Zenchat now speaks 40+ languages

Starting today, every Zenchat agent can hold a conversation in over 40 languages — without you training separate agents or translating your source content by hand.

How it works

The agent detects the visitor’s language from their first message and responds in that language, while still retrieving answers from your existing English (or any single-language) knowledge base. Ask a question in Spanish, get an answer in Spanish, sourced from the same docs your English-speaking customers use.

This matters most for teams who sell globally but write docs in one language. You don’t need a translation pipeline to support a global support surface — the agent bridges that gap at conversation time.

What changed under the hood

• Automatic language detection on the first inbound message, with a manual override in the widget for visitors who prefer to type in a different language than they’re browsing in
• Retrieval stays language-agnostic — the embedding model finds the right source chunks regardless of the query language
• Responses are generated natively in the visitor’s language rather than translated after the fact, which keeps tone and terminology consistent

No configuration is required to turn this on — it’s live for every workspace today. If you want to restrict an agent to specific languages (for compliance or brand reasons), that’s now available as a setting under Agent → Guardrails.