Most teams hear "MCP server" and picture a developer tool. The early adopters are showing it's broader than that. Localization managers, program owners, and content teams are connecting their AI assistant to Smartling and getting routine work done in a sentence or two.
Here's what that looks like in practice, based on how customers are actually using it.
What is Smartling's MCP server?
The Model Context Protocol (MCP) is an open standard that lets AI assistants connect to business tools. Smartling's MCP server lets an assistant like Claude, ChatGPT, or Gemini read from and act in your Smartling account, using plain language. Everything runs against your own jobs, glossaries, style guides, and translation memory.
The idea behind it is simple: anything you can do in Smartling, you can do from where you already work.
1. Turn a ticket into a translation job
The scenario: A request comes in through Jira as a spreadsheet of strings to translate.
What happens: Someone tags Claude in a comment. It reformats the spreadsheet, creates the Smartling job, and later returns the finished translations in the original layout.
Why it matters: The customer who built this routine estimates it replaces about 10 minutes of manual work per request. Multiply that by every request in a quarter.
2. Localize a document with your own glossary and style guide
The scenario: A product spec sheet needs to go out in French and German, and it has to keep the company's design system.
What happens: One prompt pulls the customer's glossary and style guide from Smartling and applies them to the PDF. Language managers reviewed the output and changed no words.
Worth knowing: In this case the AI assistant did the translating, guided by Smartling's glossary and style guide. No Smartling translation job ran. It's a good example of how portable your language assets become once they're available outside the platform.
3. Handle urgent small edits without the wait
The scenario: A press release gets new paragraphs the day before it needs to ship, and the language manager is out.
What happens: The new text is localized the same way as above, using approved terminology, to hit the deadline.
Worth knowing: This skips the normal review step, so it's best for urgent small edits, not as a replacement for review on content that needs it.
4. Fill a content library from past work
The scenario: A team has a spreadsheet library across 3 tabs and 13 languages, and the translations already exist somewhere in past Smartling jobs and projects.
What happens: The assistant matches strings to previous work and fills in the library. The customer estimates this would have taken hours by hand.
5. Keep project management current without building an integration
The scenario: A localization manager lives in a project management tool and needs to know where every translation job stands.
What happens: A routine runs every few hours, uses the MCP server to pull new jobs and their statuses from Smartling, and updates the project board automatically. No custom integration required.
Why it matters: This is the pattern we expect to see more of. Teams write the workflow they need once, in plain language, and it runs.
Make approved wording one search away
Beyond individual workflows, teams are using the MCP server to make their language assets easier to reach. One customer built an internal lookup app that searches their glossary and translation memory so colleagues reuse approved translations instead of guessing. They're now using the MCP server to make it faster.
That points to a bigger shift. Your glossaries, style guides, and translation memory aren't only used by Smartling during translation. You can query them and use them in your AI tool for all kinds of other work.
Where this is heading
Customers are asking for more than lookups. The most common requests are to work with connected tools (code, CMS, design), sync edits made outside Smartling back in, manage projects, and report on translation data by simply asking. Our MCP roadmap is built around those requests, and we'll share more as they ship.
Try it with your team
The best fit is a team that already works in Claude, ChatGPT, or Cursor every day and has lots of small, repeatable translation requests. Start with one routine task and see how it feels.
Programar una reunión and we'll walk you through how it would work in your environment.