a2a-agent-networking

To expose a local agent as an A2A JSON-RPC server, use the A2A builder with the withA2A() function to start an HTTP server and publish an Agent Card for discoverability. This enables other agents to discover and collaborate with your agent across distributed environments.

A lead agent can discover and connect to remote specialists across URLs using client discovery and capability matching APIs. By invoking the withRemoteAgent() function, the lead agent registers remote agents and communicates with them based on their published Agent Cards.

An Agent Card is a metadata payload containing an agent's name, description, and skills published at a well-known endpoint. It enables capability-based routing and inter-agent discovery across networks, allowing client APIs to match available specialists to specific tasks.

Yes, you can build scalable inter-agent communication across distributed environments without external dependencies. The skill provides native A2A server building and client connection APIs to establish HTTP servers and execute JSON-RPC tasks independently.

The best way to route tasks to specialized agents is by generating Agent Cards with specific skills for capability-based routing. Lead agents then use client discovery APIs to match task requirements against remote specialists' published capabilities.

You should not use A2A agent networking if your reactive agents operate within a single closed runtime where direct function calls are possible. This skill is designed for distributed environments requiring HTTP-based JSON-RPC servers and remote agent discovery across URLs.