Does my screen data get sent to Amp or its model providers?

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.

Skills or amp.mcpServers?

Skills. A tool list on its own does not tell Amp when your history is relevant, and that judgement is the whole point. Use amp.mcpServers when you want the tools without the skill files.

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 Amp Memory of What You Saw and Said

Amp will happily run for an hour on one task, which makes the context it lacks expensive. Install the official LUCI skills and it can look up what was on this screen and what was said on the call instead of asking you.

See all integrations

Amp takes long-running work seriously, and long-running work is where missing context costs the most. An hour into a task, the detail that would have saved it is often something you read this morning and never wrote down.

LUCI keeps that record on this machine and the official skills teach Amp when to consult it. Amp reads skills from .agents/skills in the project and its parent directories, and personal skills from ~/.agents/skills, ~/.config/agents/skills, and ~/.config/amp/skills.

The ~/.agents/skills folder is also read by Zed and pi, so one copy there serves three agents. If you prefer the protocol path, Amp takes a remote MCP server with a url and headers under amp.mcpServers.

What Amp can do with your memory

Context for long tasks

Amp can search this morning's screen history mid-task instead of stopping to ask, so a two-hour run does not stall on a detail you already read.

Personal or per repository

Keep the skills in ~/.agents/skills for yourself, or commit them under .agents/skills so the repository carries them.

Calls are searchable

Microphone and system audio are transcribed on this machine, so Amp can find the sentence where the requirement actually changed.

Remote MCP if you prefer

Amp accepts a url and headers for remote MCP servers, so the details from the Screen Memory panel drop straight in.

Try asking

Questions that work on day one

Search my screen memory for the config I changed last week
What was I doing before this task started?
Find the part of the call where we scoped this work
What errors did I see in the terminal this morning?

Set up Amp

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
Put them in ~/.agents/skills for yourself, or commit them under .agents/skills to share them with the repository: https://github.com/Memories-ai-labs/Luci-skills
```
npx skills add Memories-ai-labs/Luci-skills -g
```
03
Optional: connect over MCP
Add LUCI under amp.mcpServers in ~/.config/amp/settings.json, or in .amp/settings.json for one workspace, using the URL and token from the Screen Memory panel.
```
"amp.mcpServers": {
  "luci": {
    "url": "<url>",
    "headers": { "Authorization": "Bearer <token>" }
  }
}
```

Where Amp 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.

Project skills folder: .agents/skills/
Personal skills folders: ~/.agents/skills/, ~/.config/agents/skills/, ~/.config/amp/skills/
Optional MCP key: amp.mcpServers
Settings file: ~/.config/amp/settings.json, .amp/settings.json per workspace

Frequently asked questions

~/.agents/skills if the memory is yours alone, and it doubles as the folder Zed and pi read. .agents/skills in the repository if you want everyone on the project to have it. Local and built-in skills take precedence over repository skills with the same name.