Customer Agent in your customer's language

Customer Agent responds in the language your customer writes in — on every channel. Whether a shopper messages you in French, German, Japanese, or Arabic, Customer Agent detects the language and replies in it. No setup required.

What this means for your brand

• Every channel: Customer Agent automatically responds in your customer's language on web chat, email, text messaging (SMS), and WhatsApp.
• No configuration needed. Detection happens on every incoming message, and the agent follows the customer's lead if they switch languages mid-conversation.
• No per-language licensing. Multilingual is included in Customer Agent. One thing to be aware of: on SMS, messages in some languages cost more to send because of how carriers encode them — see “Why some languages cost more on SMS” below.

Which languages are supported?

Customer Agent supports 113 languages across all major European, Asian, Middle Eastern, and African writing systems. If a customer writes in one of these languages, Customer Agent will reply in it automatically.

Full list of supported languages

Languages marked with an asterisk (*) use a writing system that requires the larger UCS-2 encoding on SMS, which means replies in those languages use more message segments and cost more to send — see “Why some languages cost more on SMS” below. The asterisks only matter for text messaging; on web chat, email, and WhatsApp every language costs the same.

Afrikaans · Albanian · Amharic* · Arabic* · Armenian* · Assamese* · Azerbaijani · Bashkir* · Basque · Belarusian* · Bengali* · Breton · Bulgarian* · Burmese* · Catalan · Cebuano · Central Kurdish* · Chechen* · Chinese* · Chuvash* · Cornish · Czech · Danish · Divehi* · Dutch · Eastern Mari* · English · Esperanto · Estonian · Finnish · French · Galician · Georgian* · German · Goan Konkani* · Greek* · Guarani · Gujarati* · Hebrew* · Hindi* · Hungarian · Icelandic · Ido · Iloko · Indonesian · Irish · Italian · Japanese* · Javanese · Kannada* · Kazakh* · Khmer* · Kirghiz* · Komi* · Korean* · Lao* · Latin · Latvian · Lithuanian · Lojban · Low German · Lower Sorbian · Luxembourgish · Macedonian* · Malayalam* · Maltese · Marathi* · Mongolian* · Nepali* · Norwegian · Oriya* · Panjabi* · Persian* · Piemontese · Polish · Portuguese · Pushto* · Quechua · Romanian · Russian* · Sanskrit* · Scottish Gaelic · Serbian* · Sindhi* · Sinhala* · Slovak · Slovenian · Somali · South Azerbaijani* · Spanish · Sundanese · Swahili · Swedish · Tagalog · Tajik* · Tamil* · Tatar* · Telugu* · Thai* · Tibetan* · Turkish · Turkmen · Uighur* · Ukrainian* · Upper Sorbian · Urdu* · Vietnamese · Waray · Welsh · Western Frisian · Western Panjabi* · Yakut* · Yiddish*

Note: Some Latin-alphabet languages with extended accented characters — for example Vietnamese, Polish, Czech, or Turkish — may also use UCS-2 encoding on SMS depending on the specific characters in a message.

Where multilingual works

• Web chat: Automatic language matching on every message.
• Email: Automatic language matching on every message.
• Text messaging (SMS): Automatic language matching, with brand controls for which languages the agent uses (see below).
• WhatsApp: Automatic language matching on every message.

Managing languages on text messaging

Text messaging is the one channel where replying in another language can change what you pay (see the next section), so it comes with controls:

• All supported languages are on by default. If you do nothing, the agent simply replies in the customer's language.
• Opt out of additional-cost languages. The languages that cost more to send on SMS — the ones marked with an asterisk in the list above — can be turned off individually. You keep the ones that matter to your customers and skip the rest.
• Set a default language. When a customer writes in a language you've opted out of, the agent responds in your default language instead of going silent or handing off.

Why some languages cost more on SMS

This is a property of how text messaging works, not a Klaviyo fee.

SMS messages are billed per segment, and how many characters fit in a segment depends on the message's encoding. Messages using the basic Latin character set (GSM-7) fit about 160 characters per segment. Languages that need other scripts — the ones marked with an asterisk in the list above — require a different encoding (UCS-2) that fits about 70 characters per segment.

The practical effect: the same reply written in one of these languages uses roughly twice as many segments, so it costs more to send. That's why the language opt-out controls above exist — you can support the languages your customers actually use and opt out of the ones that would only add cost.

Email, web chat, and WhatsApp don't bill by segment, so this doesn't apply to those channels.

What's not multilingual yet

A few things to be aware of as you plan campaigns and content:

• Your uploaded knowledge content (documents, webpages, FAQs) should still be in English. The agent can answer in your customer's language from English source material, but the better your English content, the better the final answer in any language. Native multilingual content support is on our roadmap.
• Custom agents and skills you've configured still use English instructions. The agent will translate its customer-facing reply, but internal reasoning, tool calls, and brand instructions are in English.
• Brand tone of voice instructions are written in English. The agent does its best to carry your tone through to the customer's language, but very brand-specific phrasings (idioms, slogans, taglines) may not always translate naturally.

Tips for getting the best results

• Keep your English knowledge content clear and complete. This is the source of truth Customer Agent uses to answer in any language.
• Use brand voice instructions that focus on principles, not specific phrases. “Warm and friendly, never overly formal” carries across languages better than a specific English catchphrase.
• Spot-check responses in the languages your customers use most. You can do this from the Customer Agent preview panel by chatting in a target language — use the channel selector to test each channel.
• On SMS, review your language list once. Five minutes deciding which languages to keep on (and what your default should be) prevents surprise segment costs later.

FAQ

Do I need to enable this?
No. Multilingual is on by default across all channels.

Can I turn it off?
You can turn Customer Agent itself on or off for any channel in your Customer Agent settings. For languages specifically: on text messaging, you can opt out of additional-cost languages and set a default language. On web chat, email, and WhatsApp, the agent always replies in the customer's language — there's no per-language setting to manage.

Will this affect my existing English customers?
No. The agent still detects English and responds in English — exactly as it does today.

Will replying in another language cost me more?
On email, web chat, and WhatsApp — no. On SMS, languages that require UCS-2 encoding fit fewer characters per message segment, so the same reply uses more segments and costs more. Use the language controls to opt out of languages you don't want to support.

Will the agent translate my brand name or product names?
The agent tries to preserve proper nouns (brand names, product names, trademarks) without translating them. If you see a specific product or trademark being translated, let us know.

What if the language detection is wrong?
Detection is highly accurate for clear, single-language messages but can occasionally misclassify very short or code-switched messages (e.g., a one-word reply or a message that mixes two languages). The agent will follow the customer's lead if they switch languages mid-conversation.

Where is this available?
For all Customer Agent customers, on web chat, email, text messaging, and WhatsApp.

Can the agent respond to customers based on their locale — including referencing the right products, promotions, and policies?
This is not yet available and is on the roadmap. In the meantime, the agent will respond in the right language but will draw from a single global set of context and instructions.

Additional resources