Agent2Agent (A2A)
An open protocol originated by Google in April 2025 for interoperability between AI agents — letting agents from different vendors and frameworks discover each other, advertise capabilities, and collaborate. Complementary to MCP, which connects agents to tools.
Agent2Agent (A2A) is an open protocol for agent-to-agent interoperability. Where MCP connects a single agent to tools and data (the vertical axis), A2A lets separate autonomous agents discover one another, advertise capabilities, and collaborate on tasks regardless of framework (LangGraph, CrewAI, ADK) or vendor (the horizontal axis). Google Cloud announced it in April 2025 with 50+ launch partners.
Core concepts
- Agent Card — a machine-readable JSON descriptor, typically served at a well-known URL, that publishes an agent's capabilities, input/output modalities, endpoint, and auth requirements. This is the discovery mechanism, analogous to a service descriptor in microservices.
- Tasks — units of work with a lifecycle (
submitted → working → input-required → completed / canceled / failed) carrying structured payloads of text, files, or data. - Messages and Parts — communication turns between a client agent and a remote agent, each with a role and one or more typed Parts.
- Artifacts — the outputs a task produces.
Transport is HTTP plus JSON-RPC 2.0, supporting synchronous request/response, Server-Sent Events for streaming long-running tasks, and push notifications/webhooks for async work.
On June 23, 2025, at Open Source Summit North America, Google announced donating A2A to the Linux Foundation, transferring the spec, SDKs, and tooling to a vendor-neutral project. Supporting members include AWS, Cisco, Google, Microsoft, Salesforce, SAP, and ServiceNow, with 100+ companies backing it. As of mid-2026 it is the most prominent agent-to-agent standard, though the spec is still maturing.
Naming discipline: this A2A is the AI agent protocol. It is not Yamaha/AV "A2A" receiver linking, and not "account-to-account" (A2A) payments in fintech — same acronym, unrelated domains. A2A is also distinct from the overloaded "ACP" acronym: IBM/BeeAI's Agent Communication Protocol (agent-to-agent, a peer of A2A), Zed's Agent Client Protocol (editor-to-agent), and the OpenAI/Stripe Agentic Commerce Protocol (agent-to-merchant payments).
A typical stack uses MCP to call tools and A2A to delegate to other agents; Google's docs explicitly frame the two as complementary. For the full landscape see AI Agent Protocols: MCP, A2A, and ACP and What Is the Model Context Protocol.
Related Articles
View all articles
AI Agent Protocols Explained: MCP vs A2A vs ACP and the Agent Interoperability Stack
MCP and A2A are not rivals — they are complementary layers of the same stack: MCP connects an agent to tools and data, A2A connects agents to each other. Here is the whole interoperability landscape, with ACP, ANP, and AGNTCY put in their place.
Read article →
What Is the Model Context Protocol (MCP)? The USB-C Port for AI, Explained
Model Context Protocol (MCP) is the open standard — created by Anthropic in late 2024 — that lets any AI application connect to tools, data, and prompts through one uniform wire protocol. Here's how the host/client/server architecture, primitives, and transports actually work.
Read article →
MCP Security Risks: A Practical Threat Model for Teams Connecting AI Agents to Tools
MCP isn't uniquely unsafe, but every server you connect widens your attack surface. A risk catalogue, the trust model you're actually accepting, and the governance controls MSPs and security teams should put in place.
Read article →
AI Gateway Guide: What They Are, Why You Need One, and How to Choose
A comprehensive guide to AI gateways — the proxy layer between your app and LLM providers. Compare Cloudflare AI Gateway, Portkey, Helicone, LiteLLM, AWS Bedrock, Azure APIM, and more across pricing, features, and architecture.
Read article →Explore More AI Agents
View all termsAgentic AI
AI systems that autonomously plan and take multi-step actions toward a goal — invoking tools, making decisions, and adapting — rather than producing a single one-shot response.
Read more →AI Agent
An LLM-driven system that perceives context, decides, and acts via tools and APIs in a loop to accomplish tasks — often with memory and some degree of autonomy.
Read more →Function Calling (Tool Use)
A capability where an LLM outputs a structured call (tool name plus JSON arguments) matching a developer-supplied JSON-Schema tool definition, which the host application executes and feeds back to the model.
Read more →Model Context Protocol (MCP)
An open, JSON-RPC–based client-server standard introduced by Anthropic in November 2024 that lets AI applications discover and connect to external tools, data sources, and prompts in a uniform way — "USB-C for AI integrations."
Read more →