Agent2Agent (A2A) Protocol Essentials
Prerequisites: mcp-essentials
The Agent2Agent (A2A) Protocol is an open standard that defines how agents interact with each other, enabling interoperability across agents built by different teams, technologies, and organizations. Completing this lesson gets you to your first working agent link - the Track 1 outcome.
What you will learn
- What A2A standardizes and how agents advertise capabilities.
- How MAF secures A2A interactions at the Omni Gateway.
- How to describe an agent-to-agent link.
What A2A standardizes
A2A lets a client agent discover a remote agent’s capabilities via an Agent Card, then exchange messages using JSON-RPC, with streaming responses delivered over Server-Sent Events (SSE) (A2A Protocol spec, Streaming & Async in A2A).
Securing A2A with Omni Gateway
MAF protects A2A interactions at the Omni Gateway by requiring authentication and authorization, rewriting Agent Card URLs to match the public server instance, logging or blocking sensitive information in messages, and logging SSE content for compliance audits (Securing Agent Interactions with Omni Gateway).
Included A2A policies you will use later in Track 3 include A2A Agent Card, A2A PII Detector, A2A Prompt Decorator, A2A Quality Evaluation, and A2A Schema Validation (Included Policies).
This is your first agent link
Describing an agent link
# agent-network.yaml (excerpt)
agents:
- id: triage-agent
a2a:
card: https://gw.example.com/triage/.well-known/agent.json
- id: billing-agent
a2a:
card: https://gw.example.com/billing/.well-known/agent.json
links:
- from: triage-agent
to: billing-agent
Where to go next
You have completed Track 1. Continue your journey based on your role - Integration Developers and Security Engineers typically proceed to Track 3 (UAPIM & Agentic Enterprise), where you discover, govern, and orchestrate agents at scale.