About Alibaba Cloud DashScope — Qwen
An MCP server for Alibaba Cloud DashScope that calls the OpenAI-compatible REST API directly with the user's own key. It provides Qwen chat, embeddings and model listing. The project is community-built and not affiliated with Alibaba Cloud.
Key features
- 01Qwen chat completion with several models
- 02text-embedding-v3 text embeddings
- 03List models available on the account
- 04Direct API calls without an intermediary backend
Use cases
- 01Use Qwen models from Claude Code or Cursor
- 02Generate embeddings for semantic search or RAG
- 03List which DashScope models an account can use
MCP support
The server exposes three MCP tools: qwen_chat, qwen_embed and qwen_list_models.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
npx -y @three-ws/alibaba-cloud-mcpDASHSCOPE_API_KEYSecretDashScope API key. Create one at https://bailian.console.alibabacloud.com/ → API Keys.DASHSCOPE_MODEL_IDDefault chat model id. Options: qwen-max, qwen-plus (default), qwen-turbo, qwen-long.DASHSCOPE_EMBED_MODEL_IDDefault embedding model id. Defaults to text-embedding-v3.DASHSCOPE_BASE_URLCustom DashScope endpoint. Defaults to international (dashscope-intl.aliyuncs.com).DASHSCOPE_REGIONSet to 'cn' to use the China endpoint (dashscope.aliyuncs.com). Ignored if DASHSCOPE_BASE_URL is set.Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.
Tools
3Saved tool list · updated 2w ago
| Parameter | Type | Required | Description |
|---|---|---|---|
| model | string | No | Override the model id. Options: qwen-max, qwen-plus (default), qwen-turbo, qwen-long, qwen-max-latest. |
| top_p | number | No | Nucleus sampling probability mass. |
| messages | array | Yes | Conversation so far, oldest first. |
| max_tokens | integer | No | Maximum tokens to generate. |
| temperature | number | No | Sampling temperature. 0 is greedy/deterministic. |
Details
From the project README
@three-ws/alibaba-cloud-mcp
MCP server for Alibaba Cloud DashScope, Qwen chat, embeddings, and model discovery on your own account.
Install · Quick start · Tools · Models · three.ws
A Model Context Protocol server that exposes Alibaba Cloud's Qwen models, and anything else in your DashScope account, to MCP clients such as Claude Desktop, Claude Code, and Cursor. It talks directly to the DashScope OpenAI-compatible REST API with your own Alibaba Cloud API key: no intermediary backend, no telemetry, no mock data.
Community-built and not affiliated with Alibaba Cloud. Registry name:
io.github.nirholas/alibaba-cloud.
Install
npm install @three-ws/alibaba-cloud-mcp
Run it directly with npx (no install needed):
DASHSCOPE_API_KEY=sk-... npx @three-ws/alibaba-cloud-mcp
Quick start
With Claude Code, one command:
claude mcp add alibaba-cloud -e DASHSCOPE_API_KEY=sk-... -- npx -y @three-ws/alibaba-cloud-mcp
Or add the server to your MCP client config:
{
"mcpServers": {
"alibaba-cloud": {
"command": "npx",
"args": ["-y", "@three-ws/alibaba-cloud-mcp"],
"env": {
"DASHSCOPE_API_KEY": "sk-..."
}
}
}
}
Requirements
- An Alibaba Cloud account
- A DashScope API key, navigate to API Keys and create one
- Node.js ≥ 20
Tools
| Tool | Description |
|---|---|
qwen_chat | Chat completion with any Qwen model (qwen-max, qwen-plus, qwen-turbo, qwen-long) |
qwen_embed | Text embeddings via text-embedding-v3 (1024-dim default, up to 2048) |
qwen_list_models | List all models available on this DashScope account |
Models
| Model ID | Use |
|---|---|
qwen-max | Highest quality: complex reasoning, long documents |
qwen-plus | Balanced performance (default) |
qwen-turbo | Fastest and cheapest: simple tasks |
qwen-long | Long context (up to 1M tokens) |
text-embedding-v3 | Embeddings, semantic search, RAG (default) |
Environment variables
| Variable | Required | Default | Description |
|---|---|---|---|
DASHSCOPE_API_KEY | yes | - | Your DashScope API key |
DASHSCOPE_MODEL_ID | no | qwen-plus | Default chat model |
DASHSCOPE_EMBED_MODEL_ID | no | text-embedding-v3 | Default embedding model |
DASHSCOPE_BASE_URL | no | International endpoint | Custom DashScope endpoint |
DASHSCOPE_REGION | no | - | Set to cn for the China endpoint |
DASHSCOPE_TIMEOUT_MS | no | 60000 | Request timeout in ms |
Inspect
npx -y @modelcontextprotocol/inspector npx @three-ws/alibaba-cloud-mcp
Programmatic use
The package entry point (@three-ws/alibaba-cloud-mcp) exports buildServer(): a
fully-registered McpServer with no transport attached, so you can mount the three tools
inside a host of your own instead of spawning the binary. It reads DASHSCOPE_API_KEY when
called, and registration is otherwise side-effect free. The DashScope client itself is
importable from @three-ws/alibaba-cloud-mcp/src/dashscope.js (loadConfig,
DashScopeClient, DashScopeError) when you want the REST wrapper without MCP.
// example.mjs: run with: DASHSCOPE_API_KEY=sk-... node example.mjs
import { StdioServerTransport } from '@modelcontextprotocol/sdk/server/stdio.js';
import { buildServer } from '@three-ws/alibaba-cloud-mcp';
const server = buildServer();
// What this server exposes, before connecting anything.
for (const [name, tool] of Object.entries(server._registeredTools)) {
console.log(`${name}: ${tool.title}`);
}
// Serve it over stdio (what the bin does), or attach any other MCP transport.
await server.connect(new StdioServerTransport());
qwen_chat: Qwen Chat
qwen_embed: Qwen Text Embeddings
qwen_list_models: List DashScope Models
Built by three.ws · three.ws/blog/three-ws-alibaba-cloud-partnership
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.