anubis-mcp, developed by Zoedsoupe, is an open-source Model Context Protocol server for AI-assisted text localization. It connects language models to project translation files so developers can request localized changes in natural language. The server exposes task-specific tools for listing keys, fetching missing strings, and applying updates. Intended for software developers, i18n engineers, and localization managers who want programmatic, context-aware editing of translation files within their development workflow.

What tasks the tool actually automates for localization teams

anubis-mcp implements a server-side bridge that lets an AI operate on a codebase's translation assets, turning natural-language prompts into file edits. The repository documents "chat-to-localize" examples where a developer asks an assistant to translate a login page and the server writes updated strings back into the project. That workflow removes manual key lookups by exposing scripted tools the model can call to inspect and modify i18n content.

How translation correctness is produced and what affects it

Context is supplied to the connected model so translations can reflect UI intent and structure; the project includes logic for giving the model surrounding keys and file paths. Accuracy therefore depends on the connected language model's outputs, not the server itself. The server's role is to present structured context and apply edits; produced translations still require human review before merging into source control.

What inputs and hosts it accepts, and format support

The server runs under an MCP host and Node.js; it requires an MCP-compatible environment such as Claude Desktop to accept model calls, and typical installation paths include npm or cloning the GitHub repository and configuring the host settings. File-format handling covers nested JSON structures used by i18next and other common localization formats, allowing the tools to traverse and update hierarchical keys.

How it fits into developer workflows and community extension

Designed for integration into existing dev cycles, the project exposes specific tool endpoints for listing keys, fetching missing strings, and applying updates so teams can place AI-suggested edits behind pull requests or CI checks. The codebase is open source on GitHub, enabling custom extensions and community-driven adapters for additional MCP hosts or bespoke file formats.

Practical choice for teams that want programmatic, reviewable AI edits

anubis-mcp suits teams that need model-driven edits applied directly to repository files while keeping human oversight. As a practical tip, use focused, context-rich prompts and gate AI changes through code review so model-generated translations are validated before deployment. For organizations that require an automated, reviewable path from prompt to file change, it is a pragmatic option that integrates with existing developer workflows.