Documentation Generation Workflows

Automated technical documentation generation combining multiple MCPs. End-to-end processing from specification retrieval through multilingual output.

Pattern 7: Documentation Generation Workflow

Overview

A technical documentation generation flow combining multiple MCPs. Integrates four stages — specification retrieval, diagram generation, translation, and quality verification — to efficiently produce multilingual documentation.

MCPs Used

• rfcxml-mcp - Specification information
• mermaid-mcp - Diagram generation
• deepl-mcp - Multilingual support
• xcomet-mcp - Translation quality verification

Flow Diagram

This workflow integrates specification content, visualization, and translation to create multilingual documentation:

Stage Responsibilities

Stage	MCP	Input	Output
Spec Retrieval	rfcxml-mcp	RFC number	Section structure, requirements list
Diagram Generation	mermaid-mcp	Text description	SVG/PNG diagrams
Multilingual	deepl-mcp	Japanese draft	English documentation
Quality Check	xcomet-mcp	Source + translation	Quality score, error report

Design Decisions and Failure Cases

• Japanese → English order: This project adopts the flow of thinking and writing in Japanese, then translating to English. Creating high-quality content in one's native language before translation tends to produce better overall quality than writing in English first.
• Failure case: Mermaid diagram syntax errors can block the entire document generation. This can be avoided by pre-validating with mermaid:validate_and_render.
• This repository as a living example: The documentation in the ai-agent-architecture repository is created and translated based on this workflow.