LUCI Desktop

the retention and on-disk answers below cover your own data well. the case I'm curious about is other people's: a screen share where a coworker's Slack DMs are visible in the background, or a video call where someone else's face and name are on screen. that's not really "your history" anymore, it's theirs, captured without them opting into anything. is there any redaction or exclusion for windows/apps you don't own the content of, or is the assumption that if it crossed your screen it's fair game to record?

very nice product! but before installing it, i'd like to ask: do u really plan to keep it free forever?

Hi everyoneShawn here, founder and CEO of Memories.ai. We build memory for AI, and Luci Desktop is the most personal piece of that work.I built the first version for myself. I run a company across two time zones, and by evening I could not tell you which tab, doc, or call held the thing I needed. My agents were worse off. Claude Code, Cursor, and Codex are brilliant inside one window and blind to everything outside it.Luci fixes that with one idea: your screen is already the common layer across every app. Luci captures it, transcribes your meetings, and keeps all of it on your own machine. Your agents connect over a local MCP server, and there is a CLI if you would rather script it. No connectors, no per-app setup, nothing leaves your device.It is free on Mac and Windows: luci.memories.aiWhat I would love from you today: tell me the one thing you wish your agent remembered and it does not. Those answers decide what we build next. Thanks for being here on launch day.

What did you learn in early pilots about whether people want a “distilled memory” (summaries) versus a “verbatim record” (searchable screen + transcript) — and what concrete moments convinced you the record-first approach was the right tradeoff?

Congrats on the launch! Question from a heavy Claude Code user: I usually have two sessions running side by side on different parts of the same project. Can both query Luci at the same time without stepping on each other? And more interesting for me, would session B be able to recall what session A was doing on screen, so Luci works as shared context between them?

Registration is full, I was waitlisted 😭

Congrats on the launch! Is it possible for me to ask it summarize for example the tasks I worked on last week? If so, this would be absolutely useful for my daily workflow!

I would probably want different retention rules for meetings and general screen activity. Are those settings configurable separately?

​just tested it on mac to find a doc link I forgot to bookmark earlier pulled it up in seconds.

​Most tools push everything to the cloud so a purely local memory engine is a breath of fresh air.

​Usually i skip screen recording tools due to battery drain but local first architecture makes this worth trying.

​can we query the local index through CLI scripts or only via connected AI agents?

I could see meeting history helping someone confirm a detail they missed while presenting.

Hi everyone, I'm Runze, one of the makers of Luci.

By day I'm juggling a dozen browser tabs, back-to-back calls, shared docs, and product threads — the kind of screen life where your context is constantly somewhere else. That chaos is exactly what led me here.

We kept hitting the same wall with AI agents: every conversation starts from zero. You explain the project again, paste the same links again, describe the meeting again. Meanwhile the answer was already on your screen an hour ago.

Luci records what happens on your screen, keeps it on your own machine, and makes that history available to the agents you already use. Ask about a meeting you sat through, a post you scrolled past, or a detail buried in a file you opened last week. No connectors, no per-app integrations.

Everything stays local. We also integrate with Microsoft's Foundry Local so audio transcription and daily summaries run on your own hardware.

If this sounds like your day, I'd like to hear where your agents lose context most often.