Tongyi Wanxiang MCP Server

Instructions

Implementation Reference

• src/index.ts:52-62 (registration)

  Registers the MCP tool 'wanx-t2v-video-generation' with description, input schema {prompt: z.string()}, and handler that calls generateVideo and returns task info.

  server.tool(
    "wanx-t2v-video-generation",
    "使用阿里云万相文生视频大模型的文生视频能力，由于视频生成耗时比较久，需要调用 wanx-t2v-video-generation-result 工具获取结果",
    { prompt: z.string() },
    async ({ prompt }) => {
      const result = await generateVideo(prompt);
      return {
        content: [{ type: "text", text: JSON.stringify(result) }],
      };
    }
  );

• src/index.ts:55-55 (schema)

  Input schema using Zod for the tool parameters.

  { prompt: z.string() },

• src/wanx-t2v.ts:9-37 (handler)

  Core handler function 'generateVideo' that makes async API call to Aliyun Dashscope to initiate video generation task and returns the task_id.

  export async function generateVideo(prompt: string) {  
    const apiKey = config.api.apiKey;
    const url = 'https://dashscope.aliyuncs.com/api/v1/services/aigc/video-generation/video-synthesis'; 
    const headers = {  
      'X-DashScope-Async': 'enable',
      'Authorization': `Bearer ${apiKey}`,  
      'Content-Type': 'application/json'  
    };  
    const payload = {  
      model: 'wanx2.1-t2v-turbo',  
      input: { prompt },
      parameters: {
        size: '832*480',
        duration: 5,
        prompt_extend: true,
      },
    };
  
    try {  
      const res = await axios.post(url, payload, { headers });  
      const taskId = res.data?.output?.task_id;  
      if (taskId) return taskId;  
      // 如果响应里没有task_id，说明有错误  
      throw res.data;  
    } catch (err: any) {  
      // err.response?.data 可能含详细错误  
      throw err.response?.data || err.message;  
    }  
  }