribes speech, identifies speakers, logs keyboard input, and tracks application switches. The recorded material is indexed so that users can ask questions such as “what did I see in the last five minutes?” or request summaries of daily conversations, with the processing performed entirely on the device.

The tool is aimed at knowledge workers, developers, and researchers who need a private, searchable record of their screen activity without relying on cloud services. By keeping all recordings on‑premises and offering optional encryption, it preserves user privacy while allowing integration with AI assistants that can use the captured context to enhance coding, note‑taking, or task automation.

Screenpipe is an open‑source project released under the MIT license, offering a desktop application and a command‑line interface. It runs offline, consumes modest system resources, and includes filters for windows, applications, and sensitive information such as passwords or personal identifiers.