Configure multilingual response handling
Set up automatic language detection and response in the user's preferred language.
AGNT serves users in Bali who speak English, Indonesian, Japanese, Korean, Mandarin, Russian, French, German, and more. This prompt configures the language detection and response pipeline so the agent automatically detects the user's language and responds in kind — without the user needing to set a preference.
The configuration covers three layers: message-level language detection (per-turn), user memory language preference (persistent), and venue-specific language support (which languages the venue's knowledge pack has been translated for). When a venue does not have translated content, the agent responds in the detected language but flags that menu/pricing details are only available in English or Indonesian.
The prompt
<<<
You configure multilingual response handling for an AGNT agent.
LANGUAGE DETECTION:
1. On each user message, detect language using the first 50 tokens.
2. Supported languages: en, id, ja, ko, zh, ru, fr, de, es, it, nl, pt.
3. If detected language differs from user's stored preference, update memory.
4. If detection confidence < 0.7, fall back to stored preference or English.
RESPONSE RULES:
1. Always respond in the detected/preferred language.
2. Venue names, addresses, and proper nouns stay in original form (do not translate).
3. Prices always in IDR with original format (e.g., "Rp 150.000").
4. If venue knowledge pack is not available in the detected language:
- Respond in detected language for conversational parts.
- Include menu items / policies in English with a note: "(menu available in English only)".
5. Never mix languages within a single response unless quoting a venue name.
MEMORY UPDATE:
- Write detected language to user memory key: languages_spoken (array).
- Primary language = most frequently detected over last 10 messages.
>>>When to use
Use this when setting up or auditing the multilingual pipeline. Essential for Bali venues that serve international tourists — incorrect language handling is the #2 cause of user drop-off after slow responses.
Works with
Related guides
Share as social post
Configure multilingual response handling (claude-code) — Set up automatic language detection and response in the user's preferred language. https://agntdot.com/prompts/multilingual-response-config
197 / 280 chars
Related prompts
Codex fleet-swap operator prompt
Walk an operator through swapping an AGNT agent from claude_local to codex_local.
Agent hiring brief — reusable template
The CEO/CTO prompt for spawning a new sub-agent via Paperclip with clear acceptance criteria.
Venue reply drafting prompt
Draft a reply in the venue's voice, in the guest's language, with booking-aware context.