E11 · Distributed agents (A2A)

Family: Multi-agent

Agents run as independent services, sometimes from different vendors, that discover each other and talk through a protocol.

Principle

Each agent is an independent service, possibly supplied by another vendor. It presents itself through an Agent Card (identity, skills, security) and talks to the others through the A2A protocol. A2A covers communication between agents, MCP access to tools: the two are complementary.

The problem it solves

Agents that belong to different teams or organisations cannot share a single process or a single context.

When to use it

  • Organisation or team boundaries, agents supplied by third parties, compliance that requires isolation.

When to avoid it

  • The need is internal and a plain tool call covers it: prefer a tool exposed through MCP.
  • You expect measured production feedback: A2A adoption remains little measured.

Common pitfalls

The anti-patterns that misuse this pattern, best established first. Each sheet describes the symptoms, the public incidents and the remedies.

Tools and agent-computer interface

Security and permissions

Oversight, operations and governance

All anti-patterns

Synonyms by vendor

  • Agent-to-Agent (A2A) (Linux Foundation, AWS Strands)

Neighbouring patterns

Sources

Content reviewed on 2026-10-02

Pattern catalogue