Two options · MIT base · cancel anytime

Free is real. Pro is worth it.

The library is MIT — yours forever, no account required. Pro adds the extended assessment, a synthesized master file, and paste-ready integration for every major agent. Most people get value from Free; the people who want one file for all of them upgrade.

Free
Free
The library, forever.
$0 MIT licensed
  • All 5 personality tests (MBTI, Enneagram, DISC, Attachment, Big Five)
  • 43 type-matched tuning files as Markdown
  • Paste-ready snippets for 12 agent surfaces
  • Layer multiple tunings yourself for higher fidelity
  • Self-host, fork, modify, ship in your product
  • No signup, no telemetry, no lock-in
  • Extended 30-min synthesis assessment
  • Personalized master Markdown file
Take a free test →
Always free · always open
Recommended · Coming soon
Pro
One synthesized master file built from your full assessment — plus deep integration guides for every major agent.
$9 once · lifetime access
  • Everything in Free
  • Extended 30-min multi-instrument assessment, including items outside the public library
  • One synthesized master Markdown file — your soul file, tuned to all of you
  • Cross-instrument conflict resolution (when MBTI and Big Five disagree, etc.)
  • Deep integration guides — Claude, ChatGPT, Codex, Cursor, Gemini, Hermes, OpenClaw
  • Free re-tunings forever — your file evolves with you
  • Email delivery + always-accessible private permalink
  • 30-day no-questions refund if it doesn't improve your agent
Join the waitlist — $9 once →
No pre-payment · we email when Pro is live
Same model, same files
Pro doesn't gate access to better tunings. The same .md files power both tiers — Pro just synthesizes them into one personalized master file.
No lock-in, ever
Your tuning is plain Markdown. Cancel, switch agents, fork the repo — your file is still a file. Paste it anywhere.
Free covers most needs
If you're comfortable stacking a few tunings yourself, Free is genuinely complete. Pro exists for the synthesis and the convenience of one file.
FAQ

Short answers to the obvious questions.

What's a "tuning file"?

A Markdown file matched to your personality. Paste it into your AI agent's system prompt and the agent's interaction style extends with how you process — less translation, sharper alignment, fewer cycles spent reformatting answers in your head.

Browse the live chat library to preview any tuning file with a sample response.

Is the free tier actually any good?

Yes. The 43 free tuning files are exactly what we use as the source material for the Pro master file. If you stack a few (e.g., your MBTI + Enneagram + a couple of OCEAN dimensions), the result gets you roughly 80% of the way to Pro.

Pro adds: synthesis into one file, conflict resolution where systems disagree, deeper integration guides, and items outside the public library.

What's actually in the Pro master file?

One Markdown file, ~600–1200 words, including:

  • Your synthesized personality across all 5 instruments
  • Cross-instrument conflict resolution (e.g. when MBTI E and Big Five low-E disagree)
  • Custom-tuned phrases and rules for your specific combination
  • A short "operator's manual" for your agent — what to do, what to avoid, when to ask
  • A versioned changelog so you can see how the file evolves if you re-tune

You own it. It's yours forever, and you can edit, share, fork, or paste it into anything.

Which agents does this work with?

Any agent with a system prompt slot. Confirmed and paste-ready:

  • Claude — Claude Code (CLAUDE.md), Claude.ai (Project Instructions)
  • ChatGPT — Custom Instructions, ChatGPT Projects
  • OpenAI Codex CLI — AGENTS.md in project root
  • Cursor — .cursor/rules/*.mdc
  • Gemini — Gems, Code Assist, Antigravity
  • Hermes CLI — persona system_prompt or --system <file>
  • OpenClaw — AGENTS.md
  • Anything else — paste at the start of any conversation

Take any free test to get the exact paste-ready snippet for each surface.

Why does default AI feel so generic?

Because every frontier model defaults to roughly the same personality. We tested 6 frontier models (Claude, GPT-5.5, Gemini, Grok, GLM, MiniMax) on the MBTI 100 times each — 597 of 600 came back INTJ. Useful baseline. One out of sixteen, though, and it rarely matches the way you actually work.

The full study: /research.

Do you store my data?

Free tier: nothing leaves your device. Tests run client-side; tunings are static files. We don't collect anything.

Pro tier: we store your raw assessment responses encrypted, generate your master file, send it to you, and delete the raw responses 90 days after delivery. Your finalized master file stays available at your private permalink. We never sell, share, or train on your data — full stop.

Can I cancel? Refund policy?

Pro is a one-time $9, not a subscription — there's nothing to cancel. If it doesn't measurably improve how your agent talks to you, email us within 30 days for a no-questions refund. The files you've already generated are still yours.

When does Pro launch?

Soon. We're collecting waitlist signups to gauge interest and finalize the extended assessment items. Join the waitlist and we'll email you the moment it's live. No pre-payment, no commitment.

Is the library MIT-licensed?

The free library on GitHub is MIT. Fork, modify, redistribute, embed in your product — go.

Your Pro master file is your personal IP, not ours. You own it forever and can do whatever you want with it.

Can I contribute to the library?

Yes — see CONTRIBUTING.md. We welcome PRs for tightening existing tunings, adding new personality systems (HEXACO, astrology, etc.), and submitting your own soul file.

I'm not sure which type I am — can I preview a tuning?

Yes — the chat library on the homepage lets you click any of the 43 tunings and see the Markdown content, a sample response, and the system color theme it applies. No commitment, no signup — just browse.

Educational pricing?

Free Pro for students and educators with a verified .edu email when we launch. Email [email protected] from your school address.

MIT · psyduckler/agenttune Two options · zero lock-in · always MIT underneath