Google's A2A Protocol: The Future of AI Agents

TL;DR

Google's new Agent-to-Agent (A2A) protocol is an open standard that lets AI agents communicate and collaborate, finally making complex, automated workflows scalable. This is the missing piece. #VentureStep #AI #GoogleCloud

INTRODUCTION

AI agents are incredibly powerful, capable of handling specific tasks with remarkable efficiency. However, they have largely operated in digital silos, unable to communicate with each other to tackle more complex, multi-step problems. This limitation has been a major barrier to scaling AI automation to its full potential. At their recent Cloud Next 2025 conference, Google proposed a groundbreaking solution that could fundamentally change the AI landscape.

In this episode of Venture Step, host Dalton Anderson unpacks one of the most significant announcements from the keynote: the Agent-to-Agent (A2A) protocol. Drawing on his analysis of the dense, two-hour presentation, Dalton explains why this open standard for communication isn't just an incremental improvement, but a critical piece of infrastructure required for the next evolution of artificial intelligence in the enterprise.

Dalton explores what the A2A protocol is, how it builds on foundational concepts like Anthropic's Model Context Protocol for tool usage, and why its open, cross-platform nature is essential for preventing proprietary "walled gardens." Featuring insights from Google's impressive live demo with Salesforce, this episode breaks down how communicating agents will unlock a new era of sophisticated, scalable business automation.

KEY TAKEAWAYS

• The Agent-to-Agent (A2A) protocol is an open standard proposed by Google to allow AI agents from different systems and tech stacks to communicate securely and effectively.
• This protocol is the missing piece for scalability, enabling complex, multi-agent workflows that were previously impossible due to communication barriers between siloed systems.
• Unlike proprietary ecosystems like Apple's, A2A is designed to be open, preventing vendor lock-in and encouraging widespread adoption and innovation across the entire tech industry.
• Google is building a comprehensive ecosystem in Vertex AI with tools like the Agent Developer Kit and Agent Garden to simplify the process of building, managing, and deploying these collaborative agents.
• After communication is solved, the next frontier for AI is enabling agents to conduct economic transactions, such as hiring other agents or even humans for specialized tasks.

FULL CONVERSATION

Google's Vertically Integrated AI Powerhouse

Dalton: Welcome to Venture Step Podcast, where we discuss entrepreneurship, industry trends, and the occasional book review. ¹Google just had their Google Cloud Next 2025 conference in Las Vegas, and I watched the two-hour, very dense video and demos, and something had caught my eye. ²There are a couple of things, but for this episode, we'll be discussing one. ³

And I think it's one of the last pieces for agents to actually scale. And it is the agent to agent protocol. ⁴

Dalton: What is that, why is it important, and what does it mean? That's everything we'll be discussing in today's episode. 5 Once again, I'm Dalton Anderson, your host of Venture Step Podcasts.

Dalton: So Google had a wonderful conference and a keynote that was around an hour and 40 minutes. ⁶A lot of the information was industry-leading and the stuff that they were announcing was incredible. ⁷It really truly shows how much money they've spent on AI and their technology stack being vertically integrated—from servers to building their own chips to infrastructure, to oceanic fiber networks, to the cloud infrastructure they provide to themselves and their clients, to the video and text data they're able to train on, like YouTube or Chrome. ⁸

Dalton: All of that stuff combined into a wonderful AI product. ⁹Now that AI product is industry-leading, Gemini 2.5 Pro, as of about 10 days ago, was number one on Chatbot Arena. ¹⁰Their models, for compute costs and performance, are the best-performing models on the market. ¹¹So Google is just killing it in so many different areas. ¹²And one thing that was important was the agent to agent protocol. ¹³I'll get into why that's important in just one moment. ¹⁴I just wanted to give a shout-out to Google for their wonderful keynote. ¹⁵

The Precursor: Anthropic's Model Context Protocol (MCP)

Dalton: So now we're going to get into the agent to agent protocol. Before I talk about the agent to agent protocol, there's another thing that's also important to talk about, and that'd be Anthropic's Model Context Protocol, or MCP. ¹⁶Anthropic's MCP is an open protocol that allows models to integrate with external tools, whether it be Google Maps or the internet through Microsoft or Chrome. ¹⁷

Dalton: Basically, instead of building custom code to embed your product into another product, this MCP protocol establishes a standard. ¹⁸If your AI model wants to look at financial data, each company doesn't have to build out custom code to embed their product into S&P Global. ¹⁹What this MCP protocol does is say, "Here are the key places where AI models want to go, and let's just make a single protocol for these AI models to access this information." ²⁰

Dalton: One, the vendor will be able to potentially be compensated. ²¹And two, it allows people to easily build on top of what has pre-existed. ²²That's important because if everybody's utilizing their resources to build the same thing as other folks, then those resources aren't allocated in the most efficient way. ²³This MCP allows models and companies just to use the protocol instead of building out their own proprietary protocol for external tool integration. ²⁴

Dalton: It's a very important piece of providing additional information to the query user and for grounding your answers in truth. ^25252525For example, if you ask if a store is open, you might have historical training data saying it is, but things change. ²⁶The way you can ground your truth is by verifying it on Google Maps or Apple Maps, and the way to do that is with this MCP protocol built by Anthropic. ²⁷It’s also worth noting that Google is an Anthropic investor, and Anthropic trains their data on Google's TPUs, or Tensor Processing Units, which are Google's proprietary chips for AI workflows. ²⁸

Introducing the Agent-to-Agent (A2A) Protocol

Dalton: Okay, with all that out of the way, you understand the lay of the land. ²⁹There's this tool usage protocol built by Anthropic, and now there's this other protocol, the Agent-to-Agent protocol. ³⁰This allows agents to communicate with other agents. ³¹

But once you try to have a complex workflow of multiple agents, if they can't communicate with each other and share data in a safe manner, or share data at all, then it doesn't necessarily work. ³²

Dalton: Now, what this Agent-to-Agent protocol allows is for agents to communicate with different systems. ³³It doesn't have to be on your system, which is great, so it's not owned by anybody. ³⁴ It's not owned by Google; Google proposed a protocol, but it's open to the whole world to use. ³⁵Google has a pretty good head start as they have a lot of high-profile partners for it, the most notable one being Salesforce. ³⁶It basically breaks down the communication barriers between agents, allows for more complex workflows, and allows agents to be a feasible technology that can scale. ³⁷

Breaking Down Walled Gardens: Why An Open Protocol Matters

Dalton: As I mentioned before, this agent to agent protocol is not specified for one framework or technology stack. ³⁸So if you had your agent built in Google's Gemini and Vertex AI, and then you had some kind of database stuff in MongoDB or AWS, it can still work. ³⁹Other vendors, when they're building these kinds of full-suite things, it's got to be within their walled garden. ⁴⁰Think about Apple; they do that great. ⁴¹When you're on a Mac and you want to communicate between an iPhone and a Mac, that works great. ⁴²But as soon as you work outside of the ecosystem, everything breaks. ⁴³

Whereas the agent to agent protocol is built to work with any technology stack. ⁴⁴

Dalton: Everybody communicates with everybody. And not only is that the availability, but as I mentioned before, it doesn't force people to consistently utilize their resources on things that other folks will be building. 45It's kind of like building it once, and that's the way it is. 46There's less resources spent on building the same thing, and people can spend their resources on things that are more important to add more value. 47

Live Demo: A2A in Action with Salesforce

Dalton: Before I get into why this is important, I want to share the demo from the keynote. This is Patrick Marlowe, the Product Manager of Applied AI at Google. ^48484848If you want to watch it yourself, it's in the "Google Cloud Next 25 opening keynote" video at the timestamp of one hour, six minutes, and 29 seconds. ⁴⁹I think it really emphasizes the power of the Agent-to-Agent protocol. ⁵⁰

(Video clip plays from 12:12 to 18:07)

Dalton: Okay, that is insane. ⁵¹In the backend, there are agents—a shopping agent, a customer service agent—and a connection to your Salesforce instance where, for certain things they don't have authority to do, they request approval from their sales manager. ⁵²There's a human in the loop for certain transactions. ⁵³The cool thing about all of that was it seemed pretty seamless. ⁵⁴

Dalton: If you were a manager getting these requests, you could respond with what you're comfortable with, and the agent would handle it on your behalf. ⁵⁵From the customer's perspective, everything is happening asynchronously, and there's never a real lapse in engagement. ⁵⁶I wouldn't really care if I was talking to a customer service agent and it was AI if they're able to suggest the right items I need. ⁵⁷

Dalton: I could share my camera, they can identify which plant I have, tell me I have the wrong soil, and then ask if it's okay to switch out the ones I have. ⁵⁸ That's pretty cool. Then it's also able to schedule an appointment and escalate things to the manager for approvals that you would have to wait for anyways. ⁵⁹ All of that? Chef's Kiss. Such a cool technology. ⁶⁰That at scale would really pay dividends for customer service. ⁶¹

Why Agent Communication is a Scalability Game-Changer

Dalton: That live demo was 100% live, not recorded. ⁶²They did it live in front of the audience. ⁶³I think they're comfortable doing that because they already have 50 partners utilizing these AI agents and this protocol every day. ⁶⁴

Dalton: That's why I think it's important. Agents in themselves are great ideas, but they're only good if they can communicate with other things. 65

you'll never be able to build an automated agent workflow without agents being able to communicate with other agents. ⁶⁶

Dalton: You're not going to be able to train an agent to do every single thing, but you could train an agent to do certain tasks very well. ⁶⁷You'll need to communicate with other agents if you want to create a workflow, which is the bread and butter. ⁶⁸

That's what scales. That's the technology that would allow rapid adoption and advancement in a short amount of time. 69

Dalton: And that's what Google is trying to do. Google is the first hyperscaler to come out with this open protocol. 70A hyperscaler would be like Google, AWS from Amazon, or Microsoft cloud services. 71

Google's Toolkit: The Agent Developer Kit and Agent Garden

Dalton: Now that you have the Agent-to-Agent protocol, you've got to make it easy to build agents. ⁷²And that's exactly what Google has done. ⁷³Within their Vertex AI platform, they built an Agent Developer Kit. ⁷⁴That provides many complex agent workflow templates that are already pre-built and integrated with key partners like Salesforce or Workday. ⁷⁵

Dalton: That's available right now. ⁷⁶Then they built this Agent Garden, where you access and manage all your agents. ⁷⁷So you have the Agent Developer Kit for templates, and then you put it in the Agent Garden where your agents live. ^78787878The Agent Garden will allow you to learn from working examples and manage your agents. ⁷⁹What they showed on the agent dashboard, or the Agent Engine, was an agent request to hire a software developer, where the agent found candidates and set up interviews. ⁸⁰

What's Next? The Agent Economy and Economic Transactions

Dalton: So this is the future agent workflow. What's missing? ⁸¹

In my opinion, once agents can communicate with other agents, the next thing is agents being able to conduct economic actions or transactions. ⁸²

Dalton: That would mean there needs to be some kind of protocol that allows agents to hire and transact with other agents or with humans. ⁸³That's the last thing, and from there, you should be able to scale this whole agent economy in a pretty cool way. ⁸⁴I think it'd be very cool that agents could hire people or agents could hire other agents. ⁸⁵I can definitely see agents hiring humans to do, at the moment, at least manual labor, because the agents are virtual and don't have interaction with the physical world. ⁸⁶

Dalton: Maybe they can rent a robot. ⁸⁷Maybe there's a company that just builds out robots all across the country and agents could pay to rent the robot to do tasks that they're asked to do. ⁸⁸ Who knows? You might be thinking from what I'm saying, "Wow, that's crazy, that's so far out." But I don't think it's as far out as you think. ⁸⁹I think it's closer than it is further away. ⁹⁰

The Unbelievable Pace of AI Advancement

Dalton: I think you'd truly be surprised how close these things are versus how far away they were two years ago. ⁹¹Three years ago, would you say that agents could have a full-on phone conversation without a robotic voice? ⁹²That agents can request access to your camera, look at what you're showing them, and comprehend it in real time? ⁹³ All of those things three years ago...

If I would have told you three years ago, three years from now, this is what's going to be live demoed. We'll have that capability. People would not believe you. ⁹⁴

Dalton: A lot of the technologies that we have now, that are being built and rolled out and are massively available, people just aren't utilizing them yet. ⁹⁵People would say you're insane, that you're overly optimistic. ⁹⁶

Dalton: But every time I turn around, there's some kind of crazy AI news. ⁹⁷Companies are spending hundreds of billions of dollars every year on AI. ^98989898We are getting a really cool announcement at a minimum every three months. ⁹⁹Even xAI has announced stuff that is pretty advanced. ¹⁰⁰I think recently they announced you can share your camera and Grok can understand what's going on in your environment in real time. ^101101101101The level of advancement that xAI is having in such a short amount of time is incredible. ¹⁰²They've been around for like three years and they've already caught up and in some places passed their incumbents. ¹⁰³

Dalton: So I just want to keep that perspective in mind when you're thinking about what's next. ¹⁰⁴I'm surprised every day. ¹⁰⁵I just want to make sure that it stays exciting and we just keep pushing for a better world. ¹⁰⁶But of course, wherever you are in this world, good afternoon, good morning, good evening, and thank you for listening. ¹⁰⁷I can't wait for you to listen in next week. ¹⁰⁸Have a great day. ¹⁰⁹Goodbye. ¹¹⁰

RESOURCES MENTIONED

• Google Cloud Next 2025
• Google
• Anthropic
• Gemini 2.5 Pro
• Salesforce
• Vertex AI
• AWS (Amazon Web Services)
• Microsoft
• MongoDB
• Workday
• Apple
• xAI (Grok)

INDEX OF CONCEPTS

Dalton Anderson, Google Cloud Next 2025, Las Vegas, Agent to Agent protocol, A2A protocol, Gemini 2.5 Pro, Chatbot Arena, Anthropic, Model Context Protocol, MCP, Google Maps, S&P Global, TPUs, Tensor Processing Units, Salesforce, Patrick Marlowe, Vertex AI, Agent Developer Kit, Agent Garden, Agent Engine, AWS, Amazon, Microsoft, MongoDB, Workday, Apple, xAI, Grok