How MCP and A2A Expand the Agency of AI Agents Feat. Yogin Patel

Understanding MCP & Google’s A2A and how they enable intelligent agent integration across the enterprise

MCP: Three letters that could redefine how enterprises approach AI in the coming years.

As more and more companies rush to adopt generative AI, there’s a growing realization that plugging in a model isn’t enough. What’s really needed is a smarter way to give these models the right context, guardrails and data. But traditional methods — custom APIs and connectors — are too rigid and resource-intensive. MCP or Model Context Protocol, on the other hand, is a smarter way for AI agents and SaaS tools to work together seamlessly, making collaboration across systems easier and more efficient.

It’s still a fairly new concept, and if you’re wondering what it actually means in practice, you’re not alone. We spoke to Yoginkumar Patel, VP – Engineering, Sprinklr, who revealed everything you need to know about MCP and more.

Here are a few excerpts from the interview:

What is MCP in layman's terms?

Imagine you had a toy robot, and this toy robot cannot do much as of today. And now you want this toy robot to paint a picture for you and use a paint brush and let’s say a brand like 3M sells paint brushes. You want this toy robot that came from Lego to use 3M's paintbrush. Okay, now, one thing would be that both 3M and Lego would collaborate and their systems would integrate that they will build such paint brushes that'll fit into this Lego toy robot. That would be one way to do that.

Another way is to imagine there is a new toy box which is universal. Any toy robot can plug into this toy robot box. And tools such as paint brushes or any other tool that come from any other providers also fit into this box. And this box is seamless, it's universal, it's like an adapter.

So every toy robot company will now make such toy robots that can be plugged into this existing toy robot box. And every tool-making company that is building paint brushes will also make sure that they build those paint brushes such that they will fit into this toy robot box. So this toy robot box is analogical to the MCP protocol and toy robot is analogical to the AI agent. And the paint brushes are the tools that we want the AI agent to use.

Some people compare MCP to what APIs were for the internet. Is that a fair comparison?

That is fair but APIs have their own challenges. So, APIs help two software solutions share data with each other and with APIs, you need to have hardcoded data structures to be understood and contracted between those two different systems. So that's why we had something called system integrators in between to integrate one system with other systems.

And those system integrations usually took days and weeks and majority of the time went into field mapping exercises in the data structure management exercises. That field mapping part is now being easily handled by LLMs, so we don't have to explicitly map those fields so you can wrap any APIs into MCPs, and MCPs help you scale those system integrations in a sustainable way.

How does MCP change how businesses interact with their customers and what's at stake for companies who ignore the shift?

MCP will help a company's internal knowledge and data to get exposed to and utilized by AI. So, if you're already in a transformation journey, I think the fastest you can achieve that is by MCP.

So I think it would make sense for brands to have MCP as part of their strategy, and the reason to do that in fact is so that once you do that, any AI agent can now interact with that knowledge and data. So all the workflow orchestrations that you are trying to do using that knowledge can be made possible using different vendors or tools or AI agents that you might be using in your company.

If you’re wondering how MCP can benefit your organization, let’s look at a hypothetical example that will make it clear how MCP adds value to an existing workflow. 

How MCP and A2A drive AI-powered workflow automation in healthcare

Let’s consider a healthcare enterprise deploying a virtual assistant to streamline patient services. A user says:

“I need to book a cardiology appointment for next week and have my ECG sent to the doctor.”

At first glance, this may seem like a simple request — but on the backend, it demands orchestrated interactions across multiple enterprise systems, data sources and AI capabilities.

MCP: The interface layer for enterprise systems

This virtual assistant (an AI application) must:

• Check scheduling systems for availability
• Query location-based doctor directories
• Retrieve medical records from an EHR platform
• Transmit files securely to a third-party system

MCP enables this AI assistant to plug into all these systems using a shared, structured protocol. "MCP is a way to contextualize your AI agents, and in order to do that, you use this protocol,” Patel says. Instead of building custom one-off integrations, it uses standard "tools," "resources" and "prompts" defined by the MCP server to perform functions such as:

• Query available time slots
• Access files
• Interact with secure messaging APIs

MCP essentially reduces integration overhead by standardizing how AI agents access enterprise tools, accelerating time-to-value and lowering cost of deployment across departments.

Google’s A2A: The coordination layer for collaboration among multiple AI agents

Behind the scenes, this single request triggers a multi-agent workflow, including:

• A Records Agent accessing the EHR
• Provider Match Agent surfacing relevant specialists
• Scheduling Agent booking the appointment
• Compliance Agent handling secure document transmission

Each agent has a specific skill set. A2A lets them:

• Discover one another’s capabilities (using Agent Cards*)
• Share the task and updates in real time (via structured Messages and Artifacts**)
• Work in sequence or in parallel, while tracking progress

A2A supports modular, specialized AI capabilities that can be mixed and matched across workflows — creating a scalable architecture for intelligent process automation.

*An Agent Card is a public metadata file describing an agent’s capabilites

**An Artifact is structured result as the output of a task

End result

In under a minute:

• Appointment is confirmed
• Medical data is routed securely
• Notifications and follow-ups are triggered (with no human intervention)

Why this matters

• MCP lets each agent do something in the outside world (book, fetch, send)
• A2A lets multiple agents work together to complete the full journey

Together, they transform siloed AI apps into a collaborative, context-aware ecosystem — whether it's planning a vacation or managing your health.

Now, let’s see how MCP and A2A help you manage crises better.

How MCP and A2A simplify crisis communication

Here’s the scenario: A multinational company faces a sudden product recall. The brand uses an AI-driven digital operations platform that coordinates crisis communication across channels (social, email, web, support), all in real time.

Here’s how MCP and A2A work in tandem to diffuse the impact and restore customer trust.

If this blog has piqued your curiosity about MCP and you'd like to know it connects siloed systems, unlocks smarter workflows and helps brands deliver proactive, personalized CX at scale, then give our podcast featuring Yogin Patel a listen. You can also catch it on Spotify!