Position: Collaborative Agentic AI Needs Interoperability Across Ecosystems

Rishi Sharma, Martijn de Vos, Pradyumna Chari, Ramesh Raskar, Anne-Marie Kermarrec

TL;DR: This paper argues that the rapidly developing field of collaborative AI agents urgently needs minimal interoperability standards to prevent it from fragmenting into incompatible, isolated ecosystems.

Abstract: Collaborative agentic AI is projected to transform entire industries by enabling AI-powered agents to autonomously perceive, plan, and act within digital environments. Yet, current solutions in this field are all built in isolation, and we are rapidly heading toward a landscape of fragmented, incompatible ecosystems. In this position paper, we argue that interoperability, achieved by the adoption of minimal standards, is essential to ensure open, secure, web-scale, and widely-adopted agentic ecosystems. To this end, we devise a minimal architectural foundation for collaborative agentic AI, named Web of Agents, which is composed of four components: agent-to-agent messaging, interaction interoperability, state management, and agent discovery. Web of Agents adopts existing standards and reuses existing infrastructure where possible. With Web of Agents, we take a first but critical step toward interoperable agentic systems and offer a pragmatic path forward before ecosystem fragmentation becomes the norm.

Lay Summary:

AI agents, software programs powered by large language models that can browse the web, write code, book travels, and more, are increasingly being designed to work together as teams to tackle complex tasks. But right now, every major company building these agents (Google, Anthropic, IBM, and others) is doing so using their own incompatible systems. An agent built for one platform cannot talk to an agent built for another, much like how WhatsApp users cannot message Telegram users.

We argue this fragmentation is a serious problem, and that waiting for one dominant system to "win" risks locking the world into closed, monopolistic ecosystems, as happened with messaging apps. Instead, *we advocate for adopting a small set of minimal, shared standards early, before the silos harden*.

We introduce Web-of-Agents, a blueprint built on technologies the internet already uses: web addresses, HTTP requests, cookies, and simple documentation pages, to let agents from different ecosystems discover each other, communicate, and collaborate. We implement this across three existing agent frameworks with minimal code, adding negligible performance overhead. Thus, the kind of interoperability provided by our blueprint could keep the emerging world of collaborative AI open, competitive, and secure for everyone.

Link To Code: https://github.com/sacs-epfl/web-of-agents

Primary Area: Other topic (use sparingly and specify relevant keywords)

Keywords: Agents, Agentic AI, Collaborative AI, Web, Web of Agents, Decentralized AI, Interoperability, Communication

Submission Number: 874