Quick start

From an empty machine to your first /load in Claude Code. About five minutes.

1. Install the CLI

curl -fsSL https://memcrate.dev/install.sh | sh

Other platforms and methods: install guide.

2. Scaffold a vault

Pick a folder for your vault. ~/vault is the convention; anything works.

memcrate init ~/vault

You’ll get the minimum viable shape:

~/vault/
├── Core/
│   ├── Context/
│   │   ├── Profile.md
│   │   ├── Projects.md
│   │   └── Current State.md
│   └── Sessions/
└── .memcrate

The .memcrate file is a marker — it tells the CLI and skills “this folder is a Memcrate vault.”

3. Run the setup wizard

memcrate setup

Four prompts: your name, what you do, the AI tools you use, and the projects you’re working on. The wizard writes your answers into Profile.md and Projects.md. You can edit them by hand any time — they’re just markdown.

4. Install the Claude Code skills

memcrate install claude-code

This drops three skill files into ~/.claude/skills/ (or the Windows equivalent):

save/SKILL.md
load/SKILL.md
pin/SKILL.md

Claude Code picks them up automatically — no restart needed.

5. Use the verbs

Open Claude Code in any directory. At the start of a session:

/load

Claude reads your Profile.md, Projects.md, Current State.md, and the three most recent session logs, and gives you a short oriented summary.

Work for a while. When you’re done:

/save

Claude writes a structured session log to Core/Sessions/ — decisions, learnings, files touched, what’s pending.

When something from the conversation is worth keeping permanently:

/pin

Claude figures out which context file the insight belongs in (Profile, Projects, or Current State), edits it, and bumps the last_updated field.

What’s next?

Browse your vault. Open ~/vault in Obsidian, VS Code, or any text editor. You own the files.
Sync across machines. Use Obsidian Sync, iCloud Drive, Dropbox, or git — anything that mirrors a folder. The vault doesn’t care.
Read the references. The vault and the verbs explain the design in more depth.