Screenpipe

Screenpipe is built on a straightforward idea: your computer already holds a record of your working day, but it disappears the moment it scrolls off screen. By capturing screen activity and audio locally, Screenpipe turns that stream into a searchable memory — one that both you and your AI tools can query. The pitch is that an assistant is far more useful when it actually knows what you have been doing, rather than starting from a blank context each time.

The technical approach is deliberately open. Screenpipe keeps a local database and exposes a REST API and MCP server, so models such as Claude or Codex — or your own agent — can read from it. Because the history is structured, it also becomes a foundation for workflow discovery: reviewing recurring tasks across apps, seeing where exceptions occur, and deciding what is worth documenting or automating.

Privacy is handled with local-first defaults. Raw history stays on your machine, sensitive fields can be redacted on-device, and you can exclude particular apps, windows or URLs. For organisations, Screenpipe offers an enterprise workspace with device management, SSO and audit logs, so a company can adopt it without centralising everyone's raw screen history. It runs on macOS, Windows and Linux and is source-available, with SOC 2 Type 2, HIPAA, GDPR and CCPA commitments listed for enterprise use.