Automated Multilingual Technical Documentation Pipeline

Project: Automated Multi-language Technical Documentation Pipeline

(Made with DeepL MCP Server + DeepL API)

What I built

I created an automated translation pipeline that converts a large-scale Japanese technical documentation site into a fully usable English version using the DeepL API through the new DeepL MCP Server, integrated with a coding-agent workflow.

The Problem

I maintain a technical site with:

• 180 pages

• 1.5 million Japanese characters

Translating this manually would normally require:

• A technical writer + translator

• Several weeks (2–4 weeks)

• A high cost (100× more expensive than DeepL API)

How it uses DeepL

Using the DeepL MCP Server, the agent:

1. Sends each section/page through the DeepL API

2. Collects and stores translations

3. Rebuilds the site automatically (VitePress)

4. Performs QA via round-trip checks

5. Outputs a full English version

Results

• Total translated characters: 27,349

• Translation time: ~1 day

• Cost: 1,318 + 630 yen (~12 USD)

• Translation quality: Extremely high, good enough for technical content

• Final QA performed with DeepL + GPT/Claude

Why it matters

This workflow:

• Makes multi-language documentation accessible for individual developers

• Reduces translation costs dramatically

• Eliminates the biggest barrier for solo creators who want to reach a global audience

• Demonstrates the power of integrating DeepL MCP Server with coding agents

Links

• Japanese site: https://shuji-bonji.github.io/RxJS-with-TypeScript/

• English site (DeepL-translated): https://shuji-bonji.github.io/RxJS-with-TypeScript/en/

• GitHub repo: https://github.com/shuji-bonji/RxJS-with-TypeScript

Screenshots