This is the second of a three-part series about Unlocking AI ROI. If you missed the first post you can read it here.
In this blog post we’ll explore how Model Context Protocol (MCP) can help you sustain the business benefits of your agentic apps by offering a structured approach for introducing new data sources. Don’t let your ROI goals slip away—read on!
What is MCP?
MCP is a framework for designing autonomous applications across languages, platforms and third party data sources. The MCP framework enables AI agents to maintain, retrieve, and apply relevant context from multiple enterprise systems, across interactions. This improves AI decision-making by ensuring context consistency, safe access to third party data sources, and compliance.
MCP was introduced by Anthropic in November 2024 and while it has not been made an official industry standard, its adoption is on a meteoric rise. Along with its support from AWS, Microsoft and Google it is time to explore more about this transformative innovation.
Enriching Agentic Applications with MCP
AI agents need to access data from multiple systems of record to build a context that can be used by an AI model in multi-step problem solving. MCP offers a standardized way to write the glue code to access the system of records in a way that AI agents can consume. Once an MCP server is developed it can be used by any AI agent that has an MCP client. For example, if you build an MCP server to access and update data in JIRA, then the JIRA MCP server can be used by all the agents that need to read and update issues. Without an MCP server every AI agent would need to write its own custom glue code to access JIRA.
The idea of having access to disparate systems is that the agent can solve problems with minimal human intervention. Previously, a basic chat bot application was trained on an organization’s data to answer questions from those data sources. But what happens if the end user wants to know something that is from data owned by a partner organization or government agency? For example, a third party shipping service that has contracted with e-commerce retailers will have its own system for customer shipping data. Unless the retailer is able to connect directly to that third party shipper’s data, either the customer will have to manually navigate to that shipping system or the retailer’s customer service agent will have to step in. Either way, not having access to partner data sources can limit the value of an autonomous agent. MCP creates value by offering secure access to trusted data sources to maintain the agent application’s autonomy. Here’s how MCP facilitates agentic ROI:
Improved Decision-Making Through Relevant Context
MCP ensures agentic applications can build context with the relevant information from multiple data sources for multi-step problem solving. For example, when a customer wants to return a damaged product to a retailer, the agent will need to send relevant customer data to an ordering system, a payment system and a shipping system in a continuous session. MCP enables the customer context to follow the agentic application. MCP also enables the application to securely connect those various systems. Meanwhile, developers can expose systems-of-record to AI agents through a standardized interface. This ultimately lets agentic software handle the most common and mundane tasks efficiently, freeing up personnel to focus on more complex problems.
Accelerate Business Agility with Efficient Use of AI Models and Personnel
MCP can streamline end-user interactions for better customer experiences that ultimately improve market competitiveness and brand affinity. However, agentic is not a personnel replacement strategy. Organizations we speak with are still keeping humans-in-the-loop while utilizing agentic applications to help their workforce be more agile. The enterprise still needs personnel, not just to deal with complex issues, but to also define and build the organizational policies that guide their agents while still allowing for a certain level of autonomous problem solving. In this way enterprises can actually include the collective knowledge and organizational expertise in an agentic workflow as just another data source connected by an MCP Server. In our customer return scenario, the agent could be given access to the organization’s return policy knowledge base through an MCP server to understand return eligibility and then appropriately trigger the right flow based on what it finds in the knowledge base without human intervention. This frees up personnel for more nuanced scenarios.
Build an AI Agent with Tanzu Platform and Model Context Protocol
The Tanzu team developed Spring AI MCP so organizations can evolve their current or future Spring applications to include agentic capabilities. Spring AI is a polyglot application framework for AI engineering that expands upon the Spring framework ecosystem. Applications built with Spring run seamlessly on VMware Tanzu Platform, our flagship AI-ready, private PaaS. Tanzu Platform can help organizations bolster their AI ROI by helping them more efficiently develop, operate and optimize agentic applications.Watch this video from Adib Saikali, Distinguished Engineer, Tanzu Division, Broadcom to understand how organizations might architect MCP clients and servers to perform autonomous thinking patterns.
Also be sure to check out Cloud Foundry Weekly vlog series which has multiple shows dedicated to AI application delivery.
If you are looking to build new AI skills, please consider attending our Cloud Foundry Day Platform Engineering Skills for GenAI and Agentic Training Workshop on May 13th being held on the VMware Campus in Palo Alto.
