From the builder to a configured AI — step by step.

Head to the Profile Builder. Pick a preset in the top-right (Power User, Researcher, Founder, Learner, Default) — sensible defaults across all 17 dimensions. You'll tweak from there.

Left sidebar: 17 dimensions grouped into Communication, Decisions, Interpersonal, Action & Knowledge, Free text. Each dimension has 1–5 questions, multi-select everywhere. Every option has a ⓘ with a concrete example if the wording is unclear.

The right panel shows your CLAUDE.md live — updates with every click. Tabs switch between CLAUDE.md (for Claude Code), ChatGPT Custom Instructions (2 blocks), Obsidian Markdown (with wiki frontmatter), and plain system prompt.

Top of the preview panel: Copy or Download. Auto-save means you can reopen it later and tweak — no starting over.

Download the CLAUDE.md. Drop it in your project root. Claude Code reads it automatically on every start — no extra configuration.

your-project/
├── CLAUDE.md       ← goes here
├── package.json
├── src/
└── ...

ChatGPT → Profile → Custom Instructions. Paste the two blocks from the export into "What would you like ChatGPT to know about you?" and "How would you like ChatGPT to respond?". Each block caps at 1500 chars — our exporter respects that.

Obsidian: the ai-instructions.md comes with wiki frontmatter — drop it into your vault under 10-Wiki/ or wherever.
Gemini, Mistral, local models: paste the system prompt variant into your agent setup, Cursor rules, or the system field of your API call.

If Claude still feels too polite, jump back to the builder, adjust 2–3 options, download the new file, overwrite the old. Your settings persist locally — you never start at zero.