The problem with every translation tool I've used: you get one answer, and you have no idea if it's the right register for the situation. Sending a message that's too formal to a friend, or too casual to a client, can read worse than saying nothing.
konid is an MCP server I built to fix that. For anything you want to say, it returns three options ordered casual to formal, explains why each register fits a different context, and plays audio pronunciation through your speakers using node-edge-tts — no API key needed.
I set it up to run across four clients from a single server, because I switch between them depending on whether I'm coding, writing, or in a meeting:
Claude Code:
claude mcp add konid-ai -- npx -y konid-ai
Cursor / VS Code Copilot / Windsurf / Zed / JetBrains:
Same npx command in your MCP config file.
ChatGPT and Claude Cowork:
Remote endpoint: https://konid.fly.dev/mcp — paste it in ChatGPT's Developer mode under Actions, no local install.
The remote endpoint approach is worth calling out. MCP is usually described as a local tool protocol, but hosting the server on fly.dev means ChatGPT users get the same three-option output — cultural context, nuance comparison, audio — without running anything on their machine. Same behavior, different surface.
Supports 13+ languages: Mandarin, Japanese, Korean, Spanish, French, German, Portuguese, Italian, Russian, Arabic, Hindi, and more.
MIT licensed. https://github.com/robertnowell/konid-language-learning
Top comments (0)