The A2A Protocol (Agent2Agent) is an open protocol from Google that enables communication and interoperability between AI agents. Announced in April 2025, it lets agents from different vendors, built on different frameworks, discover each other, share capabilities, and delegate tasks across organizational boundaries — the standard for agent-to-agent communication.
Key Features
- Agent Discovery: Agent Cards (JSON) describe an agent's capabilities so other agents can find and understand them.
- Tasks: Stateful units of work with defined lifecycle — agents track, update, and complete tasks.
- Messages & Parts: Structured message exchange with content parts for rich data sharing.
- A2A API: HTTP + SSE based protocol for synchronous and asynchronous agent communication.
- Push Notifications: Real-time updates from agents as tasks progress.
- Vendor-Neutral: Works across frameworks, models, and vendors — no lock-in.
- Complements MCP: MCP gives agents tools; A2A lets agents talk to each other. The two work together.
- Open Governance: Donated to the Linux Foundation; 50+ partners including Microsoft, Salesforce, SAP, ServiceNow, LangChain, and CrewAI.
Use Cases
Who Should Use This Tool?
- Enterprise Architectures: Organizations building agent ecosystems across internal systems and vendors.
- Cross-Framework Integrators: Teams whose agents span LangChain, CrewAI, Google ADK, and other frameworks.
- Platform Providers: Companies exposing their agents to other systems through a standard protocol.
Problems It Solves
- Agent silos: A standard protocol lets agents from different vendors collaborate.
- Delegation across boundaries: Tasks can be handed to specialized agents anywhere.
- Framework lock-in: A2A decouples agents from any single framework.
Pricing
| Plan | Price | Features |
|---|---|---|
| Open Protocol | $0 | Free, open, and vendor-neutral; governed by the Linux Foundation. |
Advantages & Unique Selling Points
Compared to Competitors:
- Open standard: Not a proprietary SDK — an industry protocol.
- Complementary to MCP: Together they cover agent-to-tool and agent-to-agent communication.
- Industry backing: 50+ partners and Linux Foundation governance.
What Makes It Stand Out:
- Agent Cards make discovery a first-class concept.
- Stateful Tasks with push notifications for long-running work.
- Adopted by Google ADK, LangChain, CrewAI, and enterprise platforms.
Getting Started
Quick Start Guide
- Read the spec: Visit a2a-protocol.org for the protocol documentation.
- Expose an agent: Implement the A2A API on your agent (or use an SDK like Google ADK or LangChain A2A tools).
- Publish an Agent Card: Describe your agent's capabilities in JSON.
- Discover & delegate: Other agents find your card and delegate tasks to your agent.
Integration
Integrates with:
- Google ADK, LangChain, CrewAI, and other A2A-enabled frameworks
- MCP for tool access
- Enterprise platforms from Microsoft, Salesforce, SAP, ServiceNow
Frequently Asked Questions
How is A2A different from MCP?
MCP connects agents to tools and data; A2A connects agents to other agents. They're complementary, not competing.
Is A2A proprietary to Google?
No — it's an open protocol donated to the Linux Foundation with 50+ industry partners.
Do I need to rewrite my agents?
No — A2A is an interoperability layer; existing agents expose an A2A API without rewriting their core.
Alternatives
If A2A isn't the right fit, consider these alternatives:
- MCP Framework: For agent-to-tool integration (AI Ecosystem category).
- Google ADK: Google's framework with first-class A2A support.
- LangGraph: For stateful orchestration within your own system.
Tips & Best Practices
- Publish accurate Agent Cards: Discovery quality depends on honest capability descriptions.
- Pair with MCP: Use MCP for tools and A2A for inter-agent work.
- Design for long tasks: Use push notifications to track stateful, long-running tasks.
Conclusion
A2A is the open protocol that makes agent-to-agent communication a standard — discovery, task delegation, and collaboration across vendors, frameworks, and organizations. If you're building an agent ecosystem that must interoperate beyond your own stack, A2A is the industry answer.
Comments
No comments yet. Be the first to comment!
Related Tools
Google ADK
google.github.io/adk-docs
Google's open-source framework for building multi-agent AI applications. Hierarchical agent teams, shared state, code-native design, MCP support, live multimodal, and A2A interoperability — Python, Java, and Go.
Letta
www.letta.com
Open-source framework for building stateful AI agents with persistent memory. Born from the MemGPT paper — agents remember conversations indefinitely without blowing context windows, with sleep-time compute for memory processing.
OpenAI Agents SDK
openai.github.io/openai-agents-python
OpenAI's official framework for building production AI agents in Python and TypeScript. Agents, handoffs, guardrails, sessions, and built-in tracing — the successor to Swarm.
Related Insights
After I Connected Obsidian to OpenClaw, It Started Helping Me Make Decisions
Once Obsidian stopped being just a place to store notes and started working with OpenClaw, it began helping me organize context, connect information, and improve real decisions.
Stop Cramming AI Assistants into Chat Boxes: Clawdbot Picked the Wrong Battlefield
Clawdbot is convenient, but putting it inside Slack or Discord was the wrong design choice from day one. Chat tools are not for operating tasks, and AI isn't for chatting.
The Twilight of Low-Code Platforms: Why Claude Agent SDK Will Make Dify History
A deep dive from first principles of large language models on why Claude Agent SDK will replace Dify. Exploring why describing processes in natural language is more aligned with human primitive behavior patterns, and why this is the inevitable choice in the AI era.