Agent2Agent (A2A): The Protocol for Collaborative AI. Via: Dmytro Filatov.
Google just launched the Agent2Agent Protocol – a foundational layer for agentic interoperability, co-developed with over 50 global tech and services partners including Atlassian, Salesforce, Cohere, SAP, and McKinsey & Company.
1/ Why This Matters:
We’re entering a world where single-agent workflows are no longer enough.
Businesses are deploying fleets of specialized AI agents: some handle hiring, others optimize supply chains, respond to customers, or manage complex infrastructure. With the challenge of Cross-agent collaboration. Until now, most agents were locked into proprietary systems.
2/ A2A enables:
- Interoperability across vendors and frameworks
- Secure, modular communication between agents
- Task-based collaboration, including long-running workflows
- Modality-agnostic support — from text to video
- Capability discovery via JSON “Agent Cards”
3/ What This Unlocks:
- Dynamic multi-agent ecosystems
- Composable intelligence
- A world where you can mix-and-match agents like LEGO blocks — not only within your org, but across ecosystems.
This complements efforts like Anthropic’s Model Context Protocol (MCP) which was was recently integrated by OpenAI, offering rich memory and context sharing, while A2A enables real-time coordination and execution.
4/ Real Example:
Imagine your hiring agent requests candidates from a specialized sourcing agent. The candidate is vetted by another agent. Background checks are run automatically. Then onboarding begins — all without a human ever coordinating between systems.
🚀 #A2A is live on GitHub: For anyone looking to experiment or build — the full A2A Protocol is open source (Apache 2.0) and ready for action:
What’s inside:
✅ Agent Cards for discovery and negotiation
✅ Task lifecycle (submitted → working → input-required → completed)
✅ Streaming + Push Notifications for long-running agent tasks
✅ JSON-RPC, SSE, HTTP transport-friendly
✅ Integration-ready samples in Python & JS
✅ Supported by early frameworks: LangGraph, CrewAI, Genkit
World is Building toward intelligent networks, not just individual tools. Corporates will be moving towards symbiotic networks of AI, human agents and corporate resources interconnected via the likes of A2A and MCP protocols.