Agent2Agent Protocol (A2A)

01What it means

The Agent2Agent Protocol (A2A) is an open standard that lets independent AI agents work with each other: one agent discovers another through its agent card, sends it a task over JSON-RPC, and receives structured results — regardless of which framework or vendor each side is built on. The division of labor with MCP is clean: MCP connects an agent to tools and data; A2A connects agents to agents. If MCP standardizes how an agent uses a tool, A2A standardizes how an agent hires another agent.

This site runs a small live example: an A2A agent at /api/a2a that accepts message/send requests and answers questions about our services, case studies, and writing, using the same nine read-only tools as our MCP server. Its card sits at /.well-known/agent-card.json, so any A2A-capable client can find it without being told.

The protocol matters because agent ecosystems are forming across vendor boundaries — an agent from one stack delegating research or execution to an agent from another. Open inter-agent protocols keep that market composable instead of consolidating into a handful of walled gardens.

02Why it matters for buyers

• —Cross-vendor delegation without bespoke integrations: if your agents and a supplier's agents both speak A2A, coordination is a protocol matter, not a project.
• —Discovery is standardized: agents read each other's cards before any work is sent, so capabilities and terms are stated in a file, not negotiated in prose each time.
• —It is early — and that is the signal: publishing an A2A surface today is a statement that agent-mediated business is welcome, which is exactly what our readiness tool checks for.

03Common questions