Name: 模型上下文協議（MCP） (The Model Context Protocol (MCP))
Uploaded: 2025-06-26T21:27:33.000Z
Duration: 19 min 35 s

過去簡單式

Around the time, in September, we had an internal hackathon.

大約在 9 月份，我們舉行了一次內部黑客馬拉松。

to 不定詞

Everyone was free to build basically whatever they wanted to build.

每個人基本上都可以自由地建造他們想建造的任何東西。

But it turns out everyone just built on MCP.

但事實證明，每個人都是在 MCP 的基礎上建立起來的。

Everyone's ideas were, oh, but what if we made this an MCP server?

每個人的想法都是，哦，但如果我們把它變成一個 MCP 服務器呢？

I lead Cloud Relations here at Anthropic.

我是 Anthropic 公司雲關係部的負責人。

Member of technical staff at Anthropic and one of the co-creators of MCP.

Anthropic 技術人員，MCP 的共同創建者之一。

Today we're going to be talking about the model context protocol and diving in deep into what it is and what's next.

今天，我們將討論模型上下文協議，並深入探討它是什麼以及下一步要做什麼。

But first, there's a lot of talk about MCP and not a lot of maybe real deep understanding of what it is.

但首先，關於 MCP 的討論很多，但真正深入瞭解它的人可能並不多。

Can we dive into how you view MCP and what it really means to be using MCP or building on it?

我們能否深入探討一下您是如何看待 MCP 的，以及使用 MCP 或在其基礎上進行建設的真正含義？

MCP is just a way for putting my workflow into an AI application in a very simple way.

MCP 只是一種將我的工作流程以非常簡單的方式放入人工智能應用程序的方法。

I think that's how I really wanted it to be initially, or that's how we wanted it to be.

我想這就是我最初真正想要的樣子，或者說這就是我們想要的樣子。

But it's just a way to give context to an application that uses another lab.

但這只是一種為使用另一個實驗室的應用程序提供上下文的方法。

And it can be tools, it can be just raw context, whatever you like it to be.

它可以是工具，也可以是原始背景，隨你喜歡。

How is that different than you calling an API or something like that?

這與你調用應用程序接口或類似的東西有什麼不同？

It's passing this information from one place into the prompt, basically, of the model.

基本上，它是將這些資訊從一個地方傳遞到模型的提示中。

I think the question is, what do models interact with?

我認為問題在於，模型與什麼互動？

And they don't interact directly with APIs, they interact with prompts and tools and whatever you're giving the model to ingest.

它們不會直接與應用程序接口交互，而是與提示和工具交互，不管你給模型提供了什麼，它們都會攝取。

And so MCP standardizes how you take that data from, whether it's an API or some internal data source or whatever it is, how you take that data and then actually give it to the model.

是以，MCP 規範瞭如何從 API 或內部數據源或任何其他來源獲取數據，以及如何獲取數據並將其提供給模型。

What are the main aspects of this protocol that has to follow?

必須遵守的協議主要包括哪些方面？

The main part is that it's a protocol between the AI application that uses another lab.

主要部分是使用另一個實驗室的人工智能應用程序之間的協議。

And it exposes basically three main things.

It's tools, it's a thing called resources, which is just raw data that you could ingest into a RAG pipeline or whatever you wanted to do, and there's prompts.

它是工具，是一種叫做資源的東西，它只是原始數據，你可以將其導入 RAG 管道或任何你想做的事情，而且還有提示。

And that's the three main things that a server can expose for now.

這就是服務器目前可以公開的三大內容。

So tools are actions that the model can take out in the world.

是以，工具是模型可以在世界上採取的行動。

Files, data, whatever kind of context you want to give the model.

文件、數據、任何你想賦予模型的上下文。

What a user wants to put into the context window by themselves and is triggered by the user and just put into the context window and then they can edit it as they want to.

用戶自己想放入上下文窗口的內容，由用戶觸發並直接放入上下文窗口，然後就可以隨意編輯了。

That's really what prompts are for, prompt templates in the end of the day.

說到底，這就是提示模板的作用。

We typically see that being implemented as a slash command.

If you're in the AI application of your choice, you would do a slash command and it would pull in the prompt template and save you time from having to write that out, whatever it is.

如果你正在使用自己選擇的人工智能應用程序，你只需執行一個斜線命令，它就會調出提示模板，省去你寫出來的時間。

There's definitely a lot of nuance in there.

這裡面肯定有很多細微的差別。

The most basic thing is that I worked on internal developer stuff and I got very quickly frustrated about having to copy things in and out of Cloud Desktop and then copying things back and forth between my IDE.

最基本的一點是，我負責內部開發人員的工作，我很快就對必須在 Cloud Desktop 中複製東西以及在我的 IDE 之間來回複製東西感到沮喪。

And that's just really what I was thinking about, how can I solve copying, pasting, the things I care about the most between these two applications.

這就是我一直在思考的問題，如何在這兩個應用程序之間解決複製、粘貼等我最關心的問題。

And that's really the absolute origin of where MCP started, at least in my mind.

至少在我看來，這才是 MCP 的真正起源。

And then from there, I explained that to Justin, who is the other co-creator, and he really took it and ran it and then we together just built it out and built it into Cloud Desktop.

然後，我向另一位聯合創始人賈斯汀解釋了這一點，他真的接受並運行了它，然後我們一起將它構建出來，並將它構建到雲桌面中。

And I think there was a pivotal moment that you alluded to.

我認為你提到了一個關鍵時刻。

We weren't really sure, is this going to work?

But around the time, in September, we had an internal hackathon and everyone was free to build basically whatever they wanted to build.

但就在 9 月份，我們舉行了一次內部黑客馬拉松，每個人都可以自由構建他們想構建的任何東西。

And we had everything from people doing very standard things like Slack integration or things you would think of when you think MCP, up to people who steered their 3D printer as MCP.

從做 Slack 整合等標準工作的人，到把 3D 打印機當作 MCP 使用的人，我們都有。

And I love this, when it got into the real world, when Cloud got into the real world because of an MCP server.

我喜歡這一點，當它進入現實世界時，當雲因 MCP 服務器而進入現實世界時。

I remember that too, when we were doing all these hackathon projects and there was no mandate to force people to use MCP.

我也記得，當時我們在做所有這些黑客馬拉松項目，並沒有強制要求人們使用 MCP。

Why did people gravitate towards MCP for all their projects?

為什麼人們的所有項目都傾向於 MCP？

I think it really was that standardization layer.

我認為這確實是標準化層的問題。

It just made it so much easier to add context to the application because the moment that Cloud is now integrated against MCP, that means as the server builder, you can build 1, 2, 10, 20, however many servers you want and you know that it will automatically work with that application.

因為雲計算現在已經與 MCP 集成，這意味著作為服務器構建者，您可以構建 1 臺、2 臺、10 臺、20 臺服務器，無論您想要多少臺服務器，您都知道它將自動與該應用程序配合使用。

I think that just gives you the ability to only think about one side and not have to think about the other side.

我認為，這隻會讓你只考慮一方，而不必考慮另一方。

I think there's a bit of a magic moment when you teach Cloud something new using an MCP server for the first time and you see it takes action about something you care about and I feel that's a little bit of a moment of magic that I think MCP captures really well, which makes people so excited because within five minutes they have something going.

我認為，當你第一次使用 MCP 服務器教雲學習新知識時，會有一個神奇的時刻，你會看到它對你關心的事情采取行動，我覺得這是一個神奇的時刻，我認為 MCP 很好地捕捉到了這一時刻，這讓人們非常興奮，因為在五分鐘之內，他們就能有所收穫。

I've seen it myself and even experienced it where it almost feels like you take Cloud out of the box, so to speak.

我親眼目睹過，甚至親身經歷過，感覺就像把雲從盒子裡拿出來一樣。

All of a sudden, instead of just being this thing that is just right there outputting text, it's doing other things.

突然之間，它不再是一個只在那裡輸出文本的東西，而是開始做其他事情了。

It's calling other applications, fetching other data, or even operating a 3D printer, which is a really crazy thing.

它可以調用其他應用程序，獲取其他數據，甚至操作 3D 打印機，這真是一件非常瘋狂的事情。

That does feel really special and I guess MCP allows that pretty seamlessly to some degree.

這種感覺真的很特別，我想 MCP 在某種程度上允許這種無縫連接。

This was back in the end of summer, early fall, as we were doing this Hack Week and these other things.

這要追溯到夏末秋初，當時我們正在舉辦黑客周和其他活動。

When did we launch MCP and what did that look like?

I think everyone's response, as you can imagine, well, some people still have this response, is what's MCP, right?

我想每個人的反應，正如你所能想象的，嗯，有些人還有這種反應，那就是什麼是 MCP，對嗎？

We definitely could have named it better.

我們絕對可以給它取一個更好的名字。

But you still get the MPC instead of MCP and then it makes me think NPC.

但你還是會得到 MPC 而不是 MCP，這讓我想到了 NPC。

You had a lot of people asking, what is MCP?

Not just externally, but I also think internally because it was such a bottoms-up movement.

不僅是對外，我認為對內也是如此，因為這是一場自下而上的運動。

Initially people were like, oh, what is this thing?

最初人們會問，哦，這是什麼東西？

What does it mean to give the model context?

Then as people started playing around with it and seeing it for themselves, I think that's where it actually slowly caught steam.

後來，人們開始玩它，親眼看到它，我想這就是它慢慢流行起來的原因。

The turning point was when more and more clients started adopting.

轉捩點是越來越多的客戶開始採用。

I think the IDEs were the first to adopt.

我認為集成開發環境最先採用了這一技術。

That's created a lot of waves in the market to incentivize a lot more server providers to actually build servers.

這在市場上掀起了軒然大波，激勵著更多的服務器提供商實際建造服務器。

You see so many times on social media, what is MCP?

你經常在社交媒體上看到，什麼是 MCP？

A month later, a few days later, they're going to be like, this is the best thing ever.

一個月後，幾天後，他們會說，這是最棒的事情。

It's now become, I think it's fair to say, industry standard of integration protocol.

可以說，它現已成為集成協議的行業標準。

There's nothing else in my mind that rivals it.

在我心中，沒有任何其他東西能與之匹敵。

Going back to the launch, a key decision here was to actually make this open source.

回過頭來看這次發佈，一個關鍵的決定就是將其開源。

That was pretty different in comparison to maybe previous efforts in this area that had been launched.

這與之前在這一領域所做的努力相比可能非常不同。

Can you explain the reasoning behind that decision and why did we open source it?

您能解釋一下這一決定背後的原因嗎？

If you have a closed ecosystem for integrations and for a context to be provided to AI applications, then it isn't clear to the server builders or the integration builders, is that AI application going to be around forever?

如果你有一個封閉的生態系統來為人工智能應用提供集成和上下文，那麼服務器構建者或集成構建者就不清楚，人工智能應用會永遠存在嗎？

字幕列表影片播放

模型上下文協議（MCP） (The Model Context Protocol (MCP))

context

awesome

basically

intelligence