How Froots keeps your agents oriented — a local, searchable record of everything they've seen and done.

Memory is Froots' answer to the goldfish problem: every conversation, note, and agent run lands in a local memory graph on your machine — plain files you can open, grep, and back up. When you (or an agent) start new work, Froots retrieves the most relevant pieces automatically, so an agent that wakes up Monday already knows what you were working on Friday. Agents can also search memory explicitly, and you can browse and edit everything it holds.

This page is an outline — the full guide is being written. It will cover:

What gets written to memory, and when
How retrieval picks what an agent sees
Browsing, editing, and pruning memories
Backing up and exporting the graph

In the meantime, How the memory graph works on the blog covers the design in depth, and the Memory page has the short version.