Skip to main content
Lab Grimoire
TW EN
☕ Coffee
Spellbook / Kami ↗ Visit project
ai-tools

Kami

A document design system for the AI era. Uses constrained HTML/CSS/SVG rulesets to unify the visual language of eight professional document types. Natively integrates as a Claude Code Skill — AI auto-detects intent and generates print-ready PDFs with 14 built-in SVG chart types. 4,400+ Stars in two weeks, MIT licensed.

Best For

Claude Code users who want to transform AI-generated content into beautifully designed PDF documents without manual formatting. Ideal for popular science articles, research reports, personal resumes, and presentations. If you're tired of generating great content with AI but still spending time on layout, this is the tightest integration currently available.

How I Actually Use It

Installed as a global Claude Code Skill. When writing articles or reports, AI automatically detects the document intent and applies the appropriate template (One-Pager / Long Doc / Resume / Slides) to produce PDF output. SVG charts — architecture diagrams, timelines, quadrant charts — are generated inline by AI with zero external tool switching.

Where It Is Strong

  • npx skills add tw93/kami one-line install. AI auto-detects document intent with zero slash-command configuration
  • Pure HTML/CSS/SVG, so AI can directly read/write without build environments
  • Constrained design system: Warm parchment aesthetic (background #f5f4ed, accent Ink Blue #1B365D), 4-point spacing system, high visual consistency
  • 14 built-in SVG chart types (architecture diagrams, quadrant charts, candlestick charts, timelines, flowcharts), all AI-generated directly
  • Chinese, Japanese, and English typography with visual QA
  • Not just layout: resumes and presentations have quantified writing standards baked in
  • V1.0 to V1.4.0 in two weeks. Author tw93 also maintains Pake (34k+ Stars)

Where It Fails

  • Very new project (V1.0 released 2026-04-20), so breaking change risk is high and the API is unstable
  • Bus Factor = 1. Single maintainer, long-term sustainability unknown
  • No academic citation support (no APA/Vancouver/DOI), so SCI paper submissions are out
  • PDF/PPTX output only: No DOCX or HTML web output
  • Performance in GPT-4o / Gemini environments is undocumented; tightly coupled to the Claude ecosystem
  • Offline environments require manual local font setup (CDN dependency)
  • Default is SC (Simplified Chinese) fonts; Traditional Chinese coverage needs verification and replacement

Pricing, Difficulty, and Risk

  • Completely free, MIT license
  • npx skills add tw93/kami -a claude-code -g -y completes installation. Requires a Python environment (WeasyPrint)
  • Risk: New project breaking changes, single maintainer, WeasyPrint CSS support limitations

Verdict

Kami's core value is that it doesn't give you templates to choose from. Instead, AI understands design constraints and automatically produces visually consistent documents. The warm parchment aesthetic combined with zero-framework purity has carved out a distinct niche in the Claude Skill ecosystem. Version stability is the biggest open question right now. Start with non-critical documents to learn the workflow, then go all-in after V2.x stabilizes.

Source