Description

About This Project

SDLXLIFF Refiner does not use the Trados Studio APIs. It reads and writes SDLXLIFF files directly, which gives it the flexibility to support workflows that fall outside what the current Trados toolset makes possible – including tighter integration with AI tools and non-standard approaches to working with bilingual file content.

The application was built to explore different ways of working with AI in a translation context, and to address some practical requirements that existing Trados-native solutions do not cover.

The multifariousMCP server documented here is an extension of that investigation: it demonstrates how the CLI capability built into SDLXLIFF Refiner can be used as a straightforward basis for an MCP server, exposing the same commands to an AI assistant. This is a single-user, local implementation designed for use with Claude Desktop on the same Windows machine as the CLI.

Connect Claude to your Trados Studio projects

multifariousMCP is a Claude Desktop extension that lets you work with Trados Studio bilingual files through natural language. Ask Claude to inspect a project, clean up formatting tags, fix terminology across a whole project, or translate segments, files, or entire projects – all without leaving Claude Desktop. The extension works by driving the SDLXLIFF Refiner CLI under the hood, so you will need SDLXLIFF Refiner installed with sdlxliff-refiner.exe accessible to get the most out of it.

What you can do

Explore projects and files
Ask Claude what projects you have, check how much of a file is translated, or read specific segments before deciding what to do next.

Find and replace
Fix terminology or normalise spelling across a single file or an entire project. Full .NET regex support for pattern-based replacements.

Tag operations
List all formatting tags in a file, remove unwanted tags (such as review highlighting left over from the source document), or swap one tag type for another – all with a dry-run preview before committing.

Translate
Send segments to OpenAI, Anthropic, Google Gemini, or any named provider profile for translation. Segments are sent as numbered chunks with surrounding context, not one at a time. Locked segments are always skipped. Use the project’s stored prompt for consistent results.

Edit segments directly
Set a single segment by row number, bulk-apply translations from a JSON file, or clear draft segments ready for a fresh pass.

Manage prompts
View, update, or reset the per-project LLM prompt used by the translate tool and the SDLXLIFF Refiner clipboard workflow.

Requirements

• Claude Desktop (latest version)
• SDLXLIFF Refiner CLI (sdlxliff-refiner.exe) available on your machine
• Windows (the CLI requires .NET 8 on Windows)
• Trados Studio with projects on local disk

Installation

1. Download the app, which is delivered as a zip file.
2. Unzip the downloaded file and extract multifarious-mcp-1.2.0.mcpb to your desktop.
3. In Claude Desktop, go to Settings → Extensions → Advanced Settings → Install Extension.
4. Select the .mcpb file.
5. When prompted, enter the full path to sdlxliff-refiner.exe.
6. Restart Claude Desktop.

That’s it. Claude will have access to 10 tools covering the full range of SDLXLIFF Refiner operations.

Example conversations

“Remove all yellow highlighting from the Hebrew files in the ‘International Road Education Association’ project”

“Replace ‘profesores’ with ‘docentes’ across the Spanish file – show me what would change first.”

“What Trados Studio projects do I have?”

“What’s the current status of progress in the ‘Operations Report – LLM Context’ project?”

“Conduct a Translation Quality Assessment on the German files in the ‘Operations Report – LLM Context’ project. Before assessing any segment, read the entire document so you have full context. Then assess each segment against the source, considering fluency, accuracy, terminology consistency, and register. For each segment, set the confirmation level to either approved-translation or rejected-translation. For every rejected segment, add a comment explaining the specific error. Work through the file systematically – do not assess segments in isolation.”

Resulting in this in Trados Studio:

“Translate the Hungarian files in the ‘Operations Report – LLM Context’ project. Before translating confirm the project exists and check the current translation status. Use the OpenAI provider with the project prompt and use a dry-run first to tell me the chunk breakdown before proceeding. Only translate after I confirm.”