Capabilities

Agent capabilities define what features an agent supports.

Core Capabilities

Streaming

When streaming: true, the agent supports message/stream for incremental SSE responses.

Benefits: real-time feedback, lower perceived latency, progressive rendering.

Push Notifications

When pushNotifications: true, the agent can send updates to a callback URL for asynchronous workflows and long-running background tasks.

Agent Explorer has built-in support for configuring, managing, and observing push notification webhooks. See Push Notifications for full details.

State Transition History

When stateTransitionHistory: true, task responses include the full history of state changes — useful for debugging and auditing.

Extended AgentCard

When extendedAgentCard: true, the agent supports an extended AgentCard format with additional metadata.

Skills

Skills describe what an agent can do:

{
  "skills": [
    {
      "id": "code-review",
      "name": "Code Review",
      "description": "Reviews code for bugs and improvements",
      "tags": ["code", "review", "analysis"],
      "inputModes": ["text/plain"],
      "outputModes": ["text/plain"]
    },
    {
      "id": "summarize",
      "name": "Document Summarization",
      "description": "Summarizes long documents"
    }
  ]
}

Authentication

Agents may require authentication:

{
  "authentication": {
    "schemes": ["oauth2", "apiKey"]
  }
}

Agent Explorer supports OAuth 2.0 authentication flows for secured agents. See OAuth Authentication for setup details.

Input/Output Modes

Agents declare the MIME types they accept and produce:

{
  "defaultInputModes": ["text/plain", "application/json"],
  "defaultOutputModes": ["text/plain"]
}