AI ushers in the "USB-C moment", how can MCP perfectly integrate with Ethereum?

Content | Bruce

Editing & Typesetting | Ring

Design | Daisy

The "USB-C Moment" in the history of AI evolution, in November 2024, the MCP protocol released by Anthropic is triggering the Silicon Valley earthquake. This open standard, known as the "USB-C in the AI ​​world", not only reconstructs the connection between the big model and the physical world, but also hides the password to crack the AI ​​monopoly dilemma and reconstruct the digital civilization production relationship. While we are still arguing about the scale of GPT-5 parameters, MCP has quietly paved the path to decentralization to the AGI era...

Bruce: Recently, I am studying Model Context Protocol (MCP). This is the second thing that has really excited me in the AI ​​field after ChatGPT because it hopes to solve three issues I have been thinking about for years:

• Non-scientists and geniuses, how can ordinary people participate in the AI ​​industry and earn income?
• What is the win-win combination between AI and Ethereum?
• How to implement AI d/acc? Avoid centralized large-scale companies' monopoly and censorship, and AGI destroys humanity?

01. What is MCP?

MCP is an open standards framework that simplifies the integration of LLM with external data sources and tools. If we compare LLM to Windows operating systems, and applications such as Cursor are keyboards and hardware, then MCP is the USB interface that supports flexibly inserting external data and tools, and then users can read and use these external data and tools.

MCP provides three capabilities to extend LLM:

• Resources (knowledge extension)
• Tools (execute functions and call external systems)
• Prompts (pre-written prompt word template)

MCP can be developed and hosted by anyone, provided in the form of a Server, and can be offline and stopped at any time.

02. Why do I need MCP

Currently, LLM uses as much data as possible to perform a large amount of operations and generates a large number of parameters, integrating knowledge into the model, thereby achieving dialogue output corresponding knowledge. But there are several major problems:

1. Large amounts of data and operations require a lot of time and hardware, and the knowledge used for training is usually outdated.
2. Models with a large number of parameters are difficult to deploy and use on local devices, but in fact, users may not need all information to meet their needs in most scenarios.
3. Some models use crawlers to read external information for calculations to achieve timeliness, but due to crawler limitations and the quality of external data, more misleading content may be produced.
4. Since AI has not brought good benefits to creators, many websites and content have begun to implement anti-AI measures to generate a large amount of spam, which will lead to a gradual decline in the quality of LLM.
5. LLM is difficult to extend to all aspects of external functions and operations, such as accurately calling the GitHub interface to implement some operations, which generates code according to possible outdated documents, but cannot ensure that it can be executed accurately.

03. Architectural evolution of fat LLM and thin LLM + MCP

We can consider the current hyperscale model as a fat LLM, and its architecture can be represented by the following simple diagram:

After the user enters the information, the input is disassembled and reasoned through the Perception & Reasoning layer, and then a huge parameter is called for the result generation.

After MCP, LLM may focus on language parsing itself, stripping away knowledge and ability, and becoming thin LLM:

Under the architecture of thin LLM, the Perception & Reasoning layer will focus on how to parse all aspects of human physical environment information into tokens, including but not limited to: voice, tone, odor, images, text, gravity, temperature, etc., and then orchestrate and coordinate hundreds of MCP Servers through MCP Coordinator to complete the task. The training cost and speed of thin LLM will be increased rapidly, and the requirements for deploying equipment will become very low.

04. How to solve three major problems with MCP

How do ordinary people participate in the AI ​​industry?

Anyone with unique talents can create their own MCP Server to provide services to LLM. For example, a bird lover can provide services to external parties through MCP for many years of bird notes. When someone uses LLM to search for information related to birds, the current bird notes MCP service will be called. Creators will also receive income sharing as a result.

This is a more accurate and automated creator economic cycle, the service content is more standardized, and the number of calls and the output tokens can be accurately counted. LLM providers can even call multiple bird notes at the same time MCP Servers to allow users to select and score to determine who is better quality to get higher matching weights.

A win-win combination of AI and Ethereum

a. We can build an OpenMCP.Network creator motivation network based on Ethereum. MCP Server needs to host and provide stable services, and users pay LLM providers. LLM providers allocate actual incentives to the called MCP Servers through the network to maintain the sustainability and stability of the entire network, and inspire MCP creators to continue to create and provide high-quality content. This network will require the automation, transparency, credibility and censorship-resistant use of smart contracts. Signature, permission verification, and privacy protection during operation can be achieved using Ethereum wallet, ZK and other technologies.

b. Develop MCP Servers related to Ethereum chain operations, such as AA wallet calling service, and users will support wallet payments in LLM through languages ​​without exposing the relevant private keys and permissions to the LLM.

c. There are also various developer tools to further simplify Ethereum smart contract development and code generation.

Implement AI decentralization

a. MCP Servers decentralizes AI knowledge and capabilities. Anyone can create and host MCP Servers, register on platforms such as OpenMCP.Network and then obtain incentives according to calls. No company can master all MCP Servers. If an LLM provider gives unfair incentives to MCP Servers, the creator will support blocking the company, and users will replace other LLM providers to achieve fairer competition after failing to get high-quality results.

b. Creators can implement fine-grained permission control over their MCP Servers to protect privacy and copyright. Slim LLM providers should allow creators to contribute high-quality MCP Servers by providing reasonable incentives.

c. The gap in abilities of thin LLM will gradually be smoothed out, because human language has a traversal limit and its evolution is also slow. LLM providers will need to focus on high-quality MCP Servers instead of reusing more graphics cards to refine alchemy.

d. The capabilities of AGI will be distributed and downgraded. LLM is only used as language processing and user interaction. The specific capabilities are distributed in each MCP Servers. AGI will not threaten humans, because after closing MCP Servers, you can only have basic language conversations.

05. Overall review

1. The architectural evolution of LLM + MCP Servers is essentially decentralizing AI capabilities, reducing the risk of AGI destroying humanity.
2. The use of LLM enables the number of calls and input and output of MCP Servers token-level statistics and automation, laying the foundation for the construction of the AI ​​creator economic system.
3. A good economic system can drive creators to actively contribute to the creation of high-quality MCP Servers, thereby driving the development of the entire human race and achieving a positive flywheel. Creators no longer resist AI, and AI will also provide more positions and income, and will reasonably distribute the profits of monopoly commercial companies like OpenAI.
4. This economic system, combined with its characteristics and the needs of creators, is ideal for implementation based on Ethereum.

06. Future Outlook: Next Steps of Script Evolution

1. MCP or MCP-like protocols will emerge one after another, and several large companies will begin to compete for the definition of standards.
2. MCP Based LLM will appear, a small model that focuses on parsing and processing human languages, with MCP Coordinator to access the MCP network. LLM will support automatic discovery and scheduling of MCP Servers without complex manual configuration.
3. MCP Network service providers will emerge, each with its own financial incentive system, and MCP creators will register and host their own server to earn revenue.
4. If the MCP Network's economic incentive system is built using Ethereum and is based on smart contracts, the transactions of the Ethereum network are conservatively estimated to increase by about 150 times (according to a very conservative number of calls to MCP Servers per day, currently a block of 12s includes 100 txs).