Chapter 8. The Agent-to-Agent Protocol
Multi-agent architectures successfully distribute complexity, but deploying them across an enterprise introduces a strict interoperability constraint. Consider the teaching assistant from Chapter 6. If a student asked, “What are the latest breakthroughs in quantum computing?” the math and grammar agents wouldn’t be able to answer. You would need a research agent that can browse the web, read academic papers, and synthesize complex topics. You could build this agent yourself in ADK. But what if a fantastic “academic research agent” already exists, built by a different team using a different framework, like LangGraph or AutoGen?
Right now, your ADK agent and that LangGraph agent are trapped in walled gardens. They speak different languages, use different data structures, and would require a custom integration to connect. This Tower of Babel problem is a significant bottleneck that stops us from building truly internet-scale agentic systems.
This isn’t just a problem for individual developers; it is a massive challenge for modern enterprises. Imagine a large organization with multiple business units (BUs)—like human resources, finance, and engineering—operating in silos. Each unit is likely building its own agents to solve specific problems. HR might build an onboarding agent” in CrewAI, while finance is building a payroll agent in LangChain.
In this scenario, these efforts are isolated and duplicative. But what if we could allow them to be ...
Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.
Read now
Unlock full access