The statement "AI reads docs" is more significant than "AI writes docs." Tools like ChatGPT and Claude consume documentation as context—and they need different formatting than humans.
The Paradigm Shift
For eight years, I've been advocating docs-as-code. Now the field faces a fundamental transformation: documentation must serve two distinctly different audiences.
Emerging Trends
- llms.txt adoption: 600+ organizations including Anthropic and Stripe now provide AI-readable documentation summaries
- Model Context Protocol (MCP): Standardization for AI tool integration
- RAG technology: Converting documentation into searchable knowledge bases
The Market Gap
All current AI-native documentation tools operate as SaaS. This leaves enterprises with data privacy concerns without suitable options for self-hosted solutions.
docToolchain rb Features
- 🔒 Self-hosted AI documentation with user-controlled LLMs
- 📝 AsciiDoc-native formatting
- 🏛️ Architecture documentation assistance (arc42 + ADR)
- 🤖 llms.txt and MCP support
- 🐳 Docker-first infrastructure
"The question is no longer 'How do we write docs?' but 'How will AI read our docs?'"
What's Next?
docToolchain rb represents the evolution of documentation tooling for a world where AI systems are first-class consumers of our technical content. Stay tuned for the release.