A2A: The Agent2Agent Protocol
by DeepLearning.AI, Google Cloud, IBM Research
Wrap agents built in Google ADK, LangGraph, BeeAI and Microsoft Agent Framework as A2A servers and orchestrate them into one multi-agent system.
Overview
A2A: The Agent2Agent Protocol is a DeepLearning.AI short course announced on February 12, 2026, built with Google Cloud and IBM Research and taught by Holt Skinner (Developer Advocate, Cloud AI at Google), Ivan Nardini (AI/ML DevRel at Google Cloud) and Sandi Besen (AI Research Engineer and Ecosystem Lead at IBM Research). A2A is the open protocol Google Cloud launched in April 2025 and donated to the Linux Foundation, and IBM's Agent Communication Protocol later merged into it; it standardises how agents discover and communicate with each other, complementing MCP, which connects agents to external tools and data. The course runs about 1 hour 27 minutes over 14 lessons with eight code notebooks. It opens with why A2A exists and its architecture, then builds a Q&A agent with Claude Haiku 4.5 on Vertex AI, wraps it in an A2A server and calls it with an A2A client. Later lessons build a health research agent and a sequential chain agent in Google ADK, a healthcare provider agent that uses LangGraph with an MCP server, a client in Microsoft Agent Framework, and a full multi-agent system in BeeAI, before covering running agents on Agent Stack and advanced topics such as A2A extensions and security. The notebooks are public in Holt Skinner's A2AWalkthrough GitHub repository.
At a Glance
- Topic
- Agentic
- Level
- Intermediate
- Format
- Course
- Cost
- Freemium
- Duration
- ~1.5 hours, 14 lessons with 8 code notebooks plus a quiz, self-paced
- Provider
- DeepLearning.AI, Google Cloud, IBM Research
- Hands-on
- Yes — code/exercises
- Certificate
- Available
What You’ll Learn
- ✓Explain why the A2A protocol exists and how it complements MCP
- ✓Describe the A2A architecture for agent discovery and agent-to-agent communication
- ✓Wrap an existing Q&A agent running on Vertex AI into an A2A server
- ✓Build A2A clients with the A2A SDK and with Microsoft Agent Framework
- ✓Expose agents built with Google ADK, LangGraph and BeeAI as A2A-compliant servers
- ✓Orchestrate A2A agents into sequential and hierarchical multi-agent workflows end to end
- ✓Combine A2A with an MCP server inside a LangGraph healthcare provider agent
- ✓Apply advanced A2A concepts including protocol extensions and security considerations
Highlights
- •Taught by the protocol's own ecosystem: Google Cloud developer advocates plus IBM Research, whose Agent Communication Protocol merged into A2A
- •Deliberately multi-framework: ADK, LangGraph, BeeAI and Microsoft Agent Framework interoperate in one system, which is exactly what the protocol is for
- •All eight notebooks are public in the A2AWalkthrough GitHub repository, so the code can be run without watching the videos
- •Treats A2A and MCP as complementary layers in one build rather than rivals, and closes with extensions and security
- •Enrollment is free on the DeepLearning.AI platform; the graded quiz and certificate require a Pro membership
Who It’s For
Best For
- ✓AI engineers building multi-agent systems that span more than one agent framework
- ✓Platform teams deciding how internal agents should discover and call each other
- ✓Developers already using MCP who now need agent-to-agent communication
Prerequisites
- •Familiarity with Python
- •Basic understanding of AI agents and tool calling
- •Access to the Gemini API or Vertex AI helps when running the notebooks yourself
FAQ
What is A2A: The Agent2Agent Protocol?
A2A: The Agent2Agent Protocol is a 1h27m intermediate DeepLearning.AI short course, built with Google Cloud and IBM Research, for developers building multi-agent systems. You build a healthcare multi-agent system from agents written in different frameworks, expose each one as an A2A server, call them with A2A clients, and orchestrate sequential and hierarchical workflows.
Is A2A: The Agent2Agent Protocol free?
A2A: The Agent2Agent Protocol offers free content, with paid options for certificates or premium features.
What level is A2A: The Agent2Agent Protocol for?
A2A: The Agent2Agent Protocol is aimed at a intermediate audience. Recommended background: Familiarity with Python, Basic understanding of AI agents and tool calling, Access to the Gemini API or Vertex AI helps when running the notebooks yourself.
How long does A2A: The Agent2Agent Protocol take?
Expect roughly ~1.5 hours, 14 lessons with 8 code notebooks plus a quiz, self-paced. Most learners work through it at their own pace.
What will I learn from A2A: The Agent2Agent Protocol?
You'll learn: Explain why the A2A protocol exists and how it complements MCP; Describe the A2A architecture for agent discovery and agent-to-agent communication; Wrap an existing Q&A agent running on Vertex AI into an A2A server; Build A2A clients with the A2A SDK and with Microsoft Agent Framework; Expose agents built with Google ADK, LangGraph and BeeAI as A2A-compliant servers; Orchestrate A2A agents into sequential and hierarchical multi-agent workflows end to end; Combine A2A with an MCP server inside a LangGraph healthcare provider agent; Apply advanced A2A concepts including protocol extensions and security considerations.
Topics
Sources
This page was written from 3 sources, 2 on domains other than deeplearning.ai.