Back to AI Radar
EmergingAgents & OrchestrationNew entryMarch 2026

Interesting and early. Worth a spike or exploration session.

A2A Protocol

A2A is the missing piece alongside MCP — if MCP is how agents talk to tools, A2A is how agents talk to each other.

Agentic·Infrastructure

github.com

Our Take

What It Is

The Agent-to-Agent (A2A) protocol is an open standard created by Google for enabling AI agents to discover, communicate with, and delegate tasks to other agents — regardless of the framework or vendor that built them. While MCP handles agent-to-tool connectivity, A2A handles agent-to-agent coordination. The protocol is now governed by the Linux Foundation alongside MCP, with 50+ partners including Salesforce, SAP, ServiceNow, and MongoDB. V0.3 added gRPC transport support for high-performance communication.

Why It Matters

A2A enters the radar at Emerging because it fills a clear architectural gap. As multi-agent systems move to Promising, the question of how agents from different teams, frameworks, and vendors communicate with each other becomes practical, not theoretical. A2A provides the answer: a standard protocol with agent discovery (Agent Cards), task management, streaming communication, and push notifications.

The 50+ partner list is significant because it includes enterprise platforms (Salesforce, SAP, ServiceNow) where multi-vendor agent coordination is a real deployment requirement, not an academic exercise. An agent built with LangGraph needs to coordinate with an agent built on CrewAI — A2A makes that possible without custom integration work.

Key Developments

  • Mar 2026: A2A donated to Linux Foundation alongside MCP, establishing shared governance for the complete agent connectivity stack.
  • Feb 2026: V0.3 released with gRPC transport support for high-performance agent-to-agent communication.
  • Jan 2026: 50+ partners announced including Salesforce, SAP, ServiceNow, and MongoDB.
  • Dec 2025: Google open-sources A2A protocol with initial HTTP-based transport and Agent Card discovery mechanism.

What to Watch

The MCP + A2A combination could become the complete connectivity standard for agentic systems. Watch for real-world deployments where agents from different vendors actually communicate via A2A in production — the partner list is strong but production usage is what moves this to Promising. Also track whether the protocol handles security, authentication, and trust between agents from different organisations. Enterprise adoption depends on this.

Strengths

  • Architectural fit: Complements MCP perfectly — together they cover agent-to-tool and agent-to-agent communication as a complete standard.
  • Enterprise partners: 50+ partners including Salesforce, SAP, and ServiceNow signal real enterprise demand for cross-vendor agent coordination.
  • Open governance: Linux Foundation governance alongside MCP ensures vendor-neutral evolution of both protocols.
  • Performance options: Both HTTP and gRPC transport support means teams can choose latency-cost trade-offs appropriate to their use case.

Considerations

  • Early maturity: V0.3 is still pre-1.0. The protocol specification is evolving and breaking changes are possible.
  • Limited production evidence: Strong partner list, but documented production deployments using A2A are still scarce.
  • Security gaps: Cross-organisation agent communication raises trust, authentication, and authorisation questions that the protocol is still addressing.
  • Adoption dependency: A2A's value scales with adoption. If major frameworks don't implement support, it becomes another unused standard.

Resources

More in Agents & Orchestration

A2A Protocol· OpenAI Agents SDK· PydanticAI· AI Browser Use· Agentic RAG· CrewAI· Multi-agent Orchestration· OpenClaw· Chain-of-Thought· LangGraph· Model Context Protocol· Tool Use / Function Calling

Back to AI Radar