volcengine-mcp

[!License: MIT](https://opensource.org/licenses/MIT) [!MCP](https://modelcontextprotocol.io) [!Node.js](https://nodejs.org)

MCP server for ByteDance Volcano Engine. Enables text generation, multi-turn chat, and embeddings via Doubao LLM models through the Volcano Ark API.

Tools (4 total)

Tool	Description
doubao_generate	Generate text using Doubao models (single-turn completion)
doubao_chat	Multi-turn conversation with message history
doubao_embeddings	Generate text embeddings for semantic search and RAG
doubao_list_models	List available Doubao chat and embedding models

Available Models

Chat Models

Model ID	Context	Description
doubao-pro-32k	32K	General purpose, large context
doubao-pro-4k	4K	General purpose, fast
doubao-lite-32k	32K	Lightweight, large context
doubao-lite-4k	4K	Lightweight, fast
doubao-seed-1-6-250615	32K	Latest with vision support
doubao-seed-1-6-flash-250615	32K	Fast thinking, low latency
doubao-seed-1-6-thinking-250615	32K	Deep reasoning for complex tasks

Embedding Models

Model ID	Dimensions	Description
doubao-embedding	2560	Text embeddings
doubao-embedding-text-240715	2560	Latest text embeddings

Install

npm install

Configuration

{
  "mcpServers": {
    "volcengine": {
      "type": "stdio",
      "command": "node",
      "args": ["/path/to/volcengine-mcp/index.js"],
      "env": {
        "ARK_API_KEY": "your-api-key"
      }
    }
  }
}

Environment Variables

Variable	Description	Required
ARK_API_KEY	Volcano Ark API key	Yes
VOLCENGINE_API_KEY	Alternative API key variable	Yes (if ARK_API_KEY not set)
ARK_BASE_URL	API base URL	No (defaults to Beijing endpoint)
ARK_CHAT_MODEL	Default chat model	No (defaults to doubao-pro-32k)
ARK_EMBEDDING_MODEL	Default embedding model	No (defaults to doubao-embedding)

Getting an API Key

1. Sign up at console.volcengine.com
2. Complete real-name authentication (required for Chinese cloud services)
3. Navigate to Large Models > Volcano Ark > API Key Management
4. Create and copy your API key

Use Cases

• Chinese language tasks -- Doubao excels at Chinese NLP
• Low-latency inference -- Doubao Flash for real-time applications
• Deep reasoning -- Doubao Thinking for complex analysis
• Cost-effective bulk processing -- Doubao Lite for high-volume tasks
• Semantic search and RAG -- Doubao embeddings (2560 dimensions)

Dependencies

• @modelcontextprotocol/sdk -- MCP protocol SDK

License

MIT