Anthropic’s innovative model context protocol (MCP) aims to solve scattered data and improve the efficiency of the AI driver solution. Can it become the standard for perception of AI integration in context?
One of the most urgent challenges in artificial intelligence (AI) innovation today is the isolation of large language models (LLMS) and real -time data. In order to solve this problem, the AI research and security company of San Francisco and the humanity of San Francisco recently announced a unique development architecture to reshape how the AI model interacts with data.
Company’s new Model context protocol (MCP)As a open source project, it aims to improve the efficiency of AI through two -way communication between the two -way communication between the application through the AI -driven application and the real -time diverse data source. ”
The construction of this architecture is to solve the sense of frustration of increasing growth: an outdated AI output caused by the lack of connection with real -time data. Humans claim that the unified agreement can enhance the AI development and functions of the enterprise, and make them more humane through real -time environmental consciousness. According to the company, each new business data source needs to customize AI implementation to create low efficiency. MCP tries to solve this problem by providing a standardized framework that developers can generally adopt.
“The structure of this architecture is very simple: developers can reveal their data through the MCP server, or they can also build AI applications (MCP clients) connected to these servers. Developers can now build each data source according to the standard protocol instead of instead of instead of Maintain a separate connector for each data source. ” Blog postEssence “As the ecosystem matures, the AI system will maintain the background when moving between different tools and data sets, so as to replace today’s scattered integration with a more sustainable architecture.”
The AI model, including but not limited to the flagship assistant Claude of Anthropic, can integrate tools such as Google Drive, Slack and Github. Experts suggest that MCP has a service -oriented architecture (SOA) and other protocols that completely change the application of application interoperability to change business AI integration.
“Industry standard protocols with data channels between LLM and data sources are a person who changes the rules of the game. Similar to rest and SQL in the software industry, Standardized agreements such as MCP can help the team build the Genai application faster and have better reliability, ” Gideon Mendels, co -founder and CEO of the AI model evaluation platform, told me. “This is the market in the past six months that the excellent LLM model is not enough. “
Anthropic also revealed that early companies including Block and Apollo have integrated MCP into their systems. At the same time, development tool providers (such as ZED, Replit, Codlegraph and Sourcegraph) are cooperating with MCP to enhance their platforms. The partnership is designed to help AI models and agents search more relevant information through real -time data retrieval, more effectively grasp the context, and generate subtle output to improve efficiency as the subtle effects of corporate tasks such as coding.
“More like humanization and self -awareness AI models can make technology correlation, which may promote widespread adoption,” “Masha Levin, an entrepreneur living in one way. “AI still has a lot of fear. Many people regard it as a machine. Humanizing these models can help relieve these fear and promote the integration of daily life.”
Lavin also warned potential shortcomings. “There is risk, and companies may rely too much on AI to support, so that they can affect their decisions in extreme ways, which may lead to harmful consequences.”
However, the real test of MCP will be the ability to obtain extensive adoption in the crowded market and exceed competitors.
Anthropomorphic MCP and Openai and confusion: AI Innovation Standard Battle
Although the open source method of human MCP marks the significant improvement of AI innovation, it has entered a competitive landscape led by technological giants such as OpenAI and ClLEXITY.
OpenAI recently used the “work with application” function that shows similar features in ChatGPT, although the exclusive focus is to give priority to the close partnership of the public standard. This function allows ChatGPT to access and analyze the data and content of other applications, but only when the user permits, the needs of developers will be manually copied and pasted manually. Instead, ChatGPT can directly view the data from the application, and provides more intelligent context sensing suggestions because of the integration of real -time Internet data.
In addition, the company also launched the data system structure in October, called “real -time API”, which allows voice assistants to respond more effectively by extracting a new environment from the Internet. For example, voice assistants can represent users to order or retrieve related customer information to provide personalized responses. “Now, use real -time APIs, and use audio quickly in the API. Developers no longer need to suture multiple models together to provide motivation for these experiences.” Blog postEssence “Under the hood, real-time API allows you to create a long-lasting WebSocket connection to exchange messages with GPT-4O.”
Similarly, “AI Real -time Data Agreement” (referred to as “”PPlx-API“It provides developers with access to its large language model (LLM). This API allows applications to send natural language query and receive detailed real -time information from the network. Through a single API endpoint, it can provide the latest AI applications with the latest. The data retrieval and the response of context perception enable developers to build applications that remain consistent with the latest information.
“Generally, the industry tends to standardize an open source solution, but it usually takes years. OpenAI is likely to try more agreements.” MenDels said. “However, if MCP is widely used as the first standard of similar standards, we can see that technology and best practice start to standardize it.”
Can human MCP set the standard for perception of AI integration?
Despite potential, human MCP still faces major challenges. Security is the main problem, because enabling AI systems to access sensitive corporate data will increase the risk of system hooligan leakage. In addition, persuasive developers may be difficult to adopt MCP in the established ecosystem.
According to JD Raimondi, the person in charge of the data science of IT Development Corporation, another problem is the huge scale of the data. He told me: “anthropomorphicization is the leader of experiments that lead to a huge background, but the accuracy of the model is very large. Over time, they are likely to become better, and in terms of performance, there are many techniques to maintain Acceptance. “
Although the anthropomorphic assertion that MCP has improved the ability of AI to retrieve and the cultural data of up and down, the lack of specific benchmarks that support these claims may hinder the adoption of these claims. Anthropic said: “Whether you are an AI tool developer, we hope to use the existing data companies or explore the frontier of the frontier, we invite you to build the future of context consciousness AI together.”
With the function of developers to test MCP, the industry will pay attention to checking whether this open standard can be obtained as the traction of the benchmark for context perception of AI integration. Mendels suggested that standardization may be an anthropomorphic and wise movement, and allows the team to try to use different tool combinations to determine the most suitable method for their needs. Mendes pointed out: “At present, many processes in the AI ecosystem are too early to be standardized.” “As innovation occurs so rapidly, today’s best practice may be outdated next week. Only time can it be available. Determine whether a protocol like MCP can successfully standardize the context data retrieval. “
