Previously on you.
Episodic memory, organized by time
Previously keeps a single timeline of your AI interactions instead of scattering them across chat threads. Every conversation, task, and event lives in chronological order — just like your own memory.
Slices — Slices are time-bound snapshots — hours, days, or weeks of activity grouped into cohesive episodes you can browse, search, and revisit.
Strands — Strands weave related slices together across time, connecting the dots between conversations, projects, and ideas that thread through your day.
Scattered threadsSingle timeline
How recall works
Recall happens in two phases. First, a broad temporal search surfaces candidate slices across your timeline. Then, semantic matching within those slices pinpoints the exact memory you need — without digging through endless threads.
Learn more about recallDeeper reading
MIT License
Open source, self-host
Previously is MIT-licensed and designed to be self-hosted. No cloud lock-in, no telemetry, no surprises. Deploy on your own infrastructure or contribute on GitHub.