Glossary and translation memory

Two things sit on top of the API, and you manage both in the dashboard:

• The glossary is a list of terms and phrases, each with the translation you want. Brand names, product names, defined terms and the phrases your team always says one way belong here.
• The translations list holds every translation your keys have made. Search it, edit any line and import or export it as CSV.

Both names get used for the same feature in places, so this guide fixes them: glossary is the list of entries you write, and translation memory is the automatic reuse of finished translations described in Concepts. In the dashboard, the Translation Memory screen is where you manage the glossary.

Glossary or translation memory

The memory stays exact, never fuzzy. The glossary is how you bend a translation toward your own words.

Add a term

1. Sign in to the dashboard, open Translations, then choose Translation Memory and Add Entry.
2. Pick the source language and the target language. An entry belongs to one language pair, so add one per language you translate into.
3. Enter the source text and the translation you want.
4. Add a note if the model needs context, such as "product name, keep in English". Notes are passed to the model with the entry.
5. Save. The entry is Active.

To add many entries at once, switch on Bulk Entry Mode and paste the source texts and the translations as two matching lists, one per line.

You can edit an entry, Deactivate it to stop using it without deleting it, or delete it. Filter by language pair and status, or search the source and target text.

Import and export entries

Export current filters to CSV downloads the entries you are looking at, with the columns id, source_language, target_language, source_text, target_text, active and notes. Import entries from CSV reads the same columns. A row with an id updates that entry, and a row without one creates a new entry.

How entries are used

• Account-wide. Entries belong to your account, not to one key, so a term you add is used for every key on the account, through the REST API, the MCP server, LangChain tools, the n8n node, Make, Zapier and function calling.
• Active entries only. FluentC uses an active entry when the source text appears in what you send and the language pair matches. Inactive entries are ignored.
• Real-time and batch. Entries apply in both modes.
• New translations. The glossary shapes the translation FluentC makes. Text that is already stored is returned as it is, so to change a line that exists, edit it in the translations list.
• Keep entries short. A name, a term or a phrase works better than a paragraph.

Edit every translation

1. In the dashboard, open Translations.
2. Choose a site and a language. Search the original or translated text, and sort by last updated, original text or translated text.
3. Open a line, change the translation and save.

For many lines at once, choose Export current view to CSV, edit the translated_text column and leave id and type as they are, then choose Import translations from CSV. FluentC reports how many lines it saved and which rows it skipped.

Review once, send everywhere

The loop for a team with many agents is short. Send the sentences that matter, read the translations once, fix any line in the dashboard, and add recurring terms to the glossary. Every agent on the account then sends the reviewed words, through any integration. Concepts explains why identical input returns identical output, and Ten agents, one translation shows the loop in an example.

What it doesn't do

• It doesn't make the memory fuzzy. The memory matches identical text only.
• It doesn't fix your source text. If two agents word the English differently, they send different strings. Standardise the sentences you care about.
• It doesn't replace review. For regulated or high-stakes content, keep your own review process. The glossary and the dashboard make that review a one-time job per phrase.