TL;DR
A leading investment research firm needed to put its proprietary financial data inside the AI assistants its customers already use. Node8 designed and built a production MCP server on Google Cloud — turning “our data” into “the reason people get better answers from their AI.”
The opportunity: why MCP matters
Every company with proprietary data faces the same strategic question in the AI era: build your own chat experience and compete with frontier AI labs, or make the assistants people already use dramatically better with data only you have?
The Model Context Protocol (MCP) makes the second path possible. MCP is an open standard that lets any AI assistant — Claude, Copilot, Perplexity, and others — securely connect to an external data source and query it in real time. When a user connects your MCP server, their assistant pulls your data on demand and gives answers it simply couldn’t produce otherwise.
The upside:
- A new distribution channel — one-click install through AI connector marketplaces.
- A durable moat built on proprietary data rather than model quality.
- Meeting customers where they work instead of asking them to adopt another tool.
Challenge
The firm owns a deep, proprietary financial dataset — structured research, ratings, earnings estimates, filings, and financial statements — plus a large library of unstructured analyst content. Its audience of financial advisors and sophisticated investors increasingly lives inside AI assistants.
A previous attempt to win by building a proprietary chat product had shown the hard truth: standalone chat loses to the frontier assistants. The data was the asset; the delivery model had to change — securely, at the scale of millions of users, without a heavy lift on legacy systems.
Approach
Node8 led the engagement end to end, pairing MCP design expertise with a technical build partner and the Google Cloud platform:
- Structured data first. The MVP exposed the firm’s core structured datasets through the MCP standard with a full OAuth 2.0 flow, engineered for a seamless one-click connection and a path to AI connector marketplaces.
- Designed for real usage, not API sprawl. Rather than mapping every internal API to its own tool — an unusable surface of hundreds of tools — Node8 grouped related calls into logical, task-oriented tool sets that keep the assistant efficient, fast, and cheap to run.
- Unstructured content as phase two. For the analyst research library, Node8 scoped a retrieval layer driven by rich structured metadata, with Google Cloud Vertex AI Search as the platform, so assistants can discover and cite the right documents.
The Google Cloud partnership
Node8 is a Google Cloud reseller and implementation partner, and this engagement shows why that matters — we don’t just design the AI layer, we stand up and run the platform underneath it:
- Google Cloud Platform (GCP) — the infrastructure foundation for the MCP server and its data pipelines, plus Vertex AI Search for retrieval.
- Google Workspace — collaboration, identity, and productivity, resold and implemented for customers standardizing on Google. (See also our Gemini Enterprise practice.)
- Google security tooling — enterprise security available through the same partnership.
- A single commercial relationship — Node8 provisions, implements, and supports the Google footprint: one partner for the AI build and the platform it runs on.
Why it worked
- The firm monetizes proprietary data through the fastest-growing distribution channel in software — the AI assistant ecosystem — with frontier labs as partners, not competitors.
- One team is accountable for both the AI product and the infrastructure beneath it.
- The pattern is repeatable for any organization sitting on unique data: expose it through MCP, run it on Google Cloud, and become the reason people get better answers from their AI.