MCP Academy · 课时

测试、检查并强化

使用 Inspector 验证并收紧访问权限。

测试、检查并强化 是 CoddyKit 上的免费 MCP Academy 课时。这是第 3 节课，共 4 节。你可以在下方免费阅读本课时的完整内容 — 然后在浏览器中使用内置代码编辑器和全天候 AI 导师进行实践。这是 MCP Academy 学习路径的一部分，你的进度在网页和 CoddyKit 应用中同步。 MCP Academy 课程共包含 4 节课。

本课时的部分内容尚未翻译，以英文显示。

From Working to Trustworthy

Your server runs, but is it ready for real use? Now you test, inspect, and harden it so it behaves under pressure and bad input. 🛡️

Smoke Test First

Start with a quick smoke test: does the server boot and list its tools without errors? Catching a broken start early saves hours of confusion.

Open the Inspector

Launch the MCP Inspector with the dev command. It gives you a visual panel to click through every tool, resource, and prompt by hand.

# In your project folder:
# uv run mcp dev server.py

Click Each Tool

In the Inspector, call each tool with sample inputs. Seeing the live result confirms the wiring works before any AI client ever connects.

Read a Resource

Try fetching a resource by its uri in the Inspector. If the right note comes back, your path variable and read logic are doing their job.

Test the Bad Path

Feed each tool obviously bad input: empty strings, huge numbers, missing fields. A solid server answers with a clear error, never a crash.

Return Errors, Not Stack Traces

Wrap risky work and raise a clean message. The model can read a friendly error and recover, but a raw traceback only confuses it.

@mcp.tool()
def get_note(note_id: int) -> str:
    if note_id < 1:
        raise ValueError("note_id must be a positive integer")
    return read(note_id)

Validate Before Acting

Check arguments at the top of each tool. Rejecting bad values early keeps your backend safe and your error messages specific and useful.

Apply Least Privilege

Give each tool only the access it truly needs. Least privilege means a search tool can read but never delete, shrinking the blast radius.

Guard Destructive Actions

For anything irreversible, demand a clear flag or confirmation. A guard on delete stops a hasty model call from wiping real data. ⚠️

@mcp.tool()
def delete_note(note_id: int, confirm: bool = False) -> str:
    if not confirm:
        return "Set confirm=True to delete."
    remove(note_id)
    return "Deleted."

Add Automated Tests

Write a few pytest cases that call your functions directly. Automated checks let you change code later and instantly know nothing broke.

Quick Check

Let us confirm the hardening idea.

Recap

You tested with the Inspector, validated inputs, applied least privilege, and guarded destructive actions. Your server is now ready for the real world. 🎯

常见问题解答

「测试、检查并强化」课时是免费的吗？

是的 — 「测试、检查并强化」的完整文本可在网页上免费阅读。要进行交互式练习（内置代码编辑器和全天候 AI 导师）并解锁 MCP Academy 课程的其余内容，请升级到 CoddyKit PRO。 MCP Academy 课程共包含 4 节课。

「测试、检查并强化」这节课中我会学到什么？

使用 Inspector 验证并收紧访问权限。你通过在浏览器中直接运行的动手代码来练习 MCP Academy，全天候 AI 导师会在你学习这节课的过程中回答你的问题。

学习 MCP Academy 需要有经验吗？

无需任何先前经验。CoddyKit 上的 MCP Academy 课程适合初学者到高级学习者，你可以从这里开始或从头开始，按照自己的节奏学习。这是第 3 节课，共 4 节。

「测试、检查并强化」课时需要多长时间？

大多数 CoddyKit 课程大约需要 5–10 分钟。每节课都很精短且互动，所以你能稳步进步，并在网页和应用中从离开的地方继续。

我能在这节 MCP Academy 课中编写并运行代码吗？

能。每节 MCP Academy 课都包含内置代码编辑器，你可以在浏览器中直接编写并运行真实代码，并获得即时 AI 反馈 — 无需本地设置。

此课程中的所有课时

← 返回 MCP Academy