Definition
Agent-to-Agent, or A2A, is an open protocol that lets autonomous AI agents communicate and coordinate even when they were built by different teams on different stacks. It was published by Google in April 2025 under an Apache-2.0 license and is now governed by the Linux Foundation. Where a tool schema lets one agent call a function, A2A lets one agent delegate a whole task to another agent it has never met, treating that agent as an opaque service rather than a function to micromanage.
How agents find and trust each other
A2A is built on familiar web technology: HTTP, JSON-RPC 2.0, and standard authentication. Discovery works through an Agent Card, a machine-readable document published by each agent that describes its capabilities, roles, and identity, so a requesting agent can decide whom to delegate to. Tasks then move through a defined lifecycle of states, and security is handled with established mechanisms such as OAuth 2.0, API keys, and mutual TLS. The design deliberately keeps each agent's internals private, exposing only what is needed to collaborate.
Relevance to a decentralized stack
A2A is positioned as an interoperability layer for multi-agent systems that span organizational boundaries. For a sovereignty-minded operator the appeal is open governance and no vendor lock-in: agents you run can interoperate with others on equal footing. As with any networked delegation, the open questions are authentication hardening and resistance to prompt-injection across the boundary, which remain active areas of security research.
A2A complements rather than replaces tool calling defined by a tool schema, and it provides the coordination fabric that multi-step agentic loops can run over.
In Simple Terms
Agent-to-Agent, or A2A, is an open protocol that lets autonomous AI agents communicate and coordinate even when they were built by different teams on different…