It applies to the next job
An approved term reaches the next file, the next language and the next colleague without anyone forwarding it.
01 · For the people who own the output
Translate a file, or run the whole project.
A translation project is one file or set of files with its languages, its people and its review steps attached. You choose who works on each language and how much review each job gets, without a ticket and without a developer.
Bring your own freelancers and agencies at the rates you already agreed. Files come back in the format they arrived in: PDF, DOCX, PPTX, XLSX, XML, InDesign IDML, JSON and SCORM.
See what a project looks like
Not everything your company publishes starts as a file somebody sends for translation.
02 · For the people who draft and write
Write on your own terminology.
Support drafts an article with an assistant. A colleague in Milan writes the Italian page from scratch, and it was never a translation of anything. Neither of them is moving to another editor, and neither should be inventing your vocabulary.
Two assistants work inside OneLanguage, one in the segment and one on your terminology. Outside it, an MCP server lets the editor your team already uses call your account as a tool.
A token carries a scope. Read lets the assistant search. Write lets it change entries, so a term settled in a conversation is in the account the same day. Without write, the assistant is never shown the tools that would modify anything. Tokens are stored as a hash, so nobody reads one back afterwards, including us. As of September 2026 the server exposes terminology, and style guides follow.
See the MCP landing
One person asking one question is fine. Four thousand descriptions overnight is not.
03 · At volumes nobody can read
Hand us the volume nobody can read.
Supervised AI translation is a managed service. TextUnited translates on your terminology and memory, scores every segment against its source, and puts a translator on the ones that score too low. What those translators change goes back into your language data.
You keep paying a human rate on the segments that needed one. Nothing on your side routes, queues or chases anything.
Ask what this looks like on your content
All three of those need the account reachable from wherever the work already runs.
04 · For the people who run the systems
Connect the systems you already run.
Two ways to make the same connection. A team with developers calls the API. A team without them builds the same sequence in the automation tool it already runs. Either way, the systems your content already lives in reach the account.
The API is open and what your team builds on it stays yours. The MCP server is a wrapper over the same API, so anything your systems can reach, your assistants can reach on the same terms.
05 · What accumulates
Keep what all four of them produce.
Language data is the terminology, translation memory, style guides, review decisions and quality signals your organisation accumulates as it publishes in more than one language. A project, an assistant, a managed batch and a connected system all write to the same one.
It answers later questions
Segment states and the comments around them explain why a phrase reads the way it does, a year after the person who chose it moved teams.
It leaves with you
Terminology and translation memory export as TBX and TMX, which a competing system can open. Your content sits in your own tenant.
Questions
Before you talk to us
Is OneLanguage a translation management system?
Yes, and it covers more than translation. Translation projects, translation memory and terminology are all inside it. The same language data also serves content written natively in a market and content drafted by your own AI tools.
What is language data?
Language data is the terminology, translation memory, style guides, review decisions and quality signals your organisation accumulates as it publishes in more than one language. In OneLanguage it is stored so that both people and systems can query it.
Do we write our content in TextUnited?
Your teams keep writing in the tools they already use. OneLanguage holds the terminology, style guides and approved wording those tools ask for, and keeps what comes back. Translation, review and approval happen in OneLanguage.
We write a lot of content natively, not as translations. Does that fit?
Terminology and style are properties of how your company writes, not of a translation. They apply to Italian written in Milan the same way they apply to Italian translated from English.
Which systems can we connect?
Anything that can call a REST API, which in practice means your CMS, PIM, product documentation tool, help desk or code repository. Teams that prefer not to write an integration build the sequence in the automation tool they already run.
What can our AI assistant actually do with this?
As of September 2026 the TextUnited MCP server exposes terminology, and style guides follow. An assistant reads your approved terms, and with a write token it adds and changes entries. It cannot exceed the scope on the token you gave it.
Can I bring my own translators and agencies?
Yes. You add your existing freelancers and vendors at the rates you already agreed, and order from TextUnited linguists inside the same project when a language pair has nobody on it.
What happens to our language data if we leave?
It belongs to your organisation. Terminology and translation memory export as TBX and TMX, which a competing system can open.
How do we start?
Most teams begin with a free trial on their own content, then a call to set up terminology and the workflows that match how the work arrives.
Start with real work
Bring one page and the terms two departments disagree on.
Not a demo dataset. Something a colleague published last week in a language you cannot check, and the vocabulary nobody has settled.