The Open Agentic Web Challenge: building an agent economy that works for everyone

• $13,000 global challenge invites developers to build open, interoperable AI agents on today's web
• Supported by Cognizant AI Lab, MIT Media Lab's NANDA program, Mozilla, United Nations ITU-T SG17, AWS, UT Austin School of Computing, and AI Commons.

LISBON, Portugal, Nov. 12, 2026 — Web Summit -- Organizations spanning research, industry, technology standards and civil society today announced the Open Agentic Web Challenge, a global initiative to build an open, safe and interoperable web of AI agents.

The challenge invites developers, researchers, startups, businesses and students to build publicly accessible, interconnected agentic websites. Entrants can also develop the technologies and standards that let agents discover and work with one another in a safe, trusted and reliable way.

The first contest runs on DevPost until 11th January 2027. A $13,000 prize fund will recognize outstanding contributions, with winners announced at the World Economic Forum Annual Meeting in Davos in January 2027. Additional organizations are invited to join as collaborators.

Why an open agentic web matters

AI agents are beginning to discover services, communicate, negotiate and transact on the web on behalf of people and organizations. The infrastructure that will determine how these agents find one another, establish trust and exchange value is still taking shape.

Today, personal AI agents are largely confined to the platforms on which they are built. Without open standards, that isolation could deepen, making it harder for people and businesses to connect them to the services they use. This challenge explores an alternative where people can choose their own agents, organizations can make their services agent-discoverable, and independently developed agents can interact across platforms.

Rather than requiring an entirely new network, the challenge builds on the web that already exists, including familiar technologies such as HTML and JavaScript. It aims to test practical approaches to discovery, identity, multi-agent orchestration, portable agent programs, transactions and governance.

How to participate

The Open Agentic Web challenge invites developers, researchers, startups, businesses and students worldwide to build and test open components of the agentic web on today's web stack. Builders can develop their own components or explore research and open-source starting points, contributed by some of the participating organizations, including:

• MIT Media Lab's NANDA Web Agentica and NandaIndex: research prototypes for signed, portable agent packages that run in a user-controlled runtime, such as OpenClaw or the browser.
• Cognizant's Neuro® San: an open-source framework for building and orchestrating multi-agent systems, letting developers coordinate collaborating agents and test how independently developed components work together.
• Mozilla's Otari: an open-source LLM router that gives agent applications flexible access to different models.

These contributions address different parts of the problem and their use is not required to enter the challenge. Builders can enter the challenge by visiting this DevPost site.

Here's what some of the participating organizations had to say about the challenge and world wide agentic web initiative:

Babak Hodjat, Chief AI Officer at Cognizant said:

"Personal agents have arrived, but they should work for you, not a platform, and interact freely with your bank, airline and utility provider, whoever built their agents. Every business should be discoverable, not just those with access to closed platforms. We're inviting the world to build an open agent economy on the web, where everyone can participate, compete and innovate."

Ramesh Raskar, Associate Professor, MIT Media Lab and Founder of NANDA program, said:

"The open web succeeded because people could publish, discover and connect without depending on one platform. NANDA is bringing those foundations to AI agents. Our WebAgentica research makes agent programs portable, inspectable and powered by user-authorized resources. This challenge lets us test these ideas alongside complementary multi-agent frameworks."

Raffi Krikorian, CTO of Mozilla, said:

"We are excited to get open source AI tooling into the hands of builders so they can learn how to own their AI stack and get open source AI into production."

Arnaud Taddei, Chair, ITU-T Study Group 17 said:

"The standards that made the internet and the web were the ones proven in the field. This challenge is essential for the same reason: it will show what is feasible and deployable for the agentic web, and feed the work of international standards communities, including ITU-T Study Group 17 and its Focus Group FG-TIDA."

Peter Stone, Head of the School of Computing and Professor of Computer Science, The University of Texas at Austin, said:

"The Worldwide Agentic Web is an opportunity to make AI infrastructure interoperable and user-controllable. Open protocols enable people to select agents across providers, define authorization, and inspect interactions. This aligns with UT Austin School of Computing's mission to combine technical rigor with human needs in AI development."

Madisen Taylor, Director, AI Commons, said:

"Agents will soon do much of the web's reading and deciding for us. An open agentic web should record the sources they draw on, and give the people behind them a say in how their work is used. With provenance and participation built in, agents can broaden access to information while strengthening trust and human agency."

Participating organizations

The initiative spans industry, academia, civil society and international policy. Organizations involved in the initiative include:

• Cognizant AI Lab: brings deep expertise in multi-agent systems and AI safety from its AI Lab, plus open-source tooling that gives challenge participants a practical foundation for building and connecting agents.
• Mozilla: brings decades of experience championing the open web, making sure the agentic web is built on principles of openness, transparency and user control.
• MIT Media Lab's NANDA program program: contributes pioneering research on the core infrastructure of the agentic web, including open agent discovery, identity and trust.
• AI Commons: building AI's missing public record: making what agents know traceable and giving people a say in how their knowledge is used.
• ITU-T Study Group 17: the United Nations' lead group on security standards, which brings global expertise in digital identity and trust so that agents can verify who they're dealing with and act accountably across borders and platforms.
• Amazon Web Services (AWS): brings global cloud infrastructure and agentic AI services, helping builders develop, deploy and scale secure, interoperable agents.
• UT Austin School of Computing: contributes leading academic research in AI, grounding the initiative in rigorous science with access to researchers.

Organizations looking to join the challenge as collaborators can request more information at worldwideagenticweb.org.

Media contact

Paul Jarratt
[email protected] 

SOURCE World Wide Agentic Web