Best MCP servers for translation and localization
Translation from an AI agent is more than passing text through a model. For real localization you want dedicated machine translation that preserves tone and formatting, plus the ability to translate whole documents and rephrase for a target audience, and you often want a strong general LLM in the loop for context-aware, glossary-aware rewrites. The servers below combine a purpose-built translation engine with capable multilingual models, so an agent can localize copy, translate documents, and adapt voice across languages without you wiring translation APIs by hand. Each pick is a real MCP server with a verified, current install config.
DeepL
DeepL
DeepL's official MCP server: high-quality machine translation, document translation, and AI rephrasing across 30+ languages from your agent.
DeepL's official server delivers high-quality machine translation, document translation, and AI rephrasing across 30+ languages, the dedicated engine for accurate, tone-preserving localization.
Google Gemini
Ali Argun
Maintained community MCP server for Google's Gemini API: generate text, analyze images, count tokens, and create embeddings from your agent.
Gemini's server adds a strong multilingual LLM for context-aware translation and glossary-driven rewrites, handy when localization needs judgment beyond literal translation.
OpenRouter
heltonteixeira (community)
Community OpenRouter MCP server: chat with 300+ language models through one unified API, search the model catalog, and validate model IDs from your agent.
OpenRouter's server routes to many language models behind one tool, so an agent can pick the best model per language pair or compare translations without separate integrations.
Together AI
Manas Bharadwaj
Community MCP server for Together AI image generation: create high-quality images with the FLUX.1 Schnell model straight from your agent.
Together's server gives access to a broad set of open multilingual models, a flexible option for high-volume or self-hostable translation workloads from the agent.