Skip to main content
Glama
README.md
# konid

*konid* (کنید) — Farsi for "do." Take action.

A language expression coach. Tell it what you want to say, and it shows you how a native speaker would actually say it — with cultural context, tone notes, and audio pronunciation.

<p align="center">
  <img src="docs/hero.gif" alt="Google Translate says 'estoy caliente' for 'I'm hot', which actually means 'I'm horny' in Spanish" width="700">
</p>

## How it works

<a href="https://www.loom.com/share/9ae0d46f7c6d4a6caa2847351aa18ae8">
  <img src="docs/cowork-demo.png" alt="Watch the demo video" width="500">
</a>

**You:** "how do I say 'we'll see' in Chinese?"

**konid returns 3 options, casual to formal:**

1. **再说吧** (zài shuō ba) — casual, slightly evasive, can function as a soft "no"
2. **看情况吧** (kàn qíngkuàng ba) — "depends on the situation," genuinely open-minded
3. **到时候再看吧** (dào shíhou zài kàn ba) — "let's wait and see," most neutral

Plus cultural context, nuance comparison, and audio pronunciation.

Supports 13+ languages including Mandarin, Japanese, Korean, Spanish, French, German, Portuguese, Italian, Russian, Arabic, and Hindi. Any language Claude knows can be coached.

## Install

### Claude Cowork

> Requires a Pro or Max subscription.

1. Download [`konid-ai-plugin.zip`](docs/konid-ai-plugin.zip)
2. In Cowork, click **Customize** in the left sidebar
3. Click **+** next to "Personal plugins" > **Upload plugin**
4. Choose the `konid-ai-plugin.zip` file

<p align="center">
  <img src="docs/cowork-upload.png" alt="Cowork: Customize > + > Upload plugin" width="500">
</p>

That's it. Start a new task and ask *"how do I say 'nice to meet you' in Japanese?"*

### Claude Code

Requires Node.js 18+ and an [Anthropic API key](https://console.anthropic.com/settings/keys).

```bash
claude mcp add konid-ai -- npx -y konid-ai
```

Audio plays directly through your speakers.

### Other MCP clients

konid works with any MCP-compatible client. Add this to your client's MCP config:

```json
{
  "mcpServers": {
    "konid": {
      "command": "npx",
      "args": ["-y", "konid-ai"]
    }
  }
}
```

Tested with Cursor, VS Code Copilot, Windsurf, Zed, and JetBrains.

### ChatGPT

Settings > Apps > Advanced settings > Developer mode > add `https://konid.fly.dev/mcp`

## License

MIT

TDQS

A4.3/5.0

Scored across 3 tools

Disambiguation5/5

Each tool has a clearly distinct purpose: coach generates language options, speak produces audio from text, and replay repeats the last audio. There is no overlap in functionality, so an agent can select the correct tool without confusion.

Naming Consistency5/5

All tool names are single lowercase verbs (coach, speak, replay) that directly describe their action. The naming convention is perfectly uniform and instantly understandable.

Tool Count5/5

With only 3 tools, the server is tightly scoped to a simple workflow: generate a phrase, hear it, and repeat it. Each tool is necessary and the count is ideal for the intended purpose.

Completeness5/5

The tool surface covers the full language pronunciation loop: creating natural expressions, listening to them via TTS, and replaying on demand. There are no obvious gaps or dead ends for the intended use case.

Maintenance

ActivityInactive
ResponsivenessNo issues