What exactly is MCP, which is popular in the AI ​​programming circle?

‍‍

‍‍If you often use AI programming recently, you must have heard of the concept of MCP?

So what exactly is MCP?

I tried to explain it to you today.

First of all, from a professional perspective, MCP is an open, universal, and consensus-based protocol standard released by Anthropic (Claude).

Model Context Protocol (MCP)

• MCP is a standard protocol that is like a "universal interface" for large AI models, allowing AI models to interact seamlessly with different data sources and tools. Just like the USB-C interface, it provides a standardized method to connect AI models to various data sources and tools.

• MCP aims to replace fragmented Agent code integration to make AI systems more reliable and more efficient. By establishing common standards, service providers can launch their own AI capabilities based on the protocol, thereby supporting developers to build more powerful AI applications faster. Developers do not need to reinvent the wheel and can build a powerful AI Agent ecosystem through open source projects.

• MCP can maintain context between different applications/services, thereby enhancing the overall ability to perform tasks autonomously.

MCP follows a client-server architecture and consists of the following core components:

• MCP Hosts : AI applications that initiate requests, such as chatbots, AI-driven IDEs, etc.

• MCP Clients : Maintain a 1:1 connection with the MCP server inside the host program.

• MCP Servers : Provide context, tools, and prompts to MCP clients.

• Local Resources : Resources on the local computer that can be securely accessed by the MCP server, such as files and databases.

• Remote Resources : Remote resources that an MCP server can connect to, such as data provided through an API.

The workflow of MCP can be simply summarized into the following steps:

1. Connection : The MCP host connects to one or more MCP servers.
2. Request : The host sends a request to obtain data or execute a tool.
3. Processing : The server processes the request and accesses relevant data sources or external services.
4. Return : The server returns the result to the host.
5. Generate Response : The host provides information to the AI ​​model to generate a user response.

Why do we need MCP?

For example, we cannot currently use an AI application to search the Internet, send emails, publish our own blogs, etc. These functions are not difficult to implement individually, but if they are all integrated into one system, it will become out of reach.

If you don’t have a specific feeling yet, we can think about it in daily development. Imagine that in the IDE, we can use the IDE’s AI to complete the following tasks.

• Ask AI to query existing data in the local database to assist development
• Ask AI to search Github Issues to determine whether a problem is a known bug
• Code Review by sending comments on a PR to colleagues via instant messaging software (such as Slack) through AI
• Use AI to query or even modify the current AWS and Azure configurations to complete deployment

What about MCP? If other services follow the MCP standard, it will be like a universal interface, making our development more efficient.

Let’s take a simple example

Suppose you are using an AI programming assistant to help you write code. This AI assistant is an MCP host. It needs to access some external resources, such as code libraries, documents, or debugging tools. The MCP server is like an intermediary that connects these resources and the AI ​​assistant.

• When you need to find the usage of a function, the AI ​​assistant sends a request to the MCP server through the MCP client.
• After receiving the request, the MCP server searches for relevant information in the code library or document.
• After finding the information, the MCP server returns the results to the AI ​​assistant.
• Based on the information returned, the AI ​​assistant generates a code or explanation and shows it to you.

Advantages of MCP

• Standardization : MCP provides a unified communication protocol, making it easier to integrate and expand different AI applications.
• Flexibility : With MCP, AI applications can connect to a variety of data sources and tools, increasing their functionality and applicability.
• Security : MCP ensures the security of data transmission and protects user privacy and data security.

Let’s take a real-life example: suppose you are a class monitor and have to deal with a lot of class affairs every day: checking the class report card (Excel file stored in the computer), collecting feedback from classmates (chat records in the WeChat group), and arranging the duty roster (online document).

Traditional method: You need to open Excel, look through WeChat records, edit online documents, and organize information manually, which is time-consuming and laborious.

After using MCP, you can directly tell AI: "Help me check the average score of the recent math test, put the names of students who failed into the duty list, and remind them to take the make-up exam in the WeChat group." AI will automatically complete the following: Use the "universal plug" MCP to connect your computer and read the Excel scores. Use MCP to connect WeChat and find relevant chat records. Use MCP to modify online documents and update the duty list. The whole process does not require you to operate manually, and the data will not leave your device. It is safe and efficient.

Therefore, the power of MCP lies in that it does not need to reinvent the wheel. In the past, each software (such as WeChat and Excel) had to make a separate interface for AI. Now MCP has unified the standard, just like all electrical appliances use USB-C charging ports, AI can connect all tools with one interface.

Moreover, the data does not need to be uploaded to the cloud, AI processes it locally. For example, your transcript is only stored on your computer, and AI reads and analyzes it through MCP, but the data will not be leaked.

MCP will make AI better "understand" the context. For example, if you ask AI to "summarize the key points of last week's class meeting", it can automatically retrieve meeting recordings, chat records, and note documents, and combine this information to give you the answer instead of making it up out of thin air.

Therefore, MCP provides a powerful tool for AI applications to interact with the outside world more flexibly and securely.

‍