Best Practices for Bilingual (EN/JA) Descriptions in mcp-go Tools

[rs-tutt]

Title: Best Practices for Bilingual (EN/JA) Descriptions in mcp-go Tools

---

Context

I’m building MCP tools that will be consumed by LLMs (e.g. ChatGPT, Custom GPTs) via deep-search or OpenAI prompts. Each tool field has a description string, and I’m considering whether to include both English and Japanese:

• Option A: English-only descriptions
• Option B: Bilingual descriptions (English + Japanese)

My goal is to help the LLM correctly parse, explain, or render UI labels in either language without confusion.

---

Questions

1. Is it generally more effective to keep descriptions English-only, or to provide bilingual descriptions?
2. If bilingual is recommended, what formatting convention ensures the LLM cleanly separates the two languages?
3. Are there any pitfalls I should watch out for when mixing or prefixing languages?

---

Example of an “optimal” bilingual description format

Below is a generic template (not tied to any existing tool) demonstrating how I imagine separating the two languages clearly:

// Description for a hypothetical “user_id” field
description := `
English:
Unique identifier for the user. Must be a positive integer.
Example: { "user_id": 1234 }

日本語:
ユーザーを一意に識別する番号です。正の整数で指定してください。
例: { "user_id": 1234 }

• The “English:” and “日本語:” prefixes mark each block.
• Blank lines separate the language blocks.
• No mixing of EN/JP within the same sentence.

Would this style be parsed reliably by LLM-based tooling? Or is there a more robust pattern I should adopt?
Looking forward to your insights!