Does this work with Claude Desktop too, or just Claude Code?

The official skills install into Claude Code from LUCI or from Memories-ai-labs/Luci-skills. Any MCP client can also connect using the same connection details from the Screen Memory panel.

What exactly can Claude Code query?

Exact strings on screen, fuzzy recall by meaning, captures narrowed by app and time, a whole time range at once ("what was I doing between X and Y"), what was said in meeting audio, and one moment in full detail. See the MCP memory server page for the optional protocol surface.

Do I need to keep LUCI running?

For live history, yes. If LUCI is quit, the luci skill reads your Life folder instead. The memory only grows while LUCI is capturing.

Coding agent

Give Claude Code Memory of Everything You’ve Seen

Claude Code starts every session blind to what you were just doing. Install the official LUCI skills and it can search your screen history, recall the error you hit this morning, and pull up what was said in your last meeting — from a memory that lives 100% on your own machine.

See all integrations

Claude Code is remarkable at working with what’s in front of it — and amnesiac about everything that isn’t. The stack trace you saw an hour ago in another window, the API decision from Tuesday’s standup, the config you changed once last week: none of it exists unless you find it and paste it back in.

LUCI closes that gap with a local screen memory and official skills that teach Claude when to look it up. The skills live in Memories-ai-labs/Luci-skills. MCP is available if you want the tools without the skill files.

What Claude Code can do with your memory

Recall exact strings

Error messages, stack traces, IDs, names — Claude Code can keyword-search everything that has appeared on your screen and get back the matching text with its app, window, and time.

Fuzzy recall, semantically

"That article about caching I read last week" is a real query. Semantic search over your screen history finds things by meaning when you don’t remember the words.

Reconstruct your timeline

Ask what you were doing between two points in time and Claude Code gets an aggregated view of your activity — useful for standups, timesheets, and "where was I?" moments.

Search what was said

LUCI transcribes microphone and system audio locally. Claude Code can search those transcripts, so decisions made in meetings are as retrievable as text on screen.

Try asking

Questions that work on day one

Search my screen memory for the error I saw this morning
What was I working on yesterday between 2pm and 4pm?
Find the part of today’s meeting where we discussed the API migration
What did that pricing page I looked at last week say?

Set up Claude Code

01
Turn on Screen Memory in LUCI
Install LUCI on your Mac or Windows machine and enable Screen Memory. Leave the app running.
02
Install the official skills
In LUCI, open Settings → Agent access, tick Claude Code, and click Install. Or install from the official repo: https://github.com/Memories-ai-labs/Luci-skills
```
claude plugin marketplace add Memories-ai-labs/Luci-skills
claude plugin install luci-skills
```
03
Optional: connect over MCP
If you want Claude to call LUCI as a tool without the skills, copy the ready-made Claude Code snippet from the Screen Memory panel.
```
claude mcp add --transport http luci <url> \
  --header "Authorization: Bearer <token>"
```

Where Claude Code looks for it

The two skill folders are ordinary files you own. Drop them where this tool reads and it finds them on the next session.

Skills folder: ~/.claude/skills/
Written by the LUCI installer: Yes, Settings > Agent access
Optional MCP path: claude mcp add --transport http

Frequently asked questions

The memory itself stays on your machine. What an agent actually pulls becomes part of that conversation. If the agent runs in someone else’s cloud, that conversation goes there. Connect only the agents you trust, and ask them for what you are willing to put in that thread.