A technical guide to agent-to-agent (A2A) protocols — how independently built AI agents advertise their capabilities, discover each other, negotiate a task and delegate work with an auditable trail, without needing a shared codebase or vendor.
An agent cannot negotiate a task with another agent it does not know exists, or whose capabilities it cannot verify. A2A protocols solve discovery first — a standard capability card an agent publishes describing what it can do, what it costs and how to authenticate to it — so that negotiation and delegation have something real to build on instead of a hardcoded integration between two teams that happened to talk to each other.