The Pulse

Open Agentic Web Challenge offers $13,000 for interoperable AI tools

The Open Agentic Web Challenge is inviting developers and researchers to build interconnected AI agents and supporting technologies. The first contest has a $13,000 prize fund and runs until January 11, 2027.

AI.info Team · · fact-checked · against 2 sources ↓ · Published 12 Nov 2026

A new $13,000 challenge is seeking AI agents and supporting technologies that can work across platforms using the existing web. The Open Agentic Web Challenge was announced on November 12 at Web Summit in Lisbon by organizations from research, industry, standards and civil society.

Its first contest runs on DevPost until January 11, 2027. Winners are to be announced at the World Economic Forum Annual Meeting in Davos in January 2027.

Who can take part

The challenge asks developers, researchers, startups, businesses and students worldwide to build publicly accessible, interconnected agentic websites. Entrants can also work on the technologies and standards that enable agents to discover and work with one another.

The initiative is supported by Cognizant AI Lab, MIT Media Lab's NANDA program, Mozilla, United Nations ITU-T Study Group 17, Amazon Web Services, UT Austin School of Computing and AI Commons. Additional organizations are invited to join as collaborators.

The project is intended to explore an approach in which people choose their own agents and organizations make their services discoverable to those agents. The contest also invites work on open components that participants can build and test using the current web stack.

Working across platforms

In its announcement, World Wide Agentic Web said personal AI agents are largely confined to the platforms where they are built and that, without open standards, that isolation could deepen. It said the challenge explores an alternative in which independently developed agents can interact across platforms.

The challenge builds on the existing web, including HTML and JavaScript, rather than requiring a new network. It aims to test approaches to discovery, identity, multi-agent orchestration, portable agent programs, transactions and governance.

The announcement describes agents as beginning to find services, communicate, negotiate and transact on behalf of people and organizations. It says the infrastructure for agents to find each other, establish trust and exchange value is still taking shape.

Optional building blocks

Participating organizations have contributed research and open-source starting points. MIT Media Lab's NANDA program lists Web Agentica and NandaIndex as research prototypes for signed, portable agent packages. The packages are designed to run in a user-controlled runtime, such as OpenClaw or a browser.

Cognizant's Neuro San is an open-source framework for building and orchestrating multi-agent systems. It lets developers coordinate collaborating agents and test how independently developed components work together.

Another starting point is Mozilla's Otari, an open-source router that gives agent applications access to different language models. Builders can create their own components or explore the contributed starting points. Use of the contributed tools is not required to enter.

Participants' priorities

Ramesh Raskar, an associate professor at MIT Media Lab and founder of the NANDA program, said the challenge would allow its research on portable agent programs to be tested alongside complementary multi-agent frameworks. He described the programs as inspectable and powered by resources authorized by users.

Arnaud Taddei, chair of ITU-T Study Group 17, said the challenge would show what is feasible and deployable for the agentic web and feed the work of international standards communities. Madisen Taylor, director of AI Commons, said an open agentic web should record agents' sources and give the people behind those sources a say in how their work is used.