Guides · all free

Personality tuning, agent by agent.

Find setup steps, editable templates, research and developer resources for the app you use.

Start here

Start here
How to give your AI a personality.
The method: choose preferences → review → install · read →
 Research
What personality type is ChatGPT?
GPT-6, Claude, Grok and Muse in September 2026, plus the May reports · read →

Meta Muse

Explainer
Meta Muse, explained.
What it is, where it runs, what it costs, the three files · read →
 Setup
Give Muse a character: the Soul.md guide.
The tap path, what to paste, how to test and reset it · read →
 Templates
Twenty Soul.md templates.
Four downloads and sixteen role blocks · read →
 Files
Muse characters, explained.
Soul.md, Identity.md, Memory.md and three characters to paste · read →
 Comparison
Muse vs ChatGPT vs Claude.
Where the personality and the memory live in each · read →
 Research
What personality type is Muse?
Muse Spark 1.3, as reported: ISTJ 80, Helper 78 · read →

Templates and developer resources

Developers
Enneagram developer resources and rights.
Synthetic nine-group demo, JavaScript and instrument rights →

By agent

ChatGPT
Custom instructions for all 16 MBTI types.
Paste-ready blocks, with compact exports from the generator · read →
 ChatGPT
Make ChatGPT sound like you.
Personality-first vs corpus-first, honestly · read →
 Claude
Change Claude's personality.
Settings, five examples, recorded results and a testing worksheet · read →
 Claude Code
Change Claude Code's personality.
Output styles, the flag that matters, and a 43-style pack · read →
 Claude Opus 5.5
Claude Opus 5.5's personality, tested.
500 test runs, how it compares, and what to paste · read →
 GPT-6 Astra
Give GPT-6 Astra a personality.
Your personal AGENTS.md, and the retired picker · read →
 Grok Bot
Give Grok Bot a personality.
It lives in the Bot's Description · read →
 Claude Code
Annotated CLAUDE.md examples.
Four annotated files + the principles · read →
 Coding agents
The personality layer your AGENTS.md is missing.
CLAUDE.md, Cursor rules, Copilot instructions · read →

By system

Enneagram
Enneagram AI prompts for all 9 types.
Motivation-first tuning blocks · read →
 Comparison
MBTI vs Big Five, for AI purposes.
Evidence and limits from cross-model questionnaires · read →

Free tools

Tool
Custom instructions generator.
Your preferences + your app → editable instructions · open →
 Tool
CLAUDE.md generator.
Three questions → a working starter file · open →

New research and developer resources

Five new analyses, a React starter, a reproduction guide and three test kits. Protocols are labeled separately from completed analyses.

Developer resource · runnable starter
Build a nine-group scoring demo in React + TypeScript
Download a React and TypeScript starter with 36 synthetic inputs, accessible radio groups, strict validation and tied leaders. It is not a personality test. read →
 Research · descriptive reanalysis
Are AI models really INTJ? The tie-breaking effect
335 INTJ labels become 280 fully resolved results when tied axes stay unresolved. Explore the scoring policy in 400 frozen MBTI records. read →
 Research · descriptive reanalysis
Opus 5.5 vs Fable 5.1: which answers differ?
Compare all 170 questionnaire items for Opus 5.5 and Fable 5.1, with response distributions, reverse-scoring examples and downloadable CSV data. read →
 Research · descriptive reanalysis
GPT-6 Astra vs Sol: what the self-reports show
Astra and Sol differ on MBTI margins, Big Five means and DISC scores. Compare 1,000 assessments while accounting for different prompts. read →
 Research · descriptive reanalysis
What does “neutral” mean when an AI takes a test?
Astra flags 1,823 of 2,980 midpoint responses as inapplicable. See why neutral answers cannot all be interpreted as moderate personality traits. read →
 Research · descriptive reanalysis
Does a stable result mean a stable AI personality?
Compare unique answer vectors, modal results and score variation across 2,000 assessments. Stable labels can conceal changing questionnaire answers. read →
 Developer guide · reproducibility
How to reproduce AgentTune’s AI self-report research
Download historical responses and numeric scoring keys. Reproduce reported scores, preserve ties and review the questionnaire provenance correction. read →
 Test kit · protocol, not results
When does Claude follow your preferences? A test kit
Test Claude account and project instructions with 36 planned replies, fixed prompts, pass/fail criteria and a downloadable worksheet. Protocol only. read →
 Test kit · protocol, not results
Does Muse remember Soul.md changes? A persistence test kit
Check a Muse Soul.md preference across new chats, a later day and restoration. Includes exact probes and a 45-row worksheet; no measured results yet. read →
 Test kit · protocol, not results
Do personality prompts beat plain-English preferences?
A controlled 180-reply protocol compares default prompting, explicit rules and the same rules plus an INTJ label. Includes tasks, rubrics and a worksheet. read →

New in September 2026

Claude
The best Claude personality setups.
Seven paste blocks. Pick by the behavior you want, not by a label. · read →
 Claude
How to make Claude more direct.
Three editable blocks, a test task, and what the available pilot can tell us. · read →
 ChatGPT
ChatGPT custom instructions examples, by what you do.
Twenty short blocks for Custom Instructions. Pick a role, paste, edit and test. · read →
 Tests
Personality questionnaires: availability and terms.
What OEJTS, OEPS, ODAT, IPIP-50 and ECR-R measure, how they score, and where they break. · read →
 Reference
An AI personality glossary.
32 terms, one paragraph each, with the page that uses them. · read →
 Research
Does personality tuning change AI answers?
A 60-reply pilot, with an auditable reanalysis. · read →
 Research
What personality type is Claude?
Measured on Opus 5.5 and Fable 5.1 in September 2026, with the earlier versions for context. · read →
 Research
What personality type is Grok?
One self-report from Grok 4.6 in September 2026, with the May 2026 Grok 4.3 records for context. · read →
 Research
What personality type is Gemini?
From the May 2026 study of Gemini 3.1 Pro. Not yet re-tested in September. · read →
Claude Fable 5.1
Fable 5.1: results and tuning ideas.
500 sessions, exact ties and instructions to try →
New research
Astra, Sol, Grok, Opus and Fable.
Five-model study with downloadable response data →