Memmy Agent

Memmy solves a very specific frustration: every AI assistant you use starts each session without knowing anything about you. Instead of maintaining separate chat histories in Cursor, Claude Code, Codex and browser-based chatbots, Memmy builds one personal memory layer that all of them can draw from. With your explicit permission it reads supported local conversation histories, extracts the durable bits — your preferred tech stack, style rules, past decisions and ongoing project state — and stores them in searchable memory on your own machine.

The injection step is what makes it practical. When you switch tools, Memmy supplies only the memory relevant to the current task rather than dumping your entire history into the prompt window. That keeps context windows clean while preserving continuity, whether you are picking a project back up after a week or handing a decision made in one assistant over to another.

Memmy is more than a passive memory store. It is also an agent you can task directly: combining plans discussed across different agents, turning this week's technical decisions into a document, or organizing raw material. Data handling is deliberately conservative — local storage by default, permission-first collection that you can stop at any time, and per-source and per-tool controls. It is free and open source, available as a macOS and Windows desktop app plus a CLI, and it ships with a generous free token trial rather than a paywall.