A New Chapter in AI Agent Collaboration: Exploring the Potential of Agent Business Protocols

Recently, the performance improvements of AI models have become steady, and the industry's focus has shifted from technological development to practical applications. AI agents, as an emerging technology, have attracted widespread attention, but the limitations of individual agents make specialized collaboration crucial. However, there is currently a lack of standardized systems for agent collaboration in the market.

A solution called Agent Commerce Protocol (ACP) has emerged. ACP standardizes and automates collaboration between agents through four stages: "request-negotiation-transaction-evaluation." This enables agents from different platforms to achieve seamless integration and cooperation.

With the help of ACP, agents can operate as independent economic entities around the clock. For example, on-chain hedge funds and autonomous media production have showcased the potential of this technology. Currently, about 1 million agents generate $1 billion in value annually, and this figure is expected to reach $1 trillion by 2035.

AI agents are becoming the new frontier in the application of AI technology. Unlike traditional passive tools, agents are proactive systems that can fully understand tasks and make autonomous decisions. For example, when booking a restaurant, the agent can not only provide suggestions but also take into account user preferences, search for popular restaurants, and make real-time reservations.

However, a single agent finds it difficult to become an all-round expert. Each field requires specific expertise, and there are limitations in memory and computing power. Therefore, collaboration between different agents becomes particularly important. However, achieving this collaboration is not easy, as it requires addressing a series of issues such as scope of work, quality standards, pricing, and delivery evaluation.

The four main stages of ACP—request, negotiation, transaction, and evaluation—are similar to the traditional process of companies issuing requests for proposals, but the entire process is automated through smart contracts. Taking the creation of a poster as an example, the managing agent can publish a request, the designer agent provides a quote, both parties begin work after reaching an agreement, and finally, the evaluation agent reviews the results and automatically settles the payment.

The application of ACP will bring about a fundamental transformation of the agency ecosystem. Agents can work 24/7 without physical constraints and time limitations, paving the way for new business models. For example, on-chain hedge funds can operate around the clock, with agents from different specialties collaboratively analyzing the market, managing risks, and optimizing portfolios.

Another example is a media production factory driven by agents. Professional agents can handle the entire process from planning to production to distribution, achieving real-time interaction between virtual idols and fans, greatly increasing engagement.

The agent economy has gradually transitioned from concept to reality. The rapid development of technological foundations, such as the significant reduction in AI inference costs and the emergence of high-performance open-source models, has created favorable conditions for the widespread application of agents. It is expected that by 2025, approximately 1 million public agents will be operating on-chain, generating a total value of about $1 billion annually.

Despite the bright prospects, there are still some challenges that need to be overcome. For example, further improvements are needed in terms of privacy protection, especially regarding sensitive transaction information and business logic. With the development of technologies such as zero-knowledge proofs, these limitations are expected to be gradually resolved, thereby further unleashing the potential of the agency economy.