MCP Server: What You Need to Know

Financial firms are moving fast on artificial intelligence, but most are still solving the same underlying problem: getting trusted data into the AI tools their teams already use, without building a custom connection for every pairing of model and system.  

The Model Context Protocol, or MCP, was created to close that gap. This open standard has been adopted quickly across the AI industry since its introduction in 2024. 

The following blog post lays out the basics of MCP and answers the questions heads of data and AI strategy as well as asset managers, wealth firms, and other investment industry leaders ask most as they evaluate MCP server deployment for their own organizations.  

The blog post also explains how the Morningstar MCP Server applies this efficiency for investment data. 

MCP is a program that gives AI applications a consistent way to connect to the systems where data lives, including content repositories, business tools, and development environments. Anthropic introduced the open standard in November 2024. 

MCP is the layer that connects an AI agent or assistant to data and tools. An MCP server works alongside the agents and assistants firms already use rather than in place of them. 

Instead of creating a custom-built connection to every data source for every AI tool, developers can expose data through an MCP server once, and any MCP-compatible AI application can connect to it. 

MCP servers vary widely by what they expose, from general-purpose connectors for tools like databases and file systems to purpose-built servers like Morningstar's, which is scoped specifically to governed investment data and research. 

The Morningstar MCP Server grounds AI applications like Claude, ChatGPT, Copilot, and others in Morningstar's investment data and independent research.  

In practice, that means governed, AI-ready data and analytics are available directly inside the tools an investment team already uses, rather than in a separate system they have to switch into. 

An AI model is only as reliable as the data source behind it. Morningstar built its MCP server to connect AI systems to Morningstar's data and independent research rather than to a generic, unverified source.

MCP does not replace the APIs financial firms already rely on. Every MCP server still has a conventional API doing the underlying work of authentication, data retrieval, and business logic behind it. What MCP adds is a common discovery and invocation layer on top, so an AI system can find and use those capabilities without a bespoke integration for every tool. 

A traditional API assumes a developer already knows exactly which endpoint to call and how to format the request. MCP assumes an AI agent has to work that out for itself, so it adds a standardized layer for discovering, describing, and invoking tools. 

That distinction matters at scale. Connecting a growing number of AI models to a growing number of data sources through custom, one-off integrations can create a tangle of connections that becomes harder to maintain with every new pairing.

Security matters for any protocol that hands AI systems new ways to reach live data sources. MCP servers don't just expose functionality the way a typical API does. The descriptions attached to their tools can also carry instructions that quietly steer an agent's behavior, a risk known as tool poisoning. Tools to catch this poisoning are improving, but even their creators say the results should guide a human, not replace one.  

That's why human-in-the-loop governance has to be part of any MCP server deployment.  

Morningstar's approach is built around secure access limited to the data and research each user is licensed and permitted to see. It’s designed to work within a firm's existing security controls, so firms can extend AI into their workflows without giving up the oversight their compliance functions require. 

Asset and wealth managers, advisors, and institutional investors are the clearest fits for the Morningstar MCP Server. It helps asset and wealth managers accelerate research and portfolio analysis, gives advisors AI-powered client insights at scale, and lets institutional investors fold Morningstar's data into their own custom AI workflows for compliance, risk, and strategy.  

Akambo, a Melbourne-based investment management firm, is one example. As one of Morningstar's first MCP clients globally, the firm connected its existing Claude-based workflows directly to Morningstar data without building custom infrastructure. Its team now runs screens across more than 5,300 Australian-domiciled funds and queries hundreds of data points per security conversationally. 

The MCP Server's entity detection, data extraction, and analyst research functions map closely onto Akambo's existing fund screening, manager due diligence, and client reporting workflows.

MCP was designed to be quick to stand up so organizations can connect important datasets to AI tools without a lengthy development cycle.  

The broader ecosystem has matured quickly to support that. MCP now has official software development kits across every major programming language.  

The Morningstar MCP Server already connects to tools financial professionals use, including Claude, Microsoft 365 Copilot, Copilot Studio, Mistral, Perplexity, and ChatGPT, with more integrations added on an ongoing basis. 

The MCP server unlocks operational efficiencies for investment workflows. Before adopting the Morningstar MCP Server, Akambo's team had to export data, reformat it, and feed it into their AI tool as separate steps; with MCP, those steps are consolidated into a single conversational session, and fund screening and manager due diligence that once required navigating multiple platforms can now happen in one query.

MCP server deployment does not require an engineering overhaul. For most financial services firms, it means connecting AI tools your teams already trust to data they already rely on, with the governance your compliance function expects built in from the start.  

Explore the Morningstar MCP Server to see how governed, AI-ready data can fit into your firm's workflows.