qwen-omni-mcp

An MCP server that gives Claude Code and other AI agents video, image, audio, and audio-video understanding via Bailian (DashScope) using the multimodal Qwen3.7-Plus and Qwen3.5-Omni models.

Qwen3.7-Plus reads video natively — no client-side frame extraction. Qwen3.5-Omni adds native audio understanding (and audio-track awareness for video). Pass a public media URL or a local file path; the model does the rest.

Highlights

• Native video understanding — send a video URL or local file, get grounded analysis

• Image understanding — describe, Q&A, OCR

• Audio understanding — transcribe, summarize, analyze speech/sound (mp3/wav/flac/ogg/m4a/aac)

• Audio-video understanding — analyze a video's visuals and its sound track together

• Local file support — pass a local path; files are sent inline as base64 (25MB guardrail)

• npx-launchable — one line in your MCP client config

Related MCP server: Qwen Video Understanding MCP Server

Install

No global install needed. Run directly with npx:

npx -y qwen-omni-mcp

For local development:

git clone <this-repo>
cd qwen-omni-mcp
npm install            # also installs husky git hooks
cp .env.example .env   # fill in DASHSCOPE_API_KEY
npm run dev            # run from source via tsx

Configuration

All config is via environment variables (loaded from .env by dotenv):

Get a key at https://bailian.console.aliyun.com/cn-beijing?tab=model#/api-key.

The Anthropic-compatible /apps/anthropic endpoint does not support video input, so this server uses the OpenAI-compatible endpoint.

Use with Claude Code

Add to your MCP client config:

{
  "mcpServers": {
    "qwen-omni-mcp": {
      "command": "npx",
      "args": ["-y", "qwen-omni-mcp"],
      "env": {
        "DASHSCOPE_API_KEY": "your-key"
      }
    }
  }
}

For local development without publishing:

{
  "mcpServers": {
    "qwen-omni-mcp": {
      "command": "npx",
      "args": ["tsx", "src/index.ts"],
      "env": { "DASHSCOPE_API_KEY": "your-key" }
    }
  }
}

Tools

Each media tool accepts a public http/https URL or a local file path. Local files are read and sent inline as base64, with a 25MB guardrail (verified up to a 14MB video / ~18MB body on Qwen3.7-Plus, and an 8.8MB video / ~11.7MB base64 body on Qwen3.5-Omni, both HTTP 200). Files larger than 25MB must be hosted at a public URL instead. Local input is validated by extension + magic-byte signature before encoding, so non-media files are rejected.

analyze_audio / analyze_audio_video use the omni model (QWEN_OMNI_MODEL, default qwen3.5-omni-plus) and force text-only output. Audio is sent as an input_audio block in the data:;base64,<b64> form with a format field (mp3/wav/flac/ogg/m4a/aac).

Development

npm run typecheck     # strict tsc
npm run lint          # eslint, --max-warnings 0
npm run format:check  # prettier
npm test              # unit + mocked e2e (no API cost)
npm run build         # emit dist/
LIVE=1 npm run test:live   # real API calls (costs tokens)

CI (.github/workflows/ci.yml) runs the same gates on Node 20/22. secrets-scan.yml runs gitleaks. smoke-live.yml (manual / weekly) runs one real image call.

See AGENTS.md for the full set of agent rules (never bypass hooks, never commit secrets, etc.).

License

MIT