AI Agent's Exploration in the Web3 Domain: From Manus to MCP

Recently, a product named Manus, the world's first universal AI Agent, has attracted widespread attention. As an AI system capable of independent thinking, planning, and executing complex tasks, Manus demonstrates unprecedented versatility and execution ability. This has not only sparked discussions within the industry but also provided valuable product ideas and design inspiration for various AI Agent developments.

With the rapid development of AI technology, AI Agents, as an important branch of artificial intelligence, are gradually transitioning from concept to reality, demonstrating immense application potential across various industries, and the Web3 sector is no exception.

Basic Concept of AI Agent

An AI Agent is a computer program that can make decisions and execute tasks autonomously based on the environment, inputs, and predefined goals. Its core components include:

1. Large Language Models (LLM) as the "brain"
2. Observation and Perception Mechanism
3. Reasoning Process
4. Action Execution Ability
5. Memory and Retrieval Functions

The design patterns of AI Agents mainly have two development routes: one emphasizes planning ability, while the other emphasizes reflective ability. Among them, the ReAct model is the earliest and most widely used design pattern. ReAct solves diverse language reasoning and decision-making tasks by combining reasoning and acting in language models. Its typical process can be described as a cycle of "Think → Act → Observe."

According to the number of agents, AI Agents can be divided into Single Agent and Multi Agent. The core of Single Agent lies in the collaboration between LLM and tools, while Multi Agent assigns different roles to different Agents, completing complex tasks through collaborative cooperation.

Introduction to MCP Protocol

Model Context Protocol (MCP) is an open-source protocol launched by Anthropic, aimed at addressing the connectivity and interaction issues between LLM and external data sources. MCP provides three capabilities to extend LLM: Resources (knowledge expansion), Tools (executing functions, calling external systems), and Prompts (pre-written prompt templates).

The MCP protocol adopts a Client-Server architecture, with JSON-RPC protocol used for underlying transmission. Anyone can develop and host an MCP Server, and can take the service offline at any time.

The Current Status of AI Agents in Web3

In the Web3 industry, the popularity of AI Agents peaked in January this year and then significantly declined, with an overall market value shrinking by over 90%. Currently, projects that still have a voice mainly revolve around the AI Agent framework for Web3 exploration, which mainly includes three models:

1. Launch Platform Mode: represented by Virtuals Protocol
2. DAO Model: Represented by ElizaOS
3. Business Company Model: Represented by Swarms

From the perspective of economic models, currently only the launch platform model can achieve a self-sustaining economic closed loop. However, this model also faces challenges, mainly because the issued AI Agent assets need to have sufficient "attractiveness" to generate a positive flywheel.

The Exploration Direction of MCP in the Web3 Field

The emergence of MCP has brought new exploration directions for Web3 AI Agents, mainly including:

1. Deploy the MCP Server to the blockchain network to solve the single point issue and have censorship-resistant capabilities.
2. Empower the MCP Server to interact with the blockchain, such as conducting DeFi transactions and management, reducing the technical barriers.

In addition, there is a scheme based on Ethereum to build the OpenMCP.Network creator incentive network. This network aims to achieve automation, transparency, trustworthiness, and censorship resistance of incentives through smart contracts, while utilizing technologies such as Ethereum wallets and ZK for signature, permission verification, and privacy protection during the operation process.

Although the theoretical combination of MCP and Web3 can inject decentralized trust mechanisms and economic incentives into AI Agent applications, there are still some limitations in current technology, such as the difficulty of verifying the authenticity of Agent behavior with zero-knowledge proof (ZKP) technology, and the efficiency issues of decentralized networks.

Conclusion

The release of Manus marks an important milestone for universal AI Agent products. The Web3 world also needs a milestone product to break the skepticism about the practicality of Web3. The emergence of MCP brings new exploration directions for Web3's AI Agent. Although there are still many challenges ahead, the integration of AI and Web3 is an inevitable trend. We need to maintain patience and confidence while continuously exploring the possibilities in this field.