MCP Academy · 课时

在读取时构建内容

在资源被请求时计算其正文。

在读取时构建内容 是 CoddyKit 上的免费 MCP Academy 课时。这是第 2 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。

Lazy, Not Eager

A dynamic resource doesn't store a fixed body. Its content is computed on demand, only when a client actually asks to read it. ⚡

The Read Triggers the Work

Your resource function runs at the moment of a read request. Nothing happens until a client reads it, so you never waste effort building content no one wants.

Return the Body

Whatever your function returns becomes the resource's content. Return a string and the client receives that exact text as the resource body.

@mcp.resource("time://now")
def now() -> str:
    return datetime.now().isoformat()

Fresh Every Read

Because the body is rebuilt each time, the model always sees current data. Read the clock twice and you get two different timestamps.

Use the URI Variables

Combine read-time building with a template. The captured parameters shape what you compute, so the same function serves tailored content per URI.

@mcp.resource("greeting://{name}")
def greet(name: str) -> str:
    return f"Hello, {name}!"

Pull From Anywhere

At read time you can do real work: query a database, call an API, or read a file. The result of that work is what you return as the body.

@mcp.resource("stats://summary")
def summary() -> str:
    return db.fetch_summary()

Async When You Wait

If building content means waiting on I/O, make the function async. The server can serve other reads while yours awaits a slow network call.

@mcp.resource("feed://latest")
async def latest() -> str:
    return await fetch_feed()

Shape the Returned Text

You decide the format of the body. Returning JSON text is common when the model needs structured data it can reason over cleanly.

import json

def report() -> str:
    return json.dumps({"ok": True})

Keep It Quick

A read should feel snappy. If the work is heavy, cache the result so repeated reads don't redo expensive computation every single time.

Handle Missing Data

What if the requested thing doesn't exist? Raise an error from the function so the client learns the resource couldn't be read, rather than returning junk.

def get_doc(doc_id: str) -> str:
    if doc_id not in store:
        raise ValueError("not found")
    return store[doc_id]

Why This Matters

Read-time building turns a static idea into a live window on your data. The model always reads what is true right now, not a stale snapshot. 🎯

Quick Check

When does the body of a dynamic resource get produced?

Recap

You saw how returning a value from a resource function builds content on read, giving the model fresh, computed data every time. Next: listed vs templated resources. 🚀

常见问题解答

「在读取时构建内容」课时是免费的吗？

是的 — 「在读取时构建内容」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。

「在读取时构建内容」这节课中我会学到什么？

在资源被请求时计算其正文。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 2 节课，共 4 节。

「在读取时构建内容」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