M SCIENCE BLOG SERIES
Connecting Investors
to M Science Data Through MCP
FEATURING A DEEP DIVE INTO
M Science Unified Data Model (UDM)
A look at why M Science built an MCP Server, how it works with the Unified Data Model, and what it unlocks for the future of investment research.
KEY TAKEAWAYS
- The M Science MCP Server brings analyst-curated data and KPIs directly into the AI tools investors already use, including Claude, ChatGPT, and internal agents, with no separate interface to learn.
- The MCP tools sit on top of the Unified Data Model, providing access to both analyst-covered companies and the broader universe of tracked companies.
- A semantic layer adds analyst context to the underlying data, helping AI understand what a KPI means and which data cut answers are relevant to a given question, rather than simply identifying where the data resides.
- The result is broader company coverage and greater trend granularity through the same natural-language interface, whether investors are analyzing a single covered company or comparing trends across sectors.
What Is MCP? In Plain Terms
Model Context Protocol, or MCP, is an emerging standard that connects AI systems to external data sources, enabling them to interact with the data through natural language. For institutional investors, this means asking a question much as they would ask an analyst and having the AI query the underlying data for an answer, rather than relying solely on information included in the model’s training data. M Science’s MCP Server provides the connective layer that allows Claude, ChatGPT, and internal AI systems to query M Science’s Unified Data Model rather than the open web.
Multiple Tools, One Data Model
The MCP Server is not a single search interface. It provides a set of purpose-built tools, each backed by the Unified Data Model. Together, these tools provide insights on companies actively covered by M Science analysts and the broader universe of companies tracked across M Science’s underlying data feeds.
Research & KPI Retrieval
Surfaces written research, KPI definitions, and analyst context tied to a covered company, so a question about demand trends or a specific metric returns the context needed to interpret the results, not just a stand-alone figure.
Unified Data Model (UDM) Tool
The Unified Data Model (UDM) tool combines semantic KPI discovery and direct data retrieval into a single workflow. It translates an investment question into the relevant data cut, matching the query to the most relevant KPIs across M Science’s 1,440+ metrics without requiring an analyst to know which dataset, field, or KPI to select in advance. From there, the tool retrieves the underlying time-series data, including growth rates, trend lines, and other quantitative signals, across more than 1,400 public and private companies in the UDM, all queryable in plain language. Because the tool operates on the same Unified Data Model, analysts can move seamlessly from breadth to depth. They can scan a sector for companies or KPIs showing the sharpest inflections, then drill into the most relevant names without switching data sources or reestablishing context.
Breadth and Depth, in Practice
The clearest way to see how these tools extend the scope of analysis is through the questions they can help answer, from those focused on a single covered company to those spanning a broader competitive set.
Going Deep on a Covered Name
SAMPLE QUERY
“Is Spotify engagement per user improving or weakening in 2026, and what could that imply for churn and monetization?”
Answering this type of question traditionally required an analyst to pull downloads, MAU, DAU, and subscriber trends separately and connect those signals into a broader narrative. With the semantic layer translating the question into the relevant KPIs and data cuts, the MCP Server can surface the information needed to evaluate engagement and its potential implications for churn and monetization.
Going Wide Across a Sector
SAMPLE QUERY
“Show H&M’s monthly market share in France versus Zara and Uniqlo over the past 12 months.”
This is a breadth question — three retailers, one geography, twelve months of trend. Because the UDM Tool sits on the same standardized schema across the tracked universe, a multi-company comparison like this one returns as a single query rather than three separate lookups.
EXTENDING INTO UNCOVERED NAMES
SAMPLE QUERY
“Which Anheuser-Busch brands contributed most to U.S. spending growth in Q2?”
Brand-level and subsegment questions like this may fall outside the scope of a single published research report. Because the underlying data feeds extend beyond the analyst-covered universe, the MCP Server can surface relevant data for companies, brands, or subsegments that may not have dedicated analyst coverage.
Across all three examples, the mechanism is the same: the question is mapped to the relevant KPI or company dataset, while the UDM standardizes the resulting data to enable direct comparisons across companies.
Grounded in M Science, Not the Open Web
For an investment workflow, one of the most important distinctions is where the answer comes from. Responses are grounded in M Science research and data rather than relying solely on a general-purpose model’s training data or information from the open web. This grounding helps reduce the risk of hallucinated figures and outdated information while providing greater transparency into the underlying sources. For a PM or analyst using AI to accelerate idea generation, that transparency makes AI-generated output more useful: the underlying data can be traced, validated, and incorporated into the investment research process with greater confidence.
Talk to your M Science representative to learn more about accessing the MCP Server.
