Where do the skill folders go?

Anywhere pi reads: ~/.pi/agent/skills/ or ~/.agents/skills/ globally, .pi/skills/ or .agents/skills/ inside a project. npx skills add Memories-ai-labs/Luci-skills -g writes to a global location; drop the -g to keep them in one project.

Does my screen data reach whatever model pi is pointed at?

pi can be pointed at many providers, which makes this worth stating plainly. 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.

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 the pi Coding Agent a Memory

pi keeps a small tool set and loads skills only when they apply. That is exactly how a memory should be reached. Install the official LUCI skills and pi can search what you saw and what was said on this machine, without carrying any of it around.

See all integrations

pi is a deliberately small terminal agent: four core tools, a short system prompt, and skills that load only when they match the task. It also has no MCP, on purpose. The project's answer to extra capability is a skill with a README, not a protocol.

That makes pi the simplest agent to connect to LUCI, because skills are the path we recommend anyway. Put the two official skills in a folder pi reads and it can search your screen history and meeting transcripts from the next session on.

pi looks in ~/.pi/agent/skills/ and ~/.agents/skills/ globally, and in .pi/skills/ or .agents/skills/ inside a project. The ~/.agents/skills/ folder is shared with Zed and Amp, so one copy there covers three agents.

What pi can do with your memory

Skills only, which is the recommended path

pi ships without MCP and nothing is missing. The luci skill tells pi when to search your memory, and the distill skill turns a finished day into a record in your Life folder.

Nothing loaded until it is needed

pi advertises a skill and loads the body on demand, so your memory costs no context until pi actually goes looking.

Text and meaning

Exact error strings when you remember them, search by meaning when you do not. Results come back with the app, window, and time they came from.

What was said, not only what was shown

LUCI transcribes microphone and system audio on this machine, so pi can search a call the same way it searches a window.

Try asking

Questions that work on day one

What was I looking at before I started this refactor?
Search my screen memory for the exact error text from this morning
Find the moment in today's call where we picked the queue
distill my life

Set up pi

01
Turn on Screen Memory in LUCI
Install LUCI on your Mac or Windows machine, enable Screen Memory, and leave the app running.
02
Install the official skills
LUCI's in-app installer covers Claude Code, Cursor, and Codex. For pi, install from the official repo, or ask pi to install it for you: https://github.com/Memories-ai-labs/Luci-skills
```
npx skills add Memories-ai-labs/Luci-skills -g
```
03
Check that pi sees them
Start pi and call the skill by name, or just ask what you were looking at earlier. Skills are ordinary folders holding a SKILL.md, so if the folder is in place, pi finds it.
```
/skill:luci
```

Where pi 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.

Global skills folders: ~/.pi/agent/skills/, ~/.agents/skills/
Project skills folders: .pi/skills/, .agents/skills/
Settings file: ~/.pi/agent/settings.json
MCP support: None, by design. Skills are the path.

Frequently asked questions

No. Skills are the path we recommend for every agent, and pi implements them fully. The luci skill searches the same local memory an MCP client would query, and pi loads it only when a question needs it.