A2A Joins the Agentic AI Foundation’s Open Agentic Stack
At AGNTCon + MCPCon Europe, the Agentic AI Foundation announced that Agent2Agent (A2A) — the open standard for inter-agent communication — is joining as a hosted project. This fills the critical missing piece in the agentic stack: while MCP solves agent-to-tool connectivity, A2A solves agent-to-agent communication.
The Agentic Stack, Now Complete
The open agentic stack has been taking shape over the past year:
- MCP (Model Context Protocol) — agent-to-tool connectivity — 89% adoption
- Skills — 69% adoption
- A2A (Agent2Agent) — agent-to-agent communication — now part of AAIF
- AGENTS.md — agent instructions and configuration — 51% adoption
- AGNTCY — full agent lifecycle management — 37% adoption
With A2A joining the AAIF as a hosted project, the foundation now stewards two of the three core protocols that make up the agentic stack: MCP for agent-to-tool and A2A for agent-to-agent.
What A2A Solves
Agent2Agent addresses a fundamental scalability problem: when you have multiple agents working together, how do they communicate?
- Coordination — How does an orchestrator agent delegate tasks to specialist agents?
- Handoff — How does one agent pass context and state to another?
- Negotiation — How do agents with different capabilities negotiate a shared plan?
While MCP provides a standard way for an agent to call a tool (database query, API call, file read), A2A provides a standard way for one agent to invoke another agent as a peer, with structured request/response semantics.
Why This Matters for Production
At AGNTCon, the theme was clear: agentic AI has moved from experimentation to production (81% of teams report production deployment). And production agentic systems are multi-agent by default (60% of teams run multi-agent systems, with 50% across more than one stack).
Without a standard for agent-to-agent communication, each platform builds its own proprietary protocol — creating vendor lock-in and fragmentation at exactly the layer where interoperability matters most.
Integration with MCP
Both MCP and A2A are now stewarded under the AAIF umbrella, creating a unified governance model. The two protocols are designed to complement each other:
- MCP handles: agent → tool (databases, APIs, services)
- A2A handles: agent → agent (specialist agents, orchestrators, subagents)
Teams building agentic systems can use MCP to connect agents to their tools and data, then use A2A to enable those agents to collaborate with each other.
