Chipp Con 2025 Keynotes
In October we hosted over 100 people in Greenville, SC to share how they are using Chipp and AI to build the future of business and work. View all talks here.
Meet Lia & Christopher
Lia Passaglia and Christopher Penny work with Google's AI team, helping startups understand and build AI use cases on Google's cloud platform.
At Chipp Con SC, they introduced A2A (Agent-to-Agent Protocol), an open-source protocol Google developed to help AI agents communicate with each other. Watch their talk below 👇
Agents Working Alone
Right now when you build an AI agent, it's pretty isolated. If you're building a travel agent, you might try to make it know everything about hotels, flights, local guides, and more. But that's a lot for one agent to handle, and it probably won't be great at everything.
The better approach would be to have specialized agents that are really good at their specific tasks. A hotel agent that knows hotels inside and out. A flight agent that's an expert on flights. A local guide agent that knows everything about a specific city.
But here's the challenge: how do these agents talk to each other?
Enter A2A Protocol
A2A is basically a standard language that agents agree to use when communicating. Think of it like a secret handshake that all agents following the protocol understand.
Let's say you built a travel agent and the City of Greenville built their own specialized agent that knows everything about visiting Greenville. You have no idea how they built their agent. It could use any framework, any platform. But if both agents follow A2A protocol, they can communicate and collaborate without any problem.
From Lonely Agent to Lego Piece
A2A takes your agent from being isolated to being a Lego piece that can plug into any other system. Your main agent now has access to what the team called an "agent mesh," basically a network of additional specialized agents it can call on when needed.
How It Actually Works
There are two main parts to how agents communicate with A2A:
Agent Discovery
This is basically the business card exchange. When one agent connects to another, the first thing it learns is: who are you, what can you do, what tools do you have, what services can you perform?
This information gets hosted at the agent's URL. So when your travel agent needs to talk to the Greenville agent, it first learns that this agent specializes in all things about visiting Greenville and what specific tasks it can handle.
Agent Interaction
Once agents know about each other, they need to actually communicate. A2A provides an SDK that wraps around whatever you've built. It's just a layer in front of your agent design that creates the universal standard for communication.
Through this layer, agents establish their communication style. Can they interrupt each other? Do they update tasks in the background? How do they collaborate? All of this gets defined through the A2A protocol.
Why This Matters
The team emphasized that A2A is as good as people adopt it. Google open sourced it and donated it to the Linux Foundation, so anyone can use it and contribute to it.
As more builders learn this protocol and build agents that follow it, the more powerful the network becomes. It creates real interoperability between different agents, regardless of how they were built or what platform they're on.
Wrapping Up: Learning the Future Language
If you're building agents or exploring what's possible with AI agents, learning A2A is learning the future language of how these systems will work together. It's not just about building one really smart agent anymore. It's about building specialized agents that can collaborate with other specialized agents to get things done.
