Translation and localization MCP servers are turning AI agents into multilingual content managers. Instead of copying text between translation dashboards, managing locale files manually, and hunting for missing translations, these servers let agents translate content, manage translation projects, sync localization files, and maintain terminology glossaries — all through the Model Context Protocol.
The landscape spans five areas: translation API wrappers (DeepL leads with the most starred server), enterprise TMS platforms (Crowdin, Phrase, Lokalise, and Smartling with massive tool counts), developer i18n tools (framework-specific translation file management), platform-specific localization (Xcode, Android, Figma), and local translation engines (offline models for privacy-sensitive workflows).
The headline findings: DeepL’s official server remains the most popular translation MCP server at 102 stars with text, document, and glossary support. Crowdin launched Crowdin Copilot — an AI orchestration platform with Ask Agent and Task Agent that goes beyond translation into full localization management. Phrase expanded from 65 to 96 tools in v0.5.1 with broader Strings and TMS coverage. Three new TMS entrants — SimpleLocalize (official, 14 tools), IntlPull (commercial, 8 tools), and Smartling’s official CLI MCP — joined the ecosystem. Intlayer leads developer i18n tools at 699 stars (+11%), while the new dalisys/i18n-mcp (10 stars, 16 tools) brings hardcoded string detection and multi-framework support. Google Cloud Translation still has no dedicated MCP server despite Google offering 30+ managed MCP servers for other services. Most TMS servers are official vendor implementations — a refreshing contrast with categories where fragmented community servers dominate.
Category: Education & Learning
Translation API Servers
DeepLcom/deepl-mcp-server (Most Popular)
| Server | Stars | Language | License | Tools |
|---|---|---|---|---|
| DeepLcom/deepl-mcp-server | 102 | JavaScript | MIT | 8+ |
The official DeepL MCP server and the most starred translation server in the category. Provides text translation between 30+ languages, document translation (PDF, DOCX, PPTX, XLSX, HTML, TXT, and more), and text rephrasing with tone/style customization (business, academic, casual).
Key features beyond basic translation:
- Glossary support — list, retrieve, and use glossaries for consistent terminology across projects
- Formality control — adjust output formality level for languages that distinguish formal/informal registers
- Language detection — automatic source language identification
- Document translation — preserves formatting across file types
DeepL’s free API tier allows 500,000 characters per month at no cost, making this accessible for experimentation. Integrates with Claude Desktop, Claude Code, and other MCP clients. 83 commits show active maintenance. Stars grew from 95 to 102 since March 2026.
translated/lara-mcp (Translation Memory Pioneer)
| Server | Stars | Language | License | Tools |
|---|---|---|---|---|
| translated/lara-mcp | 80 | TypeScript | MIT | 11 |
Lara Translate’s MCP server stands out with full translation memory management — a capability no other translation MCP server offers. While most servers treat translation as a stateless API call, Lara Translate lets agents build and maintain translation memories over time:
- translate — core translation with language detection, context awareness, glossary, and TM integration
- list_memories / create_memory / update_memory / delete_memory — full CRUD for translation memories
- add_translation / delete_translation — manage individual translation units within memories
- import_tmx / check_import_status — bulk import from TMX files (industry-standard format)
- list_glossaries / get_glossary — terminology management
Lara Translate’s models are trained on professional translation data, emphasizing performance on non-English language pairs where general-purpose MT often struggles. Supports both HTTP and STDIO protocols with Docker deployment. April 2026 update: Released v1.0.1 with OAuth 2.0 authentication support alongside access key auth, expanding deployment flexibility. 42 commits show continued development.
neosun100/translategemma (Local/Private Translation)
| Server | Stars | Language | License | Tools |
|---|---|---|---|---|
| neosun100/translategemma | 2 | Python | — | 1+ |
The only fully local translation MCP server, powered by Google’s TranslateGemma models. For organizations that can’t send content to external APIs — healthcare, legal, government — this is the only MCP option:
- 55 language support from TranslateGemma’s multilingual models
- Multiple model sizes — 4B, 12B, and 27B parameters with Q4 and Q8 quantization
- All-in-one Docker image (82GB) with all 6 models embedded, or lightweight image with on-demand download
- Smart text chunking for handling documents longer than the model’s context window
- Web UI + REST API + MCP integration
The tradeoff is size and speed — an 82GB Docker image is impractical for casual use, and local inference on consumer hardware will be slower than API calls. But for air-gapped or privacy-sensitive environments, this is currently the only game in town.
intento/mcp-intento-translate (Multi-Engine Aggregation)
| Server | Stars | Language | License | Tools |
|---|---|---|---|---|
| intento/mcp-intento-translate | 2 | Python | Apache-2.0 | 1 |
Intento’s MCP server provides multi-engine translation aggregation — instead of committing to one translation provider, Intento routes requests to the best-performing engine for each language pair. A single translate tool accepts text, target language, and optional source language (auto-detected if omitted). Language support is always up-to-date with the Intento API, with a language-codes resource providing ISO code mappings.
Enterprise Translation Management Systems
Crowdin MCP Server (Largest Tool Count)
Crowdin’s official hosted MCP server is one of the most comprehensive MCP servers in any category, exposing 200+ tools that cover the full localization lifecycle:
- Project management — create, configure, and monitor translation projects
- File operations — upload source files, download translations, manage file structures
- Team coordination — assign translators, track progress, manage permissions
- Terminology management — maintain glossaries across projects
- Translation memory — leverage past translations for consistency and efficiency
- Reporting — generate analytics, progress reports, and workflow summaries
- Pre-translation — automated translation using TM, MT, or AI engines
The server connects via Crowdin’s hosted infrastructure — no local installation required. Works with Claude Desktop, Cursor, and other MCP-compatible clients. Crowdin is also building an MCP marketplace in the Crowdin Store for third-party localization MCPs.
Crowdin (the platform) has 500,000+ users and supports 100+ file formats, making this the most battle-tested TMS behind any MCP server.
April 2026 update: Crowdin Copilot launched — a major evolution from MCP server to AI orchestration platform. Crowdin Copilot goes beyond translation to manage your entire localization ecosystem with two built-in agents:
- Ask Agent — read-only companion that queries project stats, reviews string statuses, explores translation memory, and generates reports without risk of unintended changes
- Task Agent — full read+write access to Crowdin, can create tasks, trigger pre-translation, upload files, and manage workflows, with confirmation prompts before irreversible actions
Beyond the built-in agents, teams can create Custom Agents with tailored system prompts for specialized roles. Copilot integrates via MCP with external tools including GitHub, GitLab, Slack, Notion, Linear, PostHog, and Sentry. A standout feature is ambiguity resolution — the agent collects ambiguity flags during translation, identifies patterns, synthesizes the minimum set of questions to resolve maximum ambiguity, asks the localization manager, then returns to the pipeline with answers. Available in the Crowdin Store.
phrase/phrase-mcp-server (Official, Dual-Product)
| Server | Stars | Language | License | Tools |
|---|---|---|---|---|
| phrase/phrase-mcp-server | 3 | TypeScript | MIT | 96 |
Phrase’s official MCP server covers both of their products and has expanded significantly from 65 to 96 tools in v0.5.1 (April 24, 2026):
Phrase Strings (66 tools, up from 47):
- Project management — create, list, configure translation projects
- Job operations — create, manage, and track translation jobs
- Locale management — add, remove, and configure target languages
- Glossary creation and terminology maintenance
- Key and translation CRUD operations
- File upload and download
Phrase TMS (30 tools, up from 18):
- Project creation and management
- Async file downloads for large translation sets
- Job searching and filtering
- Translation workflow operations
Supports configurable product selection (enable Strings, TMS, or both), regional deployment (EU/US), and per-product API token authentication. 141 commits (up from 104) indicate active development. The 48% tool count increase represents one of the largest single-refresh expansions in any TMS MCP server.
AbdallahAHO/lokalise-mcp (Most Automated)
| Server | Stars | Language | License | Tools |
|---|---|---|---|---|
| AbdallahAHO/lokalise-mcp | 4 | TypeScript | MIT | 59 |
An unofficial but impressively comprehensive Lokalise MCP server with 59 tools across 11 domains and 17 pre-built automation templates:
- Strategic management — portfolio analysis, project health monitoring, team coordination
- Content operations — bulk key operations (1,000+ simultaneously), filename filtering, platform tagging
- Global expansion — language addition, progress tracking, completion analytics
- Team orchestration — user groups, permissions, workload distribution
- Workflow automation — file processing, TM + MT integration, multi-stage review pipelines
- Real-time monitoring — process dashboards, audit trails, bulk operation tracking
- Enterprise security — multi-layer authentication, secure token handling, rate limiting
The 17 prompt templates cover project management (4), localization workflows (3), advanced automation (4), and enterprise operations (5). One-click installation via Smithery. 118 commits show sustained development.
Smartling MCP Servers (Enterprise TMS — Now Two Implementations)
| Server | Stars | Language | License | Tools |
|---|---|---|---|---|
| Jacobolevy/smartling-mcp-server | 0 | TypeScript | MIT | 10+ |
| smartling/smartling-cli-mcp | 0 | JavaScript | — | 3 |
Smartling now has two MCP server implementations:
Community server (Jacobolevy) covers 10 capability areas for enterprise localization: project management, file operations, job management, string operations, QA checks, locale management, glossary tools, context management, webhook configuration, and reporting. 141 commits. Two new tools added in 2026: “Get Available Accounts for Current User” and “List File Jobs.”
Official CLI MCP (smartling/smartling-cli-mcp) — NEW as of March 2026. A Docker-based wrapper around Smartling’s official CLI tool with 3 tools (smartling-cli, smartling-ls, smartling-cat). Routes translation requests through the customer’s preferred NMT engine or LLM, applying glossaries, translation memories, and style guides. Supports text and file translation. 38 commits. Listed in the Anthropic MCP Registry.
Nativ-Technologies/nativ-mcp (Brand-Aware Translation)
| Server | Stars | Language | License | Tools |
|---|---|---|---|---|
| Nativ-Technologies/nativ-mcp | 1 | Python | MIT | 8 |
Nativ’s MCP server focuses on brand-consistent localization with 8 tools:
- translate / translate_batch — single and multi-text translation with TM, style guide, and brand voice integration
- search_translation_memory — fuzzy-search existing translations for consistency
- add_translation_memory_entry — build TM quality over time
- get_languages — access configured languages with formality settings
- get_translation_memory_stats — TM metrics and breakdown
- get_style_guides / get_brand_voice — retrieve brand guidelines for consistent output
The brand voice integration is unique — most translation servers treat every text identically, while Nativ ensures translations match your organization’s established tone and terminology.
simplelocalize/simplelocalize-mcp-server (NEW — Official TMS)
| Server | Stars | Language | License | Tools |
|---|---|---|---|---|
| simplelocalize/simplelocalize-mcp-server | 0 | TypeScript | Apache-2.0 | 14 |
SimpleLocalize’s official MCP server launched in 2026 with 14 tools for translation management:
- Bulk key creation and translation updates
- Key management — retrieve, update, delete operations
- Tag administration for organizing translation content
- Language configuration management
- Translation retrieval and search
- Hosting environment management and publishing
Installs via npx @simplelocalize/simplelocalize-mcp with API key configuration. Works with Claude Desktop, GitHub Copilot, Cursor AI, and Windsurf. 24 commits. Open to community contributions for additional tools.
IntlPull MCP Server (NEW — Commercial TMS)
IntlPull launched a commercial MCP server with 8 tools for AI-powered translation management:
- translate_strings — convert strings to target languages
- create_translation_key — generate new translation keys
- search_translations — locate existing translations
- download_translations / upload_translations — bidirectional sync with local files
- get_translation_status — monitor translation progress
- glossary_lookup — search terminology database
- tm_search — query translation memory
Key differentiators include branch management for feature development workflows and platform-specific overrides for iOS, Android, and Web. Installs via npm install -g @intlpullhq/mcp-server. Works with Claude Desktop, Claude Code, and Cursor.
Lokalise Official MCP Server (NEW — Closed Beta)
Lokalise launched an official MCP server in closed beta on February 4, 2026 — separate from AbdallahAHO’s community implementation. The official server allows connecting Lokalise to AI assistants like Claude or ChatGPT to trigger localization actions using natural language. Invitations are being extended to teams already experimenting with AI-powered development. This is the first time Lokalise has offered vendor-supported MCP access, complementing the existing community server’s 59 tools and 17 automation templates.
Developer i18n Tools
aymericzip/intlayer (Most Starred)
| Server | Stars | Language | License | Commits |
|---|---|---|---|---|
| aymericzip/intlayer | 699 | TypeScript | Apache-2.0 | 5,714 |
Intlayer is a comprehensive i18n framework for React, Next.js, Vue, and Svelte with a built-in MCP server for IDE automation. The MCP server is one component of a larger platform that includes:
- Per-component content declaration — define translations alongside your components
- MCP-powered IDE automation — run
intlayer dictionaries buildandintlayer dictionaries fillcommands through AI assistants - 231 language support via AI-powered translation
- TypeScript autocompletion for translation keys
- Visual editor for non-technical team members
- CDN delivery via Cloudflare for instant translation updates without rebuilding
- Tree-shakable dictionaries — only ship translations that are actually used
At 699 stars (+11%) and 5,714 commits (+10%), this is by far the most mature project in the i18n MCP space, though the MCP server is a relatively recent addition to an established framework.
gtrias/i18next-mcp-server (Most Developer-Focused)
| Server | Stars | Language | License | Tools |
|---|---|---|---|---|
| gtrias/i18next-mcp-server | 12 | TypeScript | MIT | 12 |
A purpose-built MCP server for managing i18next translation files with 12 tools across four areas:
Core: get_project_info, health_check, scan_code_for_missing_keys
Key management: add_translation_key, sync_missing_keys, get_missing_keys
File operations: list_files, validate_files, export_data
Analysis: coverage_report, usage_analysis, quality_analysis
The scan_code_for_missing_keys tool is particularly valuable — it analyzes your codebase for translation function calls and identifies keys that exist in code but not in your translation files. Combined with coverage_report, this gives a complete picture of your localization completeness. Configured via environment variables for project root, locales path, default language, and supported languages.
reinier-millo/i18n-mcp-server (Chunk-Based Translation)
| Server | Stars | Language | License | Tools |
|---|---|---|---|---|
| reinier-millo/i18n-mcp-server | 9 | TypeScript | AGPL-3.0 | 4 |
Takes a pragmatic approach to translating JSON language files: instead of sending entire files to a translation model at once (which can exceed context limits), it breaks translations into configurable chunks:
- Read the base language file and list of supported languages
- Extract chunks of strings (recommended: 250 entries per chunk)
- Translate each chunk sequentially
- Cache translations in memory during the workflow
- Save completed translations to language-specific JSON files
Tested with GPT-4.1 — the developer recommends generating one language at a time for faster results. The chunk-based approach makes it practical for large translation files that would otherwise exceed token limits.
better-i18n/oss (CDN-Powered)
| Server | Stars | Language | License | Commits |
|---|---|---|---|---|
| better-i18n/oss | 8 | TypeScript | — | — |
Better i18n bundles its MCP server into a broader TypeScript SDK for Next.js, React, and Expo/React Native. Key differentiators:
- GitHub Sync — translations live in your repo, changes create PRs
- Context-aware AI translation using Google Gemini (not word-by-word machine translation)
- Cloudflare CDN delivery — serve translations from the edge, update without rebuilding
- MCP server for AI coding agents to manage translations through natural language
The MCP server enables agents in Cursor, Claude Code, Windsurf, Zed, and Codex to interact with the translation workflow. Runs locally via npx @better-i18n/mcp.
dalisys/i18n-mcp (NEW — Multi-Framework i18n Management)
| Server | Stars | Language | License | Tools |
|---|---|---|---|---|
| dalisys/i18n-mcp | 10 | TypeScript | — | 16 |
A comprehensive MCP server for managing i18n translation files across multiple frameworks (React, Vue, Svelte, Angular) with 16 tools across five categories:
- Translation search and exploration — find and browse translations across all locale files
- Translation management — add, update, and organize translation entries
- Codebase analysis — detect hardcoded strings in source code and extract them for localization
- File structure validation — integrity checking across multiple language files
- Utility functions — TypeScript type generation for translation keys, unused translation cleanup
Key differentiators:
- Real-time file watching for automatic detection of translation file changes
- Hardcoded string detection — scans source code for strings that should be localized
- TypeScript type generation for compile-time safety on translation keys
- Flexible key naming — supports nested, flat, camelCase, and kebab-case styles
- Unused translation cleanup — identifies and removes dead translations
Integrates with Claude Desktop, VS Code, and Zed. At 10 stars and 13 commits, this is the most feature-rich standalone i18n MCP server, addressing workflow gaps that framework-specific servers like i18next-mcp-server don’t cover.
Platform-Specific Localization
zhangyu1818/xcode-i18n-mcp (iOS/macOS)
| Server | Stars | Language | License | Tools |
|---|---|---|---|---|
| zhangyu1818/xcode-i18n-mcp | 3 | TypeScript | MIT | 3 |
Three focused tools for Xcode String Catalog (.xcstrings) localization:
- getKnownRegions — extract supported languages from
.pbxprojfiles - getStringsToTranslate — identify untranslated strings for specific language codes
- updateTranslations — write translated strings back to String Catalog files
Supports Xcode 15+ String Catalog format (JSON-based). Type-safe implementation with Zod schema validation and comprehensive error handling. The entire codebase was generated through AI assistance.
realskyrin/android-i18n-mcp (Android)
Android localization through Git-aware string resource management:
- Change detection — uses Git diff to identify modified strings in default
strings.xmlfiles across all modules - 28 language support — configurable via
TRANSLATION_LANGUAGESenvironment variable - Auto-creation — creates missing language directories and populates them with translations
Practical for teams that want to keep translations in sync with code changes — when you modify a string in your default locale, the server detects the change and translates it to all configured languages.
doubleA411/figma-i18n-mcp (Design-to-i18n)
Extracts text from Figma frames and designs into i18n-ready JSON format:
- Automatically extracts all visible text nodes from Figma files
- Organizes text into structured JSON grouped by frames and components
- Outputs ready for direct use with i18n libraries
Bridges the gap between design and development localization — designers define text in Figma, this server extracts it into a format developers can use directly in their i18n setup.
What’s Missing
The translation and localization MCP landscape has meaningful gaps:
No Google Cloud Translation MCP server. Despite Google now offering 30+ managed remote MCP servers for BigQuery, Cloud SQL, Spanner, Compute Engine, Cloud Run, and many more — there is still no dedicated Cloud Translation MCP server. Google published a blog post on building multilingual chatbots with Gemini, Gemma, and MCP, but this is an architecture pattern, not a standalone translation server. This remains the most surprising omission in the category.
No Transifex or Mozilla Pontoon MCP servers. Two major open-source localization platforms have no MCP presence. Transifex serves large open-source projects, and Pontoon powers Mozilla’s extensive localization ecosystem.
Limited local translation options. TranslateGemma is the only fully local option, and its 82GB Docker image makes it impractical for most development workflows. There’s room for lighter-weight local translation servers using smaller models.
No cross-TMS abstraction. Organizations using multiple TMS platforms (common during migrations or with different teams) have no way to manage translations across Crowdin, Phrase, and Lokalise through a unified MCP interface.
Subtitle/caption localization partially addressed. Video Caption MCP on Apify and the Reap MCP server now offer subtitle translation in 98+ languages (covered in our Video Production & Streaming review), but no server in this category focuses specifically on subtitle translation workflow management with timing synchronization.
Limited right-to-left (RTL) tooling. No server specifically addresses RTL language challenges — bidirectional text handling, layout mirroring, or Arabic/Hebrew-specific translation quality checks.
The Bottom Line
The translation and localization MCP ecosystem earns 4.0 out of 5 (up from 3.5). The April 2026 refresh reveals meaningful maturation: Crowdin Copilot’s AI orchestration goes beyond MCP tooling into autonomous localization management, Phrase nearly doubled its tool count to 96, three new TMS vendors entered (SimpleLocalize, IntlPull, Smartling official), and dalisys/i18n-mcp brought hardcoded string detection to the developer i18n subcategory.
Best translation API wrapper: DeepL MCP Server — official, 102 stars, glossary support, document translation, free tier available.
Best TMS integration: Crowdin MCP Server + Copilot — 200+ tools plus AI orchestration with Ask Agent and Task Agent, hosted infrastructure, battle-tested platform with 500,000+ users.
Best developer i18n tool: dalisys/i18n-mcp — 16 tools with hardcoded string detection, TypeScript type generation, multi-framework support, and unused translation cleanup. Runner-up: i18next-mcp-server for i18next-specific workflows.
Best for privacy: TranslateGemma — the only fully local option with 55-language support via Google’s open models.
Best for brand consistency: Nativ MCP — unique brand voice and style guide integration for on-brand translations.
Most improved: Phrase MCP Server — 65→96 tools (+48%) in v0.5.1, now one of the most comprehensive TMS MCP servers.
The category’s strength is vendor participation — most servers are official or semi-official, which means better maintenance and alignment with platform capabilities. The addition of SimpleLocalize, IntlPull, Smartling’s official CLI server, and Lokalise’s closed beta shows the TMS industry is now fully committed to MCP. The weakness remains adoption metrics: outside DeepL (102 stars) and Intlayer (699 stars), most servers have minimal GitHub stars. But the trend is clear — enterprise TMS platforms are investing heavily in MCP, and developer i18n tools are maturing rapidly.
This review was refreshed on 2026-04-28 using Claude Opus 4.6 (Anthropic). Original review: 2026-03-15.