// Docs

KYH & importing your history.

Two ways to make your clone less generic, fast. The interview captures where you are going. The imports capture where you have been. Do both.

KYH: Know Your Human

KYH is a gentle interview your clone runs to understand you: your goals, your values, what energizes you, what drains you, how you decide. It asks one question at a time, like a good listener, not a form. You just talk; it does all the writing.

Three things to know:

It is skippable. Say "later" during setup and nothing is lost.
It is resumable. Stop mid-interview whenever you want. Run /klyr:kyh anytime and it continues where you left off, without re-asking what it already knows.
It is worth repeating. Your goals move. A short session every few weeks keeps your clone sharp about who you are now, not who you were at setup.

Anything sensitive stays under your control: before writing something durable about you, your clone names it in plain words and asks. Any question can be skipped. The session ends the moment you want.

Importing your history

If you have a past with ChatGPT or Claude, your clone can inherit it. Imports become dormant memory: searchable, and dated with their real timestamps, not the import date. A conversation from March 2024 stays a conversation from March 2024.

Five imports are available:

ChatGPT: export your data from ChatGPT (you get a file called conversations.json), then ask your clone to import it.
Claude: same idea with a Claude export.
Claude Code: your past Claude Code sessions on this machine can be swept into memory too.
Gemini: same idea with a Google Takeout export of your Gemini activity.
OpenClaw: your OpenClaw history can be inherited as well.

Request your exports (do this early)

The exports take time to arrive, so request them before you need them:

ChatGPT — in ChatGPT: your profile → Settings → Data controls → Export data → Confirm export. OpenAI emails you a download link. It can take from an hour to a few days, and the bigger your history, the longer it takes.

The one trap: that download link dies 24 hours after the email arrives. When the email lands, download the zip right away and keep it (Desktop works). The downloaded file never expires.

Claude — on claude.ai: your initials → Settings → Privacy → Export data. The email usually arrives fast. Same reflex: download the zip and keep it.

If an export arrives after you wanted it, nothing is lost: importing it later takes two minutes — ask your clone to import the file, then ask it to refine what it knows about you with what it just learned.

Always dry-run first

Every import runs a dry run before touching anything: it reads your export, counts the conversations, shows the date span, and writes nothing. You see exactly what would be imported before you commit. After a real run, it prints a report of what landed.

The same protection applies here as everywhere in memory: credential-shaped strings (passwords, keys, tokens) are redacted before anything is stored.

Import or interview? Both.

An import gives your clone your past. It does not replace KYH: history is what happened; the interview is what you want next. The strongest start is a short KYH session plus your imported history behind it.