In private markets, communication has always been the problem nobody solved. Deals arrive over WhatsApp. Terms change overnight. Sales teams maintain parallel Excel files and forward them by email. There is no shared system, no standard protocol, no infrastructure that can make sense of the volume and fragmentation of information that moves through this market every day.

EVIDENT has built one.‍

Today we are announcing the launch of our proprietary MCP infrastructure - a Model Context Protocol server built entirely in-house, designed specifically for the operational realities of private markets. 

The Problem We Were Solving

Private markets run on fragmented, informal communication infrastructure. A deal arrives via a WhatsApp message. The terms change the next morning. Another message arrives with an updated valuation. The deal closes, or disappears, and the only record of any of it is scattered across individual conversations that nobody else can access.

For a firm like EVIDENT, operating at the intersection of private markets and digital infrastructure, this creates a specific operational challenge. We track dozens of deal opportunities simultaneously supply and demand across multiple counterparties, asset classes, and geographies. 

Terms change in real time. New deals arrive constantly.

The question we set out to answer was direct: what would it look like if deal information could be captured, structured, and made queryable without requiring anyone to change the tools they already use?

The answer required building infrastructure that did not exist anywhere else in the market.

What We Built

EVIDENT now operates a proprietary MCP server, built entirely in-house. The system is integrated with our DealHub - a centralised infrastructure layer that captures, structures, and surfaces deal supply and demand across our network.

Here is how it works in practice.

A partner submits a deal through a natural language message to an AI agent. The agent, guided by the MCP server, parses that message and executes an API call to our DealHub, posting the deal to our market board with all relevant parameters: deal size, fee structure, structure type, closing date, counterparty. No manual data entry. No reformatting. The information goes directly into our infrastructure.

Another partner can then query the same system. They ask the AI agent, again through a natural language message - what is currently available at specific terms. The agent queries our DealHub, retrieves the relevant deals, and responds with structured, current information. The entire exchange happens in real time, through a channel the partner already uses, without requiring access to any platform or interface.

Partners can now post a deal, query the market, and receive structured data - in real time, always current, through the tools they already use.

This is what a data feed ecosystem looks like when it is powered by MCP. The infrastructure is invisible to the end user. What they experience is simply: ask a question, receive an accurate answer. What makes that possible is the layer we built.

Why This Is Significant

Private markets have never had a shared infrastructure for deal data. Unlike public markets where price discovery happens on exchanges with standardised data feeds, private market deal flow moves entirely through informal channels. There is no Bloomberg terminal for this information. No API. No shared protocol.

The consequence is systemic inefficiency. Market participants spend significant time managing information that should be managed by infrastructure. Deals are missed because information is fragmented. Terms are tracked manually because there is no system of record. The market operates at a fraction of its potential speed and transparency.

MCP changes the architecture of what is possible. By building our own MCP server, we have created a foundation that can connect AI agents to our proprietary deal infrastructure reliably not occasionally, not approximately, but precisely and at scale. We can now extend that infrastructure to external partners, creating a network where deal information flows through a shared, structured channel rather than through hundreds of individual conversations.

We believe this is the direction the industry is moving. Within the next six to twelve months, every financial services firm that is serious about AI integration will need to build its own MCP layer. The firms that build it first will have established structural depth in the form of integrated systems, connected partners, and accumulated data that will be difficult to replicate later.