Agentic AI Foundation: why another standard

Angie Jones is VP of Developer Experience at the Agentic AI Foundation. She hosted the conference’s day-one keynote and sat on the committee of the foundation’s first Technical Excellence Awards. Luca Berton talked to her at AGNTCon + MCPCon North America 2026 in San Jose on 23 October 2026 and asked what the foundation means, and how an agent gets from demo to production.

Where the agentic stack is built

“Essentially, this is where we say the agentic stack is being built.” The foundation is a neutral home for open source projects and protocols that are critical to building agents: MCP, A2A, AGENTS.md, Agent Skills and several others. Companies come together there to form and work on those protocols, standards and projects.

It has more than 300 member companies. Each wants to build its own innovative solutions, and some “maybe even compete with each other”. What they share is the need for an interoperability layer underneath.

Another standard?

“Sometimes people will joke and say, oh, another standard.” It sounds funny, she says, until you look at how many new problems a new space brings. Buying something on the web is pretty much solved for people. An agent buying something on your behalf is not, at least not across many websites, merchants, payment sources and credit cards. That takes “some sort of standard or maybe five along the way”. The merchant has to know that what the agent buys matches the customer’s intent, and the agent has to pay “without necessarily you exposing your credit card details to it.”

Testing what is non-deterministic

Luca asked about testing and evaluating autonomous agents. At the core of an agent is the model, which is non-deterministic: “So how do you test a non-deterministic system, right?” Evals are part of the answer, and many people are working on efficient evaluations that cover the scenarios an agent can find itself in. Another is OpenEnv, from Hugging Face, one of the foundation’s newer projects: an environment where you can try an agent in various scenarios and get more insight into how it might act.

Even then, “how it acts today might not be how it acts tomorrow.” So beyond tests, guardrails have to be in place, so that absolutely undesirable behaviour is mitigated or blocked before it happens.

Standards and developer experience

Developer experience, for her, is about enabling developers to build products and solutions. Without standards, everyone would develop on different variations that could change, and nobody would know what to build on top of.

Her example is AGENTS.md. Most clients supported it, but the Claude team had its own variety, and the community’s feedback was aimed at them: developers who use Claude for some things and other agents for others want their guidance in one document that every agent supports. According to Angie, the Claude team listened and decided to support the standard too. “This makes our lives easier, you know?”

How to get involved

There are three ways in:

• Ambassadors: 140 volunteers from every major region of the world, who enjoy helping other developers. They write blog posts, give conference talks, speak at local meetups or make videos with the lessons they have learnt.
• Chapters: about 100 around the world, where local organisers hold regular events so that developers can learn together and share ideas and techniques.
• Contributing: the foundation has ten projects, all open source and welcoming to contributors. Whatever company or part of the world you come from, you can bring an idea to their Discord channels or GitHub orgs and talk it through with the maintainers.

And next year? The events are still being planned, but the two flagship ones are announced: AGNTCon + MCPCon in San Francisco on 28 and 29 April, and in London on 15 and 16 September.

More

The episode with Angie Jones is being produced and goes out on YouTube first. Everyone the show has met there: the Agentic AI Foundation on Udienza. The day it was recorded: AGNTCon + MCPCon North America 2026: day two.