Back to AI/ML Overview
🔗 Knowledge Site · Communication Between Agents · Messages, Tasks, A2A

Agent CommunicationOne agent asks. Another answers. The rest is the contract between them.

Two agents that work together need what any two programs need: an address, a message, a way to match a reply to its request, and an agreement about what happens when the answer is late, wrong or never comes. This site starts from one request and one result and climbs, one idea at a time, to topologies, task lifecycles, Agent Cards, the Agent2Agent (A2A) protocol, its boundary with the (), security, reliability and tracing, asking one question at each step: what must both sides agree on?

🎯The problem

An intake agent hands evidence collection to an assessment agent, and an underwriting agent waits for the result. If the reply comes late, comes twice or never comes, nothing on either side says what to do. An orchestrator is one answer, not the only one. Agent communication is the contract that makes delegation between agents safe, whoever routes the messages.

🔑The question every page asks
What must both sides agree on? Addresses, payloads, correlation ids, task states, errors, versions and timeouts, shown on three topologies of the same synthetic insurance workflow.

🗺️What is inside

Messages and topologies

Sender, recipient, payload and correlation; a central orchestrator against peer-to-peer and hierarchical designs; request and reply against asynchronous tasks.

Contracts and task lifecycle

Messages, artifacts, typed schemas, errors and versions; submitted, working, input required, completed, failed and canceled.

A2A and its boundary with

Agent Cards and skill discovery, the A2A protocol and its transports, and why delegating to an agent differs from calling a tool.

Security, reliability and tracing

Identity and delegated scopes, retries, idempotency and out-of-order delivery, and one causal timeline across all three agents.

🖱️How to read it

  • New to multi-agent systems: read in order. Each chapter starts in plain English with one tiny example.
  • Experienced: jump to the task lifecycle, A2A versus or reliability. The rigorous part sits in folded blocks.
  • Eleven interactive labs, each with a plain table next to it. Every diagram is clickable.
  • The example is synthetic and its agents are scripted stand-ins. Results show how the mechanism works, not how good any model is.