Your AI conversations reveal your personality.
KnowMe analyzes your ChatGPT / Claude / DeepSeek chat history to uncover your MBTI type, personality strengths & weaknesses, and actionable life advice.
English | ็ฎไฝไธญๆ
Traditional MBTI tests ask what you think you'd do.
KnowMe analyzes what you actually do โ how you ask questions, how you make decisions, how you structure your thoughts, what topics consume your attention โ all extracted from your real conversations with AI.
No surveys. No self-reporting bias. Just your authentic behavior.
Think about it: you've had hundreds (maybe thousands) of conversations with ChatGPT. Those conversations contain your real thinking patterns and personality fingerprint. KnowMe extracts them.
| Capability | Traditional MBTI Tests | AI Chat Assessments | KnowMe |
|---|---|---|---|
| Based on real behavior (not self-report) | โ | โ | โ |
| Supports ChatGPT / Claude / DeepSeek data | โ | โ | โ |
| 50+ behavioral signal detection | โ | โ | โ |
| Bilingual signal detection (EN + CN) | โ | Partial | โ |
| 100% local โ data never leaves your device | โ | โ | โ |
| Confidence scoring per dimension | โ | โ | โ |
| Personalized advice for all 16 types | โ | Partial | โ |
| Free & open source | โ | โ | โ |
The core difference: we don't ask you to pick answers. We let your behavior answer for you.
# Clone the repo
git clone https://github.com/AIPMAndy/KnowMe.git
cd KnowMe
# Analyze your ChatGPT data (export from Settings โ Data Controls โ Export Data)
python3 scripts/collect.py --source chatgpt --file ~/Downloads/conversations.json --output data.json
python3 scripts/analyze.py --input data.json --output report.md
python3 scripts/advise.py --report report.md --output advice.md
# Read your personality report!
cat report.md# Install the skill
openclaw skill install knowme
# Then just ask your AI assistant:
# "Analyze my personality from our conversations"
# "What's my MBTI type?"
# "Give me personality insights"Your MBTI Type: INTJ (Architect)
E [โโโโโโโโโโโโโโโโโโโโ] I (28% / 72%) โ I โฌ
๏ธ (high confidence)
S [โโโโโโโโโโโโโโโโโโโโ] N (15% / 85%) โ N โฌ
๏ธ (high confidence)
T [โโโโโโโโโโโโโโโโโโโโ] F (78% / 22%) โ T โฌ
๏ธ (high confidence)
J [โโโโโโโโโโโโโโโโโโโโ] P (68% / 32%) โ J โฌ
๏ธ (medium confidence)
Each dimension includes:
- ๐ Percentage scores (not binary labels)
- ๐ฏ Confidence level (high / medium / low)
- ๐ Behavioral evidence (what you said that led to the score)
- Message length patterns (are you concise or thorough?)
- Question / exclamation rates
- Emoji usage frequency
- Topic distribution (tech vs business vs relationships vs self-growth)
- Collaborative vs independent tendencies
- ๐ช Your Superpowers โ Top 3 strengths
- ๐ฑ Growth Edges โ 3 areas most worth developing
- ๐ฏ Career Lens โ What roles and work styles fit you
- ๐ฌ Communication Blind Spots โ Issues you might not notice
- โค๏ธ Relationship Insights โ Your social patterns
- โก 30-Day Challenge โ One specific growth experiment
| Source | Command | How to Export |
|---|---|---|
| ChatGPT | --source chatgpt --file conversations.json |
Settings โ Data Controls โ Export Data |
| Claude | --source claude --file claude_export.json |
Settings โ Account โ Export Data |
| OpenClaw | --source openclaw |
Auto-reads from workspace memory files |
| Any Text | --source text --file ./chats/ |
Any .md/.txt files with User:/Assistant: markers |
๐ก Tip: More data = better analysis. We recommend 50+ messages. Merging data from multiple platforms gives the best results.
KnowMe does not call any AI API for analysis (that would be slow, expensive, and inconsistent).
Instead, it uses a signal-based scoring system:
Your chat history โ 50+ behavioral signal scan โ 8-dimension weighted scoring โ MBTI + personality profile
| What you said | Detected signal | Dimension affected |
|---|---|---|
| "Let me think about that..." | Needs solo processing time | I +3 |
| "Let's brainstorm together" | Collaborative tendency | E +2 |
| "How exactly do I do this step by step?" | Practical orientation | S +2 |
| "What if we look at it from another angle..." | Abstract thinking | N +2 |
| "The data shows..." | Data-driven reasoning | T +2 |
| "How would this make the team feel?" | People-centered concern | F +2 |
| "What's the next action item?" | Closure-driven | J +2 |
| "It depends, let's stay flexible" | Open adaptability | P +2 |
- โก Fast โ Analyzes 1000+ messages in seconds
- ๐ Private โ Zero API calls, data never leaves your device
- ๐ Explainable โ Every score has traceable behavioral evidence
- ๐ Reproducible โ Same input always gives the same output
# Collect from multiple sources
python3 scripts/collect.py --source chatgpt --file chatgpt.json --output /tmp/gpt.json
python3 scripts/collect.py --source claude --file claude.json --output /tmp/claude.json
# Merge with jq
jq -s '{messages: (.[0].messages + .[1].messages), source: "merged", total_messages: (.[0].total_messages + .[1].total_messages)}' \
/tmp/gpt.json /tmp/claude.json > /tmp/merged.json
# Analyze merged data
python3 scripts/analyze.py --input /tmp/merged.json --output report.mdpython3 scripts/analyze.py --input data.json --output report.md --json raw_scores.jsonknowme/
โโโ README.md # ไธญๆๆๆกฃ
โโโ README_EN.md # English docs
โโโ scripts/
โ โโโ collect.py # Multi-source data collector
โ โโโ analyze.py # Signal-based personality analyzer
โ โโโ advise.py # 16-type personalized advice engine
โโโ references/
โ โโโ mbti_signals.md # Signal taxonomy (50+ signals)
โ โโโ advice_frameworks.md # Advice generation frameworks
โโโ assets/ # Demo images and media
โโโ SKILL.md # OpenClaw Skill definition
โโโ CONTRIBUTING.md # Contributing guide
โโโ LICENSE # Apache 2.0 License
- Core MBTI 4-dimension analysis
- ChatGPT / Claude / OpenClaw data sources
- Personalized advice for all 16 types
- Bilingual signal detection (EN + CN)
- OpenClaw Skill integration
- ๐ Web UI (upload file โ see report in browser)
- ๐ Gemini / Copilot / DeepSeek data sources
- ๐ Big Five personality analysis module
- ๐ Timeline analysis (how has your personality shifted?)
- ๐ Anonymous benchmarking (how do you compare to 100K people?)
- ๐ AI style analysis (how do different AIs change your expression?)
PRs welcome! Especially for:
- ๐ New data source parsers (Gemini, Copilot, DeepSeek, etc.)
- ๐ New behavioral signal patterns
- ๐ Signal detection for more languages
- ๐ Visualization and Web UI
- ๐งช Big Five / Enneagram and other personality frameworks
See CONTRIBUTING.md for details.
KnowMe is a self-discovery tool, not a clinical assessment. MBTI is a preference model โ use these results as conversation starters for self-reflection, not as fixed labels. Analysis quality improves with more conversation data (50+ messages recommended).
Apache 2.0 โ Use it, fork it, improve it, share it.
If KnowMe helped you understand yourself better, give it a โญ Star!
Built with โค๏ธ by AI้ ้ฟAndy
You've had thousands of conversations with AI. It's time those conversations worked for you.