AI

Generative AI in the Real World: Understanding A2A with Heiko Hotz and Sokratis Kartakis

O’Reilly Media
O’Reilly Media•Jun 15, 2026

Why It Matters

A2A standardizes how autonomous agents interoperate, unlocking modular AI services at enterprise scale while demanding new security and governance controls.

Key Takeaways

  • •Agent-to-agent protocols enable seamless integration across diverse frameworks.
  • •Stateful communication lets agents handle asynchronous tasks and updates.
  • •A2A complements, not replaces, tool-focused MCP protocols in practice.
  • •Router agents act as orchestrators, sharing context among specialist agents.
  • •Security layers like API keys and OAuth are essential for A2A.

Summary

The podcast features Google Cloud’s Heiko Hotz and Sokratis Kartakis explaining the emerging Agent‑to‑Agent (A2A) protocol, a stateful communication layer designed to let autonomous AI agents talk to each other without code rewrites. They argue that as developers proliferate agents built on LangChain, CrewAI, Google’s Agent Development Kit and other stacks, a common language becomes essential for scaling across teams and organizations. Key insights include the need for a protocol that supports asynchronous task execution, status updates, and natural‑language descriptions of capabilities—features that differentiate A2A from the tool‑centric Model Context Protocol (MCP). A2A is positioned as complementary to MCP: MCP handles single‑tool calls with fixed schemas, while A2A orchestrates whole agents via a router that maintains shared context and memory across specialist services. The hosts illustrate the concept with concrete analogies: agents as business cards exchanged between employees, a vending‑machine versus concierge model, and a travel‑booking workflow where a router agent coordinates flight, hotel, and car agents. They also flag security concerns, noting that each agent must authenticate via API keys, OAuth, or similar mechanisms to prevent injection attacks and ensure trustworthy interactions. Implications are clear: enterprises will move from monolithic AI applications toward micro‑service‑style agent ecosystems, enabling faster development, easier testing, and more resilient fault tolerance. Standardizing A2A could become a cornerstone for cross‑team AI collaboration, driving both innovation and the need for robust security frameworks.

Original Description

Everyone is talking about agents: single agents and, increasingly, multi-agent systems. What kind of applications will we build with agents, and how will we build with them? How will agents communicate with each other effectively? Why do we need a protocol like A2A to specify how they communicate? Join Heiko Hotz and Sokratis Kartakis for a chat about A2A and our agentic future.
Follow O'Reilly on:

Comments

Want to join the conversation?

Loading comments...