How LUCI Remembers: Inside Your Life Memory

Why Memory Matters

Most AI assistants forget everything the moment you close the window. Every session starts from zero, and you become the courier — pasting back the error you saw an hour ago, the decision from Tuesday's call, the setting you changed once last week.

LUCI takes a different approach. It remembers your whole day on the computer — meetings, code, email, documents, browsing, everything in between — and keeps that record across sessions, apps, and weeks. Then it gives that memory to whichever AI agent you already use.

One memory for your whole day

At the core of LUCI is your memory: one layer that connects everything you worked on across time, instead of a separate archive per app.

A folder of screenshots is not a memory. What makes recall useful is the connective tissue: the same project showing up in a design doc, a code review, and a call three days apart, recognized as one thread. LUCI Memory holds those threads together, so asking about a project returns the whole arc of it rather than scattered fragments.

Memory with a timeline

Everything LUCI captures carries its moment. That means you can ask by time ("what was I working on Tuesday afternoon"), by topic ("the caching discussion"), or by connection ("what did we decide after that bug came up") — and get an answer anchored to when it actually happened.

Getting things back

Recall is the point: ask in your own words, get back the exact moment.

Two kinds of questions matter, and they need different answers. Sometimes you know the exact string — an error code, a client name, a number quoted on a call — and you need the literal match. Other times you only remember the shape of it: that article about caching I read last week. LUCI handles both, so you don't have to remember how you remembered.

The result isn't a grid of thumbnails to scrub through. It's the specific moment, with the app, the window, and the time around it.

Understanding, Not Just Capture

LUCI doesn't only store what appeared on your screen — it understands it, on your device. That's the difference between finding a string you already knew to search for and asking what you were actually working on.

Meetings work the same way. Every word — yours and theirs — is understood on your device and folded into the same memory as the rest of your day, so a decision made out loud is as retrievable as a line of text.

Readable by the agent you already use

A memory only you can search is half a memory. The other half is the Agent Bridge: Claude, Cursor, and Codex can read the same memory when you ask them to, so the agent already open on your machine knows what you did on Tuesday without you pasting it back in. LUCI connects itself — there is nothing to set up — and your agent reads only what it asks for.

Privacy First

All of it runs on your own machine. Capture, understanding, and storage happen on your device, encrypted at rest, and it keeps working with Wi-Fi off. When you connect an agent, it reads only the part of your memory it asks for — so you choose which agents get to ask.

There's a falsifiable test for this, and we'd rather you run it than take our word: turn off your network mid-meeting and see whether the notes keep coming.

What's Next

We're working on collaborative memory — letting teams share context while keeping individual privacy boundaries intact.