Agentic Web: Weaving the Next Web with AI Agents

TL;DR

<3-5 sentence high-level summary>The paper introduces the Agentic Web, a paradigm in which autonomous AI agents powered by LLMs execute user-delegated tasks across a network of web services, transforming the internet into an agent-driven action space. It presents a three-dimensional framework—Intelligence, Interaction, and Economy—and discusses core conditions, architectural transformations, algorithmic transitions, and system-level roadmaps needed to realize this vision. Key contributions include the formalization of agent-centric information retrieval, planning, and multi-agent coordination; the MCP and A2A communication protocols as foundational infrastructure; and a detailed examination of safety, governance, and open problems. The work underscores potential applications, societal risks, and outlines research directions toward open, secure, and scalable agentic ecosystems that blend human intent with autonomous agent behavior.

Abstract

The emergence of AI agents powered by large language models (LLMs) marks a pivotal shift toward the Agentic Web, a new phase of the internet defined by autonomous, goal-driven interactions. In this paradigm, agents interact directly with one another to plan, coordinate, and execute complex tasks on behalf of users. This transition from human-driven to machine-to-machine interaction allows intent to be delegated, relieving users from routine digital operations and enabling a more interactive, automated web experience. In this paper, we present a structured framework for understanding and building the Agentic Web. We trace its evolution from the PC and Mobile Web eras and identify the core technological foundations that support this shift. Central to our framework is a conceptual model consisting of three key dimensions: intelligence, interaction, and economics. These dimensions collectively enable the capabilities of AI agents, such as retrieval, recommendation, planning, and collaboration. We analyze the architectural and infrastructural challenges involved in creating scalable agentic systems, including communication protocols, orchestration strategies, and emerging paradigms such as the Agent Attention Economy. We conclude by discussing the potential applications, societal risks, and governance issues posed by agentic systems, and outline research directions for developing open, secure, and intelligent ecosystems shaped by both human intent and autonomous agent behavior. A continuously updated collection of relevant studies for agentic web is available at: https://github.com/SafeRL-Lab/agentic-web.

Figures (13)

• Figure 1: Illustration of the Agentic Web process cycle. The cycle begins with a user submitting a task request. The system then plans the task and identifies appropriate agents and tools. Recruited agents engage in inter-agent discussions, collaborate using their unique capabilities and resources, and execute the task. The results are reported back to the user, completing the cycle. The Agentic Web facilitates discovery, coordination, and cooperation among agents to fulfill user goals.
• Figure 2: Evolution of user-system interaction across three internet eras. In the PC Web Era, users acted primarily as content consumers with limited interaction. The Mobile Web Era introduced a bidirectional flow, enabling users to both consume and produce content. In the emerging Agentic Web Era, tasks are delegated to ai agents, who interact with information networks on their behalf. The expanding and darkening circles reflect the increasing complexity and volume of information.
• Figure 3: Timeline of Web Evolution. The three eras of web evolution are not strictly distinct. Their transitions happened gradually, with technologies, features, and business models often overlapping and coexisting across different periods.
• Figure 4: Attention Flow Evolution Across Web Eras. This diagram illustrates the transition from the PC Web, where attention follows a linear search-query-ad model, to the Mobile Web, where algorithmic systems curate feeds based on user data, and finally to the Agentic Web, where autonomous agents interpret user intent and select among competing services to execute tasks. Dashed arrows in the agentic stage indicate competitive or compositional relationships between services.
• Figure 5: Agent Workflow under the Agentic Web.