Core Objects
Core Objects
Use these references to describe an agent, exchange messages, track tasks, and return results. Each object reference explains its fields and protocol rules, with examples from the official documentation. For an overview of how the objects relate, read A2A Key Concepts and Components .
| Object | What it is | Read this page when |
|---|---|---|
| Agent Card | The JSON document an agent publishes to describe itself | You publish an agent, or need to discover and select one |
| Task | A stateful unit of work with a lifecycle | You handle long-running work, multi-turn input, or cancellation |
| Message | One turn of communication between client and agent | You send requests or return clarifying and status messages |
| Part | The content container: text, file, or structured data | You exchange files or JSON, not only text |
| Artifact | A concrete output produced by a task | You return results, or stream them in chunks |
When declaring agent capabilities and authentication, consult these Agent Card objects:
| Object | What it is | Read this page when |
|---|---|---|
| AgentSkill | One entry in the card’s skills list | You describe what your agent can do, or read another agent’s card |
| SecurityScheme | How a card declares authentication, from API keys to OAuth 2.0 | You protect an agent endpoint, or need to know how to authenticate to one |
How the objects fit together
A client reads the Agent Card to learn how to reach an agent and what it can do. It then sends a Message, whose content is one or more Parts. The agent answers with either a direct Message or a Task. A Task moves through states, and the results it produces are Artifacts, which are also made of Parts.
Related pages
- Life of a Task : when to answer with a Message and when to start a Task.
- Data Structures : the complete field reference for every object.
- Common Workflows and Examples : the objects inside full requests and responses.
Source: A2A official documentation: Key Concepts and Protocol Data Model