Qwen MCP Tool

Instructions

Implementation Reference

• src/tools/ask-qwen.tool.ts:66-87 (handler)

  The main execution handler for the 'ask-qwen' tool. Destructures input arguments, validates the prompt, and invokes the Qwen CLI executor with the parameters.

  execute: async (args, onProgress) => {
    const { prompt, model, sandbox, approvalMode, yolo, allFiles, debug } = args;
  
    // Validate prompt
    if (!prompt || !prompt.trim()) {
      throw new Error(ERROR_MESSAGES.NO_PROMPT_PROVIDED);
    }
  
    // Execute Qwen CLI
    const result = await executeQwenCLI({
      prompt,
      model,
      sandbox,
      approvalMode,
      yolo,
      allFiles,
      debug,
      onProgress
    });
  
    return result;
  },

• src/tools/ask-qwen.tool.ts:13-65 (schema)

  Zod schema for input validation of the 'ask-qwen' tool parameters including prompt, model selection, sandbox mode, approval settings, and debug options.

  zodSchema: z.object({
    prompt: z
      .string()
      .min(1)
      .describe(
        "The query or instruction for Qwen. Use @filename or @directory to include file contents. Example: '@src/ Explain this codebase structure'"
      ),
    model: z
      .enum([
        QWEN_MODELS.PRIMARY,
        QWEN_MODELS.FALLBACK,
        QWEN_MODELS.PLUS,
        QWEN_MODELS.TURBO,
        QWEN_MODELS.PRO
      ])
      .optional()
      .describe(
        `Optional model to use (e.g., '${QWEN_MODELS.PRIMARY}'). If not specified, uses the default model (${QWEN_MODELS.PRIMARY}).`
      ),
    sandbox: z
      .boolean()
      .default(false)
      .describe(
        "Use sandbox mode (-s flag) to safely test code changes, execute scripts, or run potentially risky operations in an isolated environment"
      ),
    approvalMode: z
      .enum([
        APPROVAL_MODES.PLAN,
        APPROVAL_MODES.DEFAULT,
        APPROVAL_MODES.AUTO_EDIT,
        APPROVAL_MODES.YOLO
      ])
      .optional()
      .describe(
        "Control tool execution approval: 'plan' (analyze only), 'default' (prompt for approval), 'auto-edit' (auto-approve edits), 'yolo' (auto-approve all)"
      ),
    yolo: z
      .boolean()
      .default(false)
      .describe(
        "Enable YOLO mode to automatically approve all tool calls without prompting (equivalent to approvalMode='yolo')"
      ),
    allFiles: z
      .boolean()
      .default(false)
      .describe(
        "Include all files in the current directory as context (use with caution for large directories)"
      ),
    debug: z
      .boolean()
      .default(false)
      .describe("Enable debug mode for more verbose output")
  }),

• src/tools/ask-qwen.tool.ts:88-116 (schema)

  Prompt schema defining the structure and descriptions for the 'ask-qwen' tool arguments, likely for MCP protocol integration.

    prompt: {
      name: "ask-qwen",
      description:
        "Interact with Qwen AI for code analysis, file exploration, and general queries. Supports @file references for including file contents.",
      arguments: [
        {
          name: "prompt",
          description:
            "Your question or instruction. Use @filename or @directory to reference files.",
          required: true
        },
        {
          name: "model",
          description: `Optional model selection (${QWEN_MODELS.PRIMARY}, ${QWEN_MODELS.TURBO}, etc.)`,
          required: false
        },
        {
          name: "sandbox",
          description: "Enable sandbox mode for safe code execution",
          required: false
        },
        {
          name: "approvalMode",
          description: "Control approval for tool execution (plan/default/auto-edit/yolo)",
          required: false
        }
      ]
    }
  };

• src/tools/index.ts:9-11 (registration)

  Registers the askQwenTool ("ask-qwen") along with other tools in the tool registry.

  registerTool(askQwenTool);
  registerTool(pingTool);
  registerTool(helpTool);