What MCP actually is
Model Context Protocol (MCP) is a standard that lets AI models like Claude connect directly to external tools and data sources - databases, CRMs, APIs, file systems - in a structured, repeatable way. Instead of copying data into a chat window or waiting for a report, a user can ask a question and the model fetches live data to answer it. In practical terms: a finance lead opens Claude, asks "what is our net new ARR this week compared to last week, broken down by segment," and gets a real answer drawn from live HubSpot and accounting data.
Why dashboards are not enough
Dashboards are good at answering the questions you thought to ask when you built them. They are static by design. The moment a stakeholder asks a question that was not anticipated in the original build - and they always do - someone needs to write a query, export the data, or ask an analyst. This creates a dependency bottleneck. The people who need information most are often the least technical, and the people who can retrieve it are always busy.
"What MCP enables is not just faster data retrieval. It is a shift from 'I need to ask someone to pull this for me' to 'I can just ask.'"
What a RevOps MCP implementation looks like
In one recent engagement, we connected Claude to a client's HubSpot instance, their accounting platform, and a Google Sheet they used for headcount planning. Within a week, the questions they were asking had changed. Instead of monthly pipeline reviews, they were doing ad-hoc checks throughout the week. The behaviour shift was immediate and visible.
What it takes to build it
An MCP implementation requires three things: a clear map of which data sources will be connected and what queries they need to support, well-designed server endpoints that handle authentication and data normalisation, and a testing process that validates the model's responses against known-correct data. The model is only as reliable as the data it is reading and the instructions it is given about how to interpret it.
Who this is for
MCP is highest value for teams that have good data but poor access to it - where the information exists but retrieving it requires a technical intermediary. Finance teams, operations leads, and founders who want operational awareness without building a dedicated analytics function are the primary beneficiaries.