Financial institutions have begun deploying Model Context Protocols (MCPs) as they have expanded the use of artificial intelligence applications and agents. The technology developed by Anthropic is an open-source contextual layer that helps coordinate models and data, enabling AI applications to connect with a multitude of other platforms and processes.
In the first of a series of articles on their use by capital markets participants, Data Management Insight spoke to Harald Collet, chief executive at Alkymi, about how MCPs are empowering private markets participants and, crucially, their shortcomings.Data Management Insight: How are MCPs benefiting financial markets participants?
Harald Collet: One of the great things about the MCP is that it provides a uniform language for different AI platforms and processes to communicate. At a high level, MCP provides a standardised way for AI systems, applications, and data sources to communicate. It enables models to access context, invoke tools, and exchange information across systems without requiring custom integrations for every connection.
DMI: What are MCPs making redundant?
HC: MCP has the potential to reduce much of the custom integration work traditionally required to connect AI systems and workflows. Rather than building one-off connections between every application, organizations can leverage a common protocol that simplifies how AI agents discover, access, and interact with systems.
Think back to the days of robotic process automation. It was going to free up and automate a bunch of processes. But one of the things you ended up with when you were doing robotic process automation is an enormous amount of integration headaches, like making sure that the robots could talk to each other, that you orchestrated the robots, and so on.
MCP helps address some of those challenges by providing a standardised way for agents to access tools and services across systems. A user can generate a prompt and get it with context and have that sent to a third-party system that can then orchestrate an output and send it back into that environment.
The quality of outcomes depends on the quality of the underlying data, governance, and workflows. Organizations still need trusted operational systems capable of providing validated information to those agents.
So you have a much more complex process, which is really more in the world of process engineering, where you’re starting to think about if I had a blank canvas, what could I put together to make my specific process better? We’re starting to do some of that with our business partners. That’s where organizations can move beyond simple use cases and begin orchestrating more complex workflows across multiple systems and agents.DMI: It sounds like an ideal solution for lots of pain points within private markets.
HC: In private markets, alternatives and private credit workflows are uniquely complex and document-heavy. With AI agents, Alkymi now brings structure and real-time intelligence directly into these workflows for data sets such as transaction notices, capital account statements, schedules of investment, loan agent notices, compliance certificates, and financial statements.
Because the data is unstructured, it’s fragmented, it oftentimes requires deeper insight, combining multiple data sets. So if you’re looking to monitor your private credit deals, you might want to know the cash flows. For private credit deals, these are called loan agent notices. And for each deal, you’ll also get a compliance certificate, which says I’m in compliance with the loan agreement and that I’m supposed to have this ratio of collateral versus the debt.
This monitoring aspect, combining multiple data sets, is something that these systems are particularly good at doing on the fly.
DMI: Can MCPs bridge the gaps in legacy tech systems?
HC: MCP, at its core, is about enabling systems to connect. But not all systems can connect. If your data is locked in a walled garden or a legacy tech system without an API, MCP alone can’t get to it. Something has to expose it first.
DMI: That sounds like there may be a limit to MCPs’ ability to knit together disparate systems.
HC: What it doesn’t solve is what it bridges to. If you have a legacy system that is not built with AI at its core, adding an MCP interface is for naught. It would be no different than having a REST API. You don’t make it agentic by having the protocol. What you really need to be able to have is a system that then can take a natural language input, a prompt, and generate an output based on a prompt. And so that means it’s a big advantage for those systems that were built with AI at their core and an architecture at the core that opens up those systems and makes them more powerful. So, MCP is not a silver bullet. It is a connectivity layer. The value ultimately depends on the systems it connects to and the quality of the data and workflows behind them. Organizations still need modern platforms capable of understanding, validating, and acting on information in an intelligent way.
DMI: How are MCPs helping Alkymi’s clients?
HC: What we’re seeing is a huge wave of our clients experimenting with and beginning to adopt systems like Claude, Microsoft CoPilot, and Google Gemini that can support investment workflows. In the Alkymi environment, they are now able to go in and configure these different tools. Data that was previously locked in siloed environments can now be connected and configured. You’re essentially chatting with your Alkymi environment directly in Claude, and you can start generating an incredible amount of powerful insights on your investments that up until now were stovepiped – data would leave our platform, and it might go into an accounting platform where it was much more difficult to get insights on-the-fly.
Our clients want secure access to trusted operational data while maintaining governance and control. MCP provides a mechanism for connecting those AI experiences to the validated data and workflows already managed within Alkymi.
DMI: What sort of flexibility is built into the platform?
HC: It’s great when your data is available in platforms like Snowflake, Databricks, Aladdin, or SimCorp. The challenge is that the data often remains fragmented across systems, making it difficult to generate insights or answer questions that span multiple applications.
By adding MCP, we can connect Alkymi to a broader ecosystem of applications and AI tools. That allows users across the back office, middle office, and front office to access and interact with trusted operational data without having to move everything into a single system. And it allows you to generate ad-hoc insights that up until now have been very difficult to generate because the data has been fragmented across multiple systems where those systems have been stovepiped and the data was never really intended to be shared. It was intended to be sent down into the accounting system. It was intended to be sent down into the system record.
The flexibility comes from being able to combine information across platforms and generate ad hoc analytics that were previously difficult to produce because the data was fragmented and locked within individual workflows. In this new model, users can query data through Alkymi to answer questions in real time. You can see cash flows, monitor compliance, identify exceptions, and generate insights across portfolios in ways that were previously much more difficult to achieve.
The goal is not to replace existing systems, but to make trusted data and workflows more accessible wherever users and AI agents need them.
DMI: You have partnered with a number of companies on new capabilities. What is the latest news on that?
HC: We recently made an announcement with Finbourne in London that focuses specifically about private credit and monitoring private credit assets. We’re working closely with them on agentic integrations, where Alkymi becomes a tool within a broader set of workflows orchestrated on the Finbourne platform.
We’re doing the same thing with SimCorp, where SimCorp has announced their agentic AI and we’re then making the Alkymi system available for the SimCorp agent launchpad that they announced at their global user conference recently. Someone can now sit at their SimCorp agent launchpad and ask questions and use the Alkymi tool to generate specific insights.
More broadly, we’re enabling Alkymi to serve as a trusted source of operational intelligence within agentic ecosystems. Our goal is to make validated, workflow-ready investment data available wherever users and AI agents need it. The industry conversation often focuses on the models and agents themselves, but in practice the bigger challenge is ensuring those agents can access trusted data and workflows. That’s where we believe Alkymi plays a critical role.
Subscribe to our newsletter
