Skip to main content
PickFu runs surveys across 15 countries, each answered by native-speaking respondents in their own language. Localization is automatic and driven by the audience country — you don’t translate anything yourself.
If you’re an AI agent or app calling PickFu, don’t pre-translate the survey. Write the question, context, and options in your working language, set the survey’s country, and PickFu translates them into the respondents’ language for you. PickFu detects the source language, so text that’s already in the target language is left untouched — there’s no double-translation. Translating it yourself is unnecessary and makes results harder to compare across languages.

​
Targeting a country localizes the survey

Set the survey’s two-letter country code (and, optionally, targeting traits). PickFu then:
  1. Translates the question, context, and options into that country’s native language.
  2. Shows the survey to native-speaking respondents from that market.
  3. Collects answers — and their required written explanations — in the respondent’s language.
You set country when you create or update a survey:
  • REST API — the country field on POST /surveys and PATCH /surveys/{id}
  • MCP — the country argument on save_survey
  • CLI — the --country flag on pickfu survey create and survey update
Translation runs when you create a survey, and when you update it with its questions included (the normal save path). Changing only country on an existing draft — without resending the questions — won’t re-translate the existing text; include the questions in the same update to localize them into the new market’s language.
To discover which targeting traits are available in a given market, use GET /traits/targeting?country=DE, the MCP list_available_targeting tool, or pickfu targeting list --country DE. See the full list of supported countries.

​
Reading results in each language

A translated survey keeps the original text and adds translations for the languages PickFu generated for that survey — in practice English plus the audience’s language. You read back whichever you need:
  • Responses — each answer’s explanation is the respondent’s original text. Pass a language code (most often EN) to GET /surveys/{id}/responses, and each answer also gets a translation field holding that explanation in the requested language.
  • Survey content — GET /surveys/{id} returns the question, context, and option translations alongside the originals; pass language to select one.
  • MCP — the get_survey and get_survey_responses tools accept the same language argument.
  • Exports — GET /surveys/{id}/exports renders PDFs and PNGs per language; each entry carries its own language code.
In practice this means an agent can collect feedback from, say, a German audience and read every comment back in English without translating anything itself.

​
Requesting a language that isn’t available

PickFu serves only translations it has already stored. It doesn’t translate on demand when you read results. If you pass a language that has no stored translations for the survey, GET /surveys/{id} and GET /surveys/{id}/responses return HTTP 422 with code: TRANSLATION_UNAVAILABLE instead of returning untranslated or empty fields:
availableLanguages lists the survey’s source language plus every language with stored translations. Retry the request with one of those codes, or omit language to get the original text. Language codes are case-insensitive. The MCP get_survey and get_survey_responses tools surface the same error. The tool result includes the TRANSLATION_UNAVAILABLE code, the message, and the available languages, so an agent can retry with a supported code.

​
Controlling what gets translated

By default, all three elements — question, context, and options — are translated for the target audience. Sometimes you want finer control: translate the question into German, for instance, but keep the option labels verbatim because they’re proper names, brand names, SKUs, or model numbers that shouldn’t be localized.
Surveys created through the API, MCP, or CLI translate every element for the target audience. To keep a specific element untranslated — for example, option labels that are proper or brand names — use the per-element translation toggles (question / context / options) in the PickFu survey builder in the web app.

Surveys overview

Question types, audience targeting, and how results are returned.

MCP server

Create and read surveys — with country and language — from an AI assistant.

CLI

Use --country and read localized results from the command line.

API reference

The country and language parameters on the survey endpoints.