wove for AI Agents

Wove: MCP server for AI-driven text localization and translation

wove, developed by Mits Pl, is an MCP server that automates AI-driven text localization and translation for software projects. The app connects AI agents to localization workflows so models can translate, update JSON resource files, and maintain consistent terminology across regions. Key functions include Model Context Protocol integration, JSON localization support, and automated pipelines for UI strings and documentation. The target audience is software developers and localization engineers seeking agent-driven i18n tooling.

Agents can discover and invoke localization services inside MCP environments

The server is built as a Model Context Protocol host, which means agents can find and call its APIs rather than relying on manual CLI scripts. It explicitly lists compatibility with MCP hosts such as Claude Desktop and runs as a Node.js server, so developers deploy it inside existing agent-enabled environments. Deployment facts:

• MCP discovery and invocation
• Node.js runtime requirement
• Designed for agent use rather than manual-only workflows

Translation outputs depend on the chosen language model and supplied context

The tool uses advanced language models to produce context-aware translations, so output quality varies with the underlying model and the domain information provided. It supplies application domain metadata to support consistent terminology, which improves results for subject-specific strings. Practical implication: teams should review model-generated strings for technical or regulated content because translations reflect model behavior and prompt specificity.

It processes JSON resource files and can update keys programmatically

The server is optimized for the common software localization format JSON, letting connected agents read, translate, and write back resource files while preserving key structures. That design supports automated pipelines for UI text and documentation, enabling agents to operate on project files directly instead of producing standalone translation artifacts. Workflow note: test runs with representative JSON sets reveal key-mapping behavior before full-scale application.

Open-source code and agentic design suit engineering and localization teams

The codebase is public on GitHub and the project is positioned for community audit and contribution, so teams can inspect logic and adapt behavior. Mits Pl builds the server for integration into developer workflows, and the project is noted within the MCP developer community as a utility for extending agent capabilities into localization. Team fit: engineers comfortable with Node.js and agent orchestration gain the most immediate value.

Wove is practical for teams ready to validate model outputs

Use wove when you want agent-driven localization integrated into development workflows, and start with a small corpus of strings to establish model prompts and verification steps. Expect translation quality to reflect the chosen model and the prompts you give it, so incorporate human review for specialized terminology. In practice, wove suits engineering teams that accept model-dependent outputs and plan for systematic verification before release.