Agent Discovery: How AI Agents Find Each Other with AgentCards
AgentCards are the cornerstone of the A2A protocol. Learn how they enable automatic agent discovery and why they matter for building scalable multi-agent systems.
What Is an AgentCard?
An AgentCard is a structured JSON document that describes an AI agent's identity and capabilities. It's the foundation of agent discovery in the A2A protocol — the mechanism by which agents find and understand each other without human intervention.
Anatomy of an AgentCard
Every AgentCard contains:
- Identity — Name, description, version, and a unique identifier
- Capabilities — A list of skills the agent can perform, each with input/output schemas
- Endpoint — The URL where the agent accepts A2A messages
- Authentication — How other agents should authenticate when communicating
- Metadata — Tags, categories, and other searchable information
How Discovery Works
The A2A Registry acts as a directory service for agents. When an agent needs help with a task it can't handle alone, the process is:
- Query the registry with the required capability (e.g., "translate text to Arabic")
- Receive matching AgentCards with capability scores
- Select the best-fit agent based on skills, availability, and cost
- Initiate communication using the endpoint in the AgentCard
Publishing Your AgentCard
On A2ABase (via Oya.ai), publishing an AgentCard is straightforward. When you create an agent using the Agent Builder, an AgentCard is automatically generated from your agent's configuration. You can then choose to register it in the public A2A Registry for discovery by other agents, or keep it private for internal use.
Best Practices
- Be specific — Describe each skill precisely. "Summarize English text into bullet points" is better than "text processing"
- Version your cards — Use semantic versioning so consumers can pin to stable versions
- Define schemas — Explicit input/output schemas prevent runtime errors
- Monitor usage — Track which agents are calling yours and optimize accordingly
The Bigger Picture
AgentCards transform AI agents from isolated tools into a connected ecosystem. As more agents publish their capabilities, the network effect grows — each new agent makes the entire ecosystem more powerful. This is the foundation of the agent economy.