How to Debug and Test MCP Services with AutoDev MCP Debugger

Key Features of AutoDev MCP Debugger

Test MCP Service : Verify that the MCP service is operational.

Debug MCP Tool : Generate mock data or manually provide input to debug a tool.

Execute MCP Tool : Run a tool directly and view its result.

Test Model‑Driven Tool Calls : Send a requirement to a model and observe which tool the model selects.

Documentation: https://ide.unitmesh.cc/mcp/mcp-debugger

Quick Start

Install AutoDev version 2.0.8 (available on GitHub). Create a file with the .mcp.json extension, for example filesystem.mcp.json, and add the following configuration:

{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "/Volumes/source/ai/auto-dev"]
    }
  }
}

Click the Preview button to check whether the MCP service is running.

Testing MCP Tool Calls

You can test a tool in two ways:

Mock data generation: In the list view, click the tool’s Test button. AutoDev generates mock data and sends it to the tool.

Manual input: In the list view, click the tool’s Details button, paste JSON data, and AutoDev sends it to the tool.

Testing MCP Tool Integration

To test multiple tools or refine a tool description, use the input box at the bottom of the debugger UI:

Select the target model and configure its parameters.

Enter the natural‑language requirement.

Press Send and wait for the model to return a response. The debugger parses the response, extracts the invoked tools, and allows you to execute them individually or collectively while showing detailed timing information.

Additional Notes

The integration prompts are based on modifications of Anthropic’s prompt templates. Domestic models perform well in front‑end scenarios, but other models can also be used for testing.