Lingo.dev MCP Server: Open-Source AI i18n Engine

Overview

The Lingo.dev MCP (Model Context Protocol) server is an open-source assistant-facing service that provides structured, framework-aware i18n guidance to AI coding assistants. Instead of letting an LLM guess at project structure and file mutations, the MCP server exposes verified implementation patterns for popular web frameworks so assistants can perform coordinated, repeatable localization tasks (routing, middleware, components, config) with predictable results.

MCP is useful when you want AI-assisted setup for localization—especially for frameworks where i18n involves changes across multiple files (Next.js, React Router, TanStack Start). It integrates with agent-enabled IDEs and gives assistants the context they need to apply correct, framework-specific changes while avoiding common hallucinations about APIs or project layout.

Features

• Framework-specific i18n knowledge for Next.js (App & Pages Router), React Router, and TanStack Start
• Structured guidance for AI assistants (Claude Code, Cursor, GitHub Copilot Agents, Codex)
• Locale-aware routing, automatic locale detection, and language switcher generation
• Safe, repeatable generation patterns that minimize broken setups
• Integrates with the Lingo.dev localization pipeline (CLI, SDK, Compiler)
• Configurable LLM provider: use Lingo.dev Engine or plug your own (OpenAI, Anthropic, Google, etc.)

Installation / Configuration

The MCP server is part of the Lingo.dev toolkit. Below are common steps to initialize localization and run the CLI, plus examples of linking an LLM provider.

Initialize a project and create baseline configuration:

# Initialize i18n config in your repo (creates i18n.json)
npx lingo.dev@latest init

# Run translation pipeline (syncs locales)
npx lingo.dev@latest run

Sample i18n.json (project configuration):

{
  "$schema": "https://lingo.dev/schema/i18n.json",
  "version": "1.10",
  "locale": {
    "source": "en",
    "targets": ["es", "pt-BR"]
  },
  "buckets": {
    "json": {
      "include": ["locales/[locale].json"]
    }
  }
}

Optional provider override (example: OpenAI):

{
  "provider": {
    "id": "openai",
    "model": "gpt-4o-mini",
    "prompt": "Translate from {source} to {target}"
  }
}

Configuring the MCP server for your IDE agents typically requires:

• Running the MCP server locally or hosting it on an internal endpoint
• Pointing your agent/IDE integration to the MCP endpoint (URL + optional auth token)
• Granting the agent access to the repository or workspace so it can propose file edits

Example assistant prompt you can send after MCP is configured:

Set up i18n with the following locales: en, es, and pt-BR. The default locale is 'en'.

The assistant will then apply framework-specific changes (routing, detection, language switcher, config files).

Note: LLM-generated changes are non-deterministic. Always review generated code before committing or deploying.

Available Resources

• GitHub repository (source & docs): https://github.com/lingodotdev/lingo.dev/blob/main/mcp.md
• Official docs: https://lingo.dev/en/mcp
• CLI: npx lingo.dev@latest (init, run)
• SDK & Compiler: runtime and build-time tools available via the Lingo.dev package (npm)
• Schema and configuration reference: https://lingo.dev/schema/i18n.json

Use Cases

• Quick i18n bootstrap for a React/Next.js project
  • Prompt your agent to “Set up i18n with en, es, fr” and let MCP drive a consistent set of edits across routing, middleware, and layout files.
• Add new locales to an existing app
  • Update i18n.json targets and run npx lingo.dev@latest run to extract strings and generate translations; a lockfile (i18n.lock) prevents re-translating unchanged content.
• Migrate from ad-hoc translations to a maintained pipeline
  • Use the CLI to centralize translations across JSON, YAML, CSV, PO, and Markdown files; the MCP server helps agents modify app code to consume the organized locales.
• Agent-assisted refactor with safe patterns
  • When adopting a new routing strategy (e.g., Next.js App Router), the MCP server ensures the assistant follows verified patterns for locale routing and middleware placement.

Practical Tips

• Review automated edits: AI tools can make useful but imperfect changes—run tests and linting after applying patches.
• Use i18n.lock to avoid duplicate translation costs when using an LLM provider.
• Prefer the Lingo.dev Engine for translation continuity, but configure custom providers if you have existing LLM contracts.
• Start MCP locally when experimenting, then move to a hosted internal endpoint for team-wide agent workflows.

Supported Frameworks & IDE Agents

• Frameworks: Next.js (App & Pages Router v13–16), React Router (v7), TanStack Start (v1)
• IDE/Agent integrations: Claude Code, Cursor, GitHub Copilot Agents, Codex (OpenAI)

For more details, examples, and advanced configuration patterns, see the MCP documentation and the repository link above.