Case in point – we’re now in the era of agentic AI. People are increasingly hearing about AI agents, but few are familiar with something called NANDA.

In fact, searching for this term online can be confusing. Most results refer to a nursing association. The NANDA that is set to shape the future of the global Internet actually originated from MIT, and it's mostly known (if you’ll excuse the quantum reference) among data scientists and others in similar roles.

However, it’s likely to become a major force in our technology landscape within just a few years.

What is NANDA?

NANDA essentially serves as a complete platform for agent interactions. It's a protocol designed for a new AI-powered Internet, adapted and evolved to handle the capabilities of large language models (LLMs).

One of the most notable writers on NANDA, Rahul Shah, describes it as a “full stack protocol” where agents possess cryptographic identities – more on that shortly.

“NANDA does not replace A2A or MCP,” Shah explains, referring to Agent to Agent protocols and the Model Context Protocol developed during what could be described as the 'AI API race.' “Instead, it offers the naming, verification, and economic framework that allows agents to operate securely, scalably, and independently in real-world, distributed environments. Its purpose is to foster a self-sustaining ecosystem where valuable agents are rewarded and trusted — while spammy or harmful agents can be excluded through cryptographic audit trails.”

In terms of platform features, there’s an agent registry, and the system employs dynamic resolution logic to route agent transactions. There’s also auditing functionality and use of distributed ledger technology, where NANDA leverages zero-knowledge proofs to verify agent actions.

But all of this is a somewhat technical way to explain what NANDA really is.

Let’s approach it differently, using a more intuitive analogy tied to how AI agents resemble people.

AI Agents Line Up to be Counted

In many ways, the idea of an AI agent mirrors that of a digital person – in other words, we see these agents as possessing the cognitive abilities individuals have. We can assign them names and avatars, making them seem very human indeed. They can pass numerous Turing tests. They exist as distinct entities. They're like people.

If you extend that analogy further, NANDA acts as a kind of organizational framework for these agents. At a company, you have an organizational chart. If you're forming teams for a game of softball, you create a roster or list of names. A teacher in a classroom typically has some kind of document identifying each student.

This is essentially what NANDA facilitates. It provides a system for identifying and understanding AI agents — effectively asking: “who are they? And what do they do?”

All of this occurs within multi-agent systems where AI agents collaborate to generate solutions.

More on NANDA

I attended a panel at IIA focused on AI, where some of the leading experts in the field discussed NANDA and related topics.

My colleague Ramesh Raskar described this as working with the “building blocks” for emerging agentic systems.

Investor Dave Blundin spoke of a “variety of useful functions” and the need for a micropayments system for services.

“When this happened on the internet, nobody could figure out the revenue model, and then everything shifted to ad-based revenue, because it was simple: ‘throw up some banner ads and go with it,’” he said. “That won’t work with AI agents. You don’t want these things advertising to users.”

Aditya Challapally highlighted three key risks involved in building these systems: trust, culture, and orchestration.

“When we talk about culture, we mean things like: ‘what are the societal standards for how an agent can interact with you?’ For instance, can an agent send you a DM on LinkedIn on behalf of someone else, or should it disclose it's an agent? Establishing that cultural norm is important. And then the third part is orchestration — which is about how agents communicate with each other through a structured protocol setup?”

Panelist Anil Sharma expressed a vision for the new protocol’s potential.

“I would love to see application sustainability,” he said. “I hope to see this applied to social impact, particularly in fields such as agriculture and beyond … since that’s where data and value are trapped across ecosystems, extending into non-profits and government systems.”

And panelist Anna Kazlauskas emphasized the importance of data ownership.

“You can imagine, a few years from now, you might have an AI agent — maybe even ten AI agents — capable of autonomously performing tasks, possibly earning income on your behalf and collaborating with others,” she explained. “One risk, though, is if all those agents run on a single platform. As your AI agents begin generating real economic value, having sovereignty and true ownership over that becomes crucial.”

Blundin, speaking about the “unbundling” of services, mentioned a related concern: that AI could develop services more efficiently than traditional companies, keeping businesses on edge — made possible by a protocol like NANDA.

That gives a bit more insight into how NANDA will function and what it aims to accomplish.

Coming Soon

So, although NANDA hasn’t been widely discussed yet, it soon will be. I found it helpful to use that analogy to illustrate how emerging protocols will treat AI agents like people – assigning them names, identities, jobs, and roles as they work together, ideally in service of humanity and for our collective benefit.

The above is the detailed content of Line Up And Identify Yourselves: AI Agents Get Organized With NANDA. For more information, please follow other related articles on the PHP Chinese website!