Translation looks like one API call. Until the first InDesign file.
Prompt in, text out works in a demo. In production it breaks on tags, layouts, drifting terminology and a missing audit trail, usually around month three. Own the interface, the models and the prompts. The file filters, the term matching and the audit trail are ours to maintain.
You keep the ownership. We keep the maintenance.
A clean line between what your engineers build and what they never have to touch.
- 01The interfaceYour own frontend, or our UI white-labelled inside your portal.
- 02The models and promptsClaude, ChatGPT, DeepL, Mistral or your own model. Your keys, your routing.
- 03The agentsYour assistants and internal tools call termbase.io like any other tool.
- 04The dataBYOK and EU data residency on Azure. Nothing leaves the region unless you configure it.
- 01File filters and parsersInDesign, Word, Excel, PowerPoint, PDF, HTML, JSON and the proprietary formats nobody else reads, with the layout intact.
- 02Terminology, deterministicApproved terms matched and enforced before and after every model call, not left to the prompt.
- 03Memory, segments, QA gatesTranslation memory, segment scoring and an expert check wherever confidence is low.
- 04The audit trailWhich model, which termbase, which edit, and when. For every document.
Plug in where your stack already is.
Use the engine headless, hand it to your agents, or run our interface under your brand.
- Translation, review, export and audit by API
- Files in, files out, layout preserved
- Dedicated API users per integration
- Works with any MCP client
- Terminology and file pipeline as tools
- Every call in the audit trail
- Your brand, your domain
- Partner and client accounts
- Live in weeks, not quarters
One config. Your agent plays by your rules.
Add termbase.io to any MCP client. Your agent gets your terminology, the file pipeline and the audit trail as tools.
- 01In your IDE
Check UI strings and docs against the approved termbase while you code, in Cursor or VS Code.
- 02In your assistant
Staff drop a document into Claude Desktop or your internal chatbot. The termbase.io pipeline runs underneath.
- 03In your workflows
Automated pipelines that translate, check and deliver without anyone opening a file.
{
"mcpServers": {
"termbase": {
"url": "https://<workspace>.termbase.io/mcp",
"headers": {
"Authorization": "Bearer <TERMBASE_API_KEY>"
}
}
}
}What happens between your call and your model.
- 01ParseFile filters extract the text and keep layout and structure.
- 02MatchTranslation memory and segmentation.
- 03EnforceApproved terminology goes into the request.
- 04TranslateYour model, your keys, your routing rules.
- 05CheckQA gates score every segment; uncertain ones go to an expert.
- 06RebuildBack into the original format, logged in the audit trail.
Bring your models. Keep your systems.
Native connectors where they exist, custom adapters where they don't, scoped as a fixed one-time project.
Adobe PDF
Adobe InDesign
HTML 5
Microsoft Word
Microsoft Excel
Excel Columns
PowerPoint
JSON
Apps & Platforms
Websites & More+ proprietary formats
“Working with termbase.io has been a true partnership from day one. Rather than forcing us into a fixed workflow, it gives us a versatile toolbox to meet the unique needs of each of our clients — and feature requests are often implemented surprisingly quickly.”
ANT KAHRAMANOGLU — QA MANAGER, LANGUAGE TECHNOLOGY, T'WORKS