Multi-Agent AI Systems & the A2A Protocol

NOTE: You will probably need to use a personal laptop for the labs environment. Corporate machines may restrict access to resources we need.

Single agents hit a ceiling. The real power — and the real complexity — comes from multiple specialized agents working together. 57% of organizations already deploy multi-step agent workflows, 16% have cross-functional agents, and 81% plan expansion. Google's A2A (Agent-to-Agent) protocol, now at v1.0 with 150+ organizations under Linux Foundation stewardship and production deployments at Microsoft, Salesforce, and ServiceNow, standardizes how agents discover each other via Agent Cards, negotiate task delegation via JSON-RPC 2.0/gRPC, and exchange structured results with multi-modal payloads. MCP defines agent-to-tool communication; A2A defines agent-to-agent communication — together they form the complete interoperability stack.
This full-day workshop teaches participants to design, build, and orchestrate multi-agent systems. They'll work hands-on with the four leading frameworks — LangGraph (126K+ stars, graph-based orchestration), CrewAI (44,300+ stars, 5.2M monthly downloads, role-based crews), AutoGen/Microsoft Agent Framework 1.0 (conversational GroupChat + A2A native support), and Google ADK (8,200+ stars, SDKs in Python/TypeScript/Java/Go) — and connect agents across frameworks using A2A for true cross-platform interoperability. Participants build three progressively complex multi-agent systems, culminating in a heterogeneous A2A-connected agent network.

Prerequisite: Experience building single AI agents (through the AI Accelerator, the Implementing AI Agents in Python workshop, or similar). Comfort with Python and LLM APIs. Familiarity with an agent framework helps but isn’t required. You’ll need a free GitHub account; a free Groq API key is recommended for faster responses.

Covers:

multi-agent architecture patterns (supervisor, hierarchical, collaborative, competitive, swarm)
how agents communicate (delegation, negotiation, consensus, escalation)
LangGraph supervisor teams and CrewAI role-based crews
A2A v1.0: Agent Cards and discovery, the task lifecycle, messages and artifacts, live streaming, the JSON-RPC, gRPC, and HTTP transports, and security features such as signed cards and standard sign-in methods
connecting agents across frameworks with A2A
choosing among LangGraph, CrewAI, Google ADK, and Microsoft Agent Framework
production concerns: budget limits, preventing deadlocks, and tracing every hop

Format: Three hours of live instruction (two 90-minute blocks with a 30-minute break), with short, focused lectures between five hands-on labs.

About Brent Laster

Hi, I'm Brent Laster - a global trainer and book author, experienced corporate technology developer and leader, and founder and president of Tech Skills Transformations LLC. I've been working with and presenting at NFJS events for many years now and it is always exciting and interesting.

Through my decades in programming and management,I've always tried to make time to learn and develop both technical and leadership skills and share them with others Regardless of the topic or technology, my belief is that there is no substitute for the excitement and sense of potential that come from providing others with the knowledge they need to help them accomplish their goals.

In my spare time, I hang out with my wife Anne-Marie, 4 children and 2 small dogs in Cary, North Carolina where I design and conduct trainings and write books. You can find me on LinkedIn (linkedin.com/in/brentlaster), Twitter (@brentclaster) or through my company's website at www.getskillsnow.com.

More About Brent »