An A2A task is a stateful unit of work with an explicit lifecycle. It lets independent agents collaborate on long-running operations and multi-turn work while exchanging relevant context and results. section 7.8.1
In an A2A interaction, a client agent delegates work with the context needed for the task. A server agent processes the request in its own isolated context and returns results. An agent can act as a server for one interaction and a client when delegating another subtask. section 7.8.1
Delegating travel planning
A travel-planning orchestrator can send travel dates to a flight-booking agent, location constraints to a hotel agent, and budget preferences to an itinerary agent. Each receives relevant context rather than the orchestrator’s entire prompt history. section 7.8.1
Figure 7.3 Interaction flow between a client and server using the A2A communication protocol
The interaction flow places delegated work in the context of communication between client and server agents. section 7.8.1
Other systems use task for related but distinct concepts. In CrewAI, a task defines expected outputs and dependencies within a coordinated workflow. In an evaluation, a task is a single test with defined inputs and success criteria; multiple trials can measure consistency. Neither definition should be assumed to describe the A2A task lifecycle. section 10.4.2 section 8.2.1
A2A continues to evolve. Check the current specification before deploying a production implementation. section 7.8.1

