mcp
Tools
The tools the wxrks MCP server offers, with their arguments. Locales are wxrks locale codes, such as en_us or pt_br.
translate
Translates one text into one or more languages with the same pipeline as the wxrks Portal: translation memory matches, glossary terms, machine translation and your account's AI instructions are applied by wxrks. It calls Translate (Context Sensitive).
| Argument | |
|---|---|
text | Required. The text to translate. |
target_locales | Required. 1 to 10 target locales. |
source_locale | Detected when omitted. |
org_unit_uuid | An org unit, for its context and model routing. |
instructions | Optional tone, formality, audience, domain, creativity (0 to 5), guideline and free-form user_prompt. |
search_translation_memory
Finds existing translations in your translation memories by fuzzy similarity, for one text or up to 20 segments. A score of 1.0 is an exact match, and 1.01 an exact match whose surrounding context matched too.
| Argument | |
|---|---|
target_locale | Required. The locale of the translations to return. |
query or segments | One text, or up to 20. |
source_locale | Picks the translation memories when tm_uuids is omitted. |
tm_uuids | Up to 10 translation memories to search. By default, the ones matching the locales. |
threshold | Minimum similarity, from 0.49 to 1.0. Default 0.7. |
max_hits | Matches per segment, up to 50. Default 5. |
check_terminology
Checks source and target pairs against your glossaries. It reports glossary terms whose expected translation is missing from the target, forbidden terms that were used, and do-not-translate terms that were changed. Without targets, it lists the terms in the source and their expected translations.
| Argument | |
|---|---|
pairs | Required. 1 to 50 pairs of source and target. |
source_locale, target_locale | Required. |
glossary_uuids | The glossaries to check. By default, every glossary the user can access. |
include_ok | Also reports the terms that were translated as expected. |
Recognizing terms, inflected forms included, is done by wxrks. Whether the expected translation is present is checked word by word, so the result can differ from the editor's QA on tagged or heavily inflected text.
review_translation
Reviews one segment or paragraph with AI (wxrks Smells). It reports each issue with a category, a severity (BLOCKER, CRITICAL, MAJOR, MINOR or INFO) and an explanation, and a confidence score where 1.0 means no issue was found.
| Argument | |
|---|---|
source, target | Required. The text and its translation. |
source_locale, target_locale | Required. |
org_unit_uuid | An org unit, for its model routing. |
detect_language
Returns the locale of a text, from its first 200 characters.
| Argument | |
|---|---|
text | Required. |
list_translation_memories
Lists the translation memories the user can access, with their ids for search_translation_memory.
| Argument | |
|---|---|
name | Part of the name. |
source_locale, target_locale | Only memories for these locales. |
limit | Up to 100. Default 25. |
list_glossaries
Lists the glossaries the user can access, with their ids for check_terminology.
| Argument | |
|---|---|
name | Part of the name. |
limit | Up to 200. Default 50. |
whoami
Returns the name, email, id and status of the wxrks user the tools act as. It takes no arguments.