@sonpiaz/kyma-dub-mcp

MCP stdio server for kyma-dub — dub videos into another language with a natural, time-aligned AI voiceover, plus live model/voice discovery. Brings kyma-dub into MCP-capable desktops (Claude desktop, Cursor, ChatGPT desktop).

Prerequisite

The kyma-dub CLI must be installed and on PATH (it does the work; this server just wraps it):

curl -fsSL https://github.com/sonpiaz/kyma-dub/releases/latest/download/install.sh | bash

Set KYMA_API_KEY (in ~/.config/kyma-dub/env or your environment). Override the CLI path with KYMA_DUB_BIN if it isn't on PATH.

Install & build

cd mcp-server
npm install
npm run build

Configure your host

Claude desktop (claude_desktop_config.json) / Cursor (~/.cursor/mcp.json):

{
  "mcpServers": {
    "kyma-dub": {
      "command": "npx",
      "args": ["-y", "@sonpiaz/kyma-dub-mcp"],
      "env": {
        "KYMA_API_KEY": "kyma-xxxxxxxx",
        "KYMA_DUB_BIN": "/Users/you/.local/bin/kyma-dub"
      }
    }
  }
}

Set KYMA_DUB_BIN to the absolute path of the kyma-dub CLI (run which kyma-dub). Desktop apps spawn the server with a minimal PATH that usually omits ~/.local/bin, so without this the server reports tag=missing-dep:kyma-dub.

Published on npm: @sonpiaz/kyma-dub-mcp. (For local dev, point command to node + dist/index.js instead.)

Tools

| Tool | What it does | |---|---| | dub_start | Start dubbing a local video. Long-running → returns a job_id. | | dub_status | Poll a dub job: running / done (with output_path) / error. | | list_models | Translation-capable models on Kyma right now (live). | | list_voices | Voices + gender/age/use-case/accent/language labels; library to search the shared catalog. | | recommend_voice | Suggest voices (+ best model) for a need; smart lets a Kyma model pick. | | preview_voice | Synthesize a short sample so the user can hear a voice first. | | whatsnew | What models/voices Kyma added since last check. |

Resources

| URI | Content | |---|---| | kyma-dub://models | Live translation models (JSON) | | kyma-dub://voices | Live account voices with labels (JSON) |

Dubbing is job-based (dub_start → dub_status) because a dub takes minutes — this avoids host call-timeouts. Discovery tools are synchronous.

License

MIT © Son Piaz