WigAI for AI Agents

Local MCP server for AI-assisted software localization

WigAI, from Fabb, is an MCP server built to embed AI into software localization workflows, connecting language models to a project’s translation files. The tool reads, translates, and edits localization keys via the Model Context Protocol, with support for context-aware handling, pluralization, and multiple file formats. It targets developers and localization engineers who need AI-generated draft translations integrated directly into their codebase and version-control process.

What tasks can you actually use it for?

The tool functions as a bridge between language models and on-disk localization assets, so you can automate routine i18n tasks: generate draft translations, detect missing keys, and manage pluralization and variable interpolation. It works with common formats such as JSON, .strings, .stringsdict, and .xcstrings, and it can suggest values for absent keys. These capabilities let AI assistants operate directly on translation files rather than producing separate copy to be pasted back.

How accurate is the tool's context-aware output?

The server provides context-aware localization by surfacing placement and usage information to the model, which improves phrasing choices for UI text. Because the tool sends strings to an AI model via the Model Context Protocol, generated translations reflect the chosen model’s behavior and should be reviewed for critical content. Handling of plural rules and interpolation reduces mechanical errors that often occur when tools treat strings as isolated fragments.

Is it practical to integrate into a developer workflow?

The implementation is developer-focused: it runs locally in a Node.js environment and is written in TypeScript, and it is designed to work with MCP-compatible hosts such as Cursor, Windsurf, and Claude Desktop. Operating on local files allows teams to verify changes in their normal code-review and version-control processes. The project is open source, which lets teams adapt the server to specific repository layouts and CI checks.

A practical assistant for engineering-led localization workflows

The tool is a pragmatic option for engineering teams that want model-generated draft translations inside their repos while retaining editorial control. Expect to pair generated text with human review and to invest modest setup effort integrating it into existing CI and review gates. Treated as a drafting aid rather than a final translator, it can reduce repetitive work without replacing specialist review.