Agent Explorer

Built-in Mock Agent

Agent Explorer includes a simple built-in mock agent for quick testing without any setup.

Testing a mock agent in the Chat tab

Setup

Create a project with this URL:

http://localhost:3000/api/mock-agent

The mock agent provides:

  • A valid AgentCard at /api/mock-agent/.well-known/agent-card.json
  • Message sending with A2A 1.0 SendMessage and legacy 0.3 message/send
  • Synthetic get/cancel task responses for quick client testing
  • Compatibility with older sample callers: omitted message IDs are generated automatically, and message/stream remains an alias that returns one JSON response (not an SSE stream)

AgentCard Capabilities

The built-in mock agent advertises:

  • Streaming: Disabled
  • Push Notifications: Disabled

Special Commands

Send these text commands to trigger specific behaviours:

CommandBehaviour
delay NAdd N seconds delay (max 10s). E.g., delay 3
status STATESet task state. E.g., status working
simulate errorReturn a simulated error response
Any other textReturn Echo: followed by the message text

Supported Status States

status working, status completed, status failed, status input-required, status canceled, status submitted

Endpoints

EndpointMethodDescription
/api/mock-agentPOSTSend a message
/api/mock-agent/.well-known/agent-card.jsonGETFetch AgentCard

Limitations

The built-in mock agent is for quick testing only:

  • No real AI processing
  • No task storage; get/cancel responses simulate states without retrieving earlier messages
  • No streaming or push notifications; unsupported methods return a method-not-found error
  • Limited simulation scenarios
  • Single shared agent for all users

For more advanced scenarios, use Custom Mock Agents which support visual flow logic, skill routing, LLM responses, and per-user configuration.