MusicGPT MCP Server

Instructions

Implementation Reference

• src/index.ts:829-852 (handler)

  The core handler function that implements the generate_music tool logic by making an API call to the MusicGPT /MusicAI endpoint with the provided arguments and returning a status response.

  private async handleGenerateMusic(args: any) {
    if (!args.prompt) {
      throw new McpError(ErrorCode.InvalidParams, "prompt is required");
    }
  
    const response = await this.axiosInstance.post("/MusicAI", {
      prompt: args.prompt,
      music_style: args.music_style,
      lyrics: args.lyrics,
      make_instrumental: args.make_instrumental || false,
      vocal_only: args.vocal_only || false,
      voice_id: args.voice_id,
      webhook_url: args.webhook_url,
    });
  
    return {
      content: [
        {
          type: "text",
          text: `Music generation started!\n\n${JSON.stringify(response.data, null, 2)}\n\nUse get_conversion_by_id with the task_id or conversion_id to check the status.`,
        },
      ],
    };
  }

• src/index.ts:122-160 (schema)

  The input schema and description for the generate_music tool, defining parameters like prompt, music_style, lyrics, etc., used for validation in MCP.

    name: "generate_music",
    description: "Generate custom music from a text prompt using AI. Can create songs with or without lyrics, instrumental tracks, or vocal-only versions.",
    inputSchema: {
      type: "object" as const,
      properties: {
        prompt: {
          type: "string",
          description: "Natural language prompt for music generation (keep under 280 characters for best results)",
        },
        music_style: {
          type: "string",
          description: "Style of music to generate (e.g., Rock, Pop, Jazz, Hip-Hop)",
        },
        lyrics: {
          type: "string",
          description: "Custom lyrics for the generated music",
        },
        make_instrumental: {
          type: "boolean",
          description: "Whether to make the music instrumental (no vocals)",
          default: false,
        },
        vocal_only: {
          type: "boolean",
          description: "Whether to generate only vocals of output audio",
          default: false,
        },
        voice_id: {
          type: "string",
          description: "Voice model ID to use for vocals (use get_all_voices to find IDs)",
        },
        webhook_url: {
          type: "string",
          description: "URL for callback upon completion",
        },
      },
      required: ["prompt"],
    },
  },

• src/index.ts:669-670 (registration)

  The switch case in the main tool execution handler that registers and dispatches 'generate_music' calls to the specific handleGenerateMusic function.

  case "generate_music":
    return await this.handleGenerateMusic(args);