Screenpipe is a Y Combinator-backed, local-first AI memory layer that continuously records your computer activity and turns it into searchable context for your AI agents. The open project has more than 22,000 stars on GitHub and is under very active development, with updates in October 2026.
Screenpipe captures screen frames, on-screen text through accessibility data and OCR, and system and microphone audio, building a timestamped history of your workday that stays on your device by default. You can ask an AI what you saw, decided, or worked on, and it answers from that history. Beyond recall, Screenpipe analyzes recurring work across your apps to map workflows and surface the steps most worth automating, then drafts follow-ups, reviews, and SOPs from real evidence rather than memory.
Screenpipe is also a platform, not just an app. Developers can build small automations, called pipes, on top of the local API and database, and a growing ecosystem of integrations connects the captured history to note-taking tools, CRMs, and other agents. Users stay in control throughout: capture can be paused on a schedule, specific applications and websites can be excluded, and sensitive text can be redacted before it is stored. The combination of total recall and local ownership is what distinguishes it from cloud recording tools that ship your workday off to someone else’s servers.
Screenpipe offers a free tier for local capture, AI search, and meeting notes. Paid plans add longer history, more AI usage, model choices, sync, and team controls, so it is a freemium product.
Screenpipe is for knowledge workers, founders, and teams who want their AI assistants to work from real context โ what actually happened on their computer โ instead of starting every conversation from zero.