Microsoft joins Google's A2A party. AgenticAI future gets a lot more real!
This Week in Products, we explore how MCP and A2A protocols make AI truly agentic and what it means for the future of products we build.
The Big Shift: A2A is the Missing Piece in the Agentic AI Puzzle
A couple of weeks ago, I had written that AI’s next big leap isn’t a model. It’s a protocol.
Back then, we spoke about MCP – the Model Context Protocol. A quiet but powerful plug-and-play protocol that lets your AI seamlessly access, understand, and act on live business context. MCP enables two-way connections between AI agents and your real-world business systems: CRMs, calendars, internal tools, knowledge bases, and more.
OpenAI, Google, and Anthropic have all adopted MCP as a common protocol.
This week, another piece of the puzzle shaping the future of agentic AI has clicked into place: Microsoft has embraced Google’s Agent2Agent (A2A) protocol - an open standard to connect one AI agent to another.
If MCP gave agents the ability to act in the real world with real business systems, A2A gives them teammates.
This Week in Products, we explore what these developments mean.
But, before we dive in…
🤖 What are AI Agents?
Unlike traditional AI tools that respond to a single prompt or command (like ChatGPT that you use every day), AI agents are designed to act autonomously, proactively, and often collaboratively.
Think of them as intelligent digital workers, not just answering questions, but taking initiative, handling tasks, and making decisions on your behalf.
They can:
Plan multi-step tasks (like booking travel, summarizing reports, or managing emails),
Use tools and APIs to get things done (e.g., send messages, run code, access data),
Learn and adapt over time based on context, feedback, and outcomes.
These agents aren’t just chatbots, they’re the early scaffolding of a future where software doesn’t just serve you, it works with you.
🧐How does the A2A protocol help AI Agents?
With the A2A protocol, suddenly, we're not talking about a single AI agent fetching data or responding to prompts.
We're talking about networks of agents, each with their own purpose, collaborating to get complex jobs done across tools, clouds, and contexts.
Let’s decode what’s actually happening under the hood:
🧩 A2A ≠ Just a Chat Between Bots
This isn’t “Claude talking to ChatGPT.” This is about agents with defined roles like sales agents, finance agents, scheduling agents, collaborating like departments in a company.
🎯 Structured Communication, Not Prompt Soup
A2A defines how agents share goals, maintain state, invoke actions, and return results, all in a secure, observable way.
🌍 Cross-Cloud, Cross-Vendor
With A2A, agents built on OpenAI can talk to those running on Gemini, Claude, or anything else. Microsoft, Google, and Anthropic are making sure you’re not locked into one provider’s ecosystem.
🧱 MCP and A2A Together = Agent Stack
MCP lets agents plug into business tools. A2A lets agents plug into each other. Together, they’re becoming the protocol layer for the agentic web.
🧠 It’s Not Just About AI Doing Things
It’s about AI reasoning collectively across boundaries. Think less Jarvis. More... mini-org chart of specialists, each handling a part of a process.
Driving the News
Microsoft CEO Satya Nadella calls A2A and MCP "key to enabling the agentic web.”
Big words. But, totally justified.
Copilot Studio and Azure AI Foundry have also announced support for A2A, allowing developers to build agentic systems where agents can work together, not just alone.
The protocol is open, hosted on GitHub, and already being shaped by working groups involving the biggest names in AI.
Charles Lamanna (Microsoft Cloud VP): “We are entering a new era in business where agents will not only act independently but also work together as a team.”
🔍Reading Between the Lines
Microsoft isn’t just dabbling here with A2A.
Over 10,000 orgs are building agentic systems via their new Agent Service.
Copilot Studio is already in use by 90% of the Fortune 500.
This isn’t a side project for Microsoft. With Google and Microsoft (two of the largest tech companies on earth) adopting A2A, this is the foundation of what’s next for the future of AgenticAI as AI agents get a standard protocol to talk to each other.
🌍 The Big Picture
MCP gave AI the hands to act. A2A gives it the ability to coordinate/collaborate
Together, they’re building the foundation for an agentic web, a shift as fundamental as the browser or the app store.
And if you’re building for the future, this is your stack:
MCP to access real-world tools and data.
A2A to collaborate with other agents.
Open standards to avoid lock-in.
Composable design to scale with elegance.
If you're a product founder or leader, here’s how to get ahead:
Understand Your Product’s Role in the Agentic Stack: Can your product expose data or actions via MCP? Can it host an agent that plugs into others via A2A?
Design Workflows, Not Just Features: Ask: What if multiple agents were solving this for the user instead of them clicking around?
Build with Composability in Mind: Let users bring their own agents. Let your agents talk to others. Become a node in a bigger system.
Start Experimenting with A2A: Join the GitHub spec. Test interoperability. Contribute ideas. These standards are still young. Now’s your chance to shape them.
📰What’s going around tech?
Funding Opportunities, Coding Tools, and New Architectures…
Google launches AI startup fund offering access to new models and tools
Google's AI Futures Fund supports startups using DeepMind’s AI tools, providing access to models, expert guidance, cloud credits, and potential funding.
Read More→
Amazon Is Developing an AI Tool to Generate Code and Assist Developers
Amazon is developing "Kiro," an AI tool that generates code in real-time, creates technical documents, and optimizes code, with web and desktop apps supporting multimodal inputs and third-party AI integration.
Read More→
Sakana Unveils 'Continuous Thought' Architecture to Help AI Reason Like Humans
Sakana AI's Continuous Thought Machines (CTM) is a new AI architecture that mimics human reasoning by processing data step-by-step using neuron-level timing and synchronization. Unlike traditional models, CTMs adaptively handle complex tasks like maze-solving with human-like attention patterns.
Read More→
Tech Leaders Urge U.S. to Boost AI Infrastructure
OpenAI, Microsoft, and AMD executives told the Senate that no single country can dominate AI, urging faster permitting for data centers and chip factories to keep the U.S. competitive with China. They emphasized the need for skilled workers and clear export policies to drive AI adoption globally.
Read More→
Microsoft Reveals Windows 11 Start Menu Redesign Concepts
Microsoft has revealed its alternative Start menu designs for Windows 11, exploring innovative layouts like phone integration and floating widgets, though the final redesign (publicly available version) prioritizes familiarity and faster app access.
Read More→
A video I found insightful
An Engineer’s “Taste” Matters as Much as Code
“Taste” is what we usually talk about when it comes to clothes, sneakers, food, or whatever you love to indulge in. But having a taste in engineering? That’s a perspective you don’t hear about often, one that feels refreshingly off the radar. Michael Truell, CEO of Cursor, brings it up in this video.
He dives into how the job of a software engineer is gradually evolving beyond just writing code, to where engineers will instead act more like logic designers, defining how software should behave, both in how it works and how it looks (this is where taste comes in).
He gets at a bigger shift where we might be moving toward a future where engineers won’t need to write out every line of code. Instead, they could just describe what they want in plain English, and the system would do the rest.
Watch the video for more insights
Chart Drop!
Insights for your next AI product!
📬I hope you enjoyed this week's curated stories and resources. Check your inbox again next week, or read previous editions of this newsletter for more insights. To get instant updates, connect with me on LinkedIn.
Cheers!
Khuze Siam
Founder: Siam Computing & ProdWrks