Case study · Cortex
How Cortex uses real analytics data to power customer QBRs
Cortex set up AgentCat in a day ahead of their MCP server's GA launch. Here's how the team turned MCP usage data into faster QBRs and a sharper roadmap.
Hours → minutes
customer QBR research time
Many → one
internal workflows simplified into a single product
50%
of sessions map to two core use cases
“Bringing in MCP usage into QBRs and customer calls has become simple. What previously took hours of sifting through various sources of data now just takes 15 minutes of browsing a customer's Session Replays and Agent Goals.”
Impact
- Reduced QBR research time from hours to 15 minutes
- Consolidated QBR prep from multiple internal products into one workflow
- Identified that 50% of sessions map to two core use cases
- Shaped the MCP roadmap around measured user segments
About Cortex: Cortex is the Engineering Operations Platform that runs mission control for the AI software factory. Their customers include organizations such as H&R Block, Canva, and Skyscanner.
We first started talking with the team while their MCP server was still in limited beta. They were planning to GA it at their developer conference, and wanted to quickly get analytics and observability to see how their customers were using it. We worked closely with Cristina Buenahora, VP of Strategic Initiatives, to get everything set up in a day. Today, she and the customer success team use AgentCat to prepare for customer QBRs, simplifying a process that used to take several hours into just minutes of browsing Session Replays and Agent Goals.
THE PROBLEM
They knew MCP was a popular feature, but didn't know how exactly customers were using it
As coding agents moved into daily engineering workflows, Cortex made an early bet: engineering leaders shouldn't have to leave their IDE or agent to answer questions like "who owns this service?" or "does this meet our standards?" So Cortex built an MCP server to put its data directly where engineers and agents work.
The signal was immediate. Usage climbed, and the MCP came up in customer call after customer call. The open question wasn't whether it mattered. It was how customers were using it, and Cortex didn't have the data to say.
"People were mentioning the MCP server often. But we didn't have any data to understand how they were using it."
They built some in-house dashboards pieced together from API calls and logs, but it still required several people from the customer success and engineering teams to manually pull and compile whatever data they could get their hands on before QBR calls.
They knew the customers were using their MCP server. They just didn't have the right data to know how they were using it.
THE SOLUTION
From black box to per-customer dashboards
We got in touch with the team a week before their EVOLVE developer conference, where they were planning on officially launching their MCP server in GA. They wanted to see if they could get our SDK set up in a week. We got things set up for them in a day.
The conference came, and the MCP launch was met with strong interest and installs. And now that they had AgentCat set up, they could see what customers were doing, the major use cases and how to prioritize their MCP roadmap for the rest of the year.
Today, Cortex has integrated AgentCat into customer QBRs. Before the call, they pull up a customer's dashboard in AgentCat and view how many engineers are using the MCP server, where agent sessions ran into errors, and the top use cases and agent goals.
Segmenting users based on intentions, not actions
The insights didn't stop there. AgentCat also helped identify different use cases within its users.
Some users used the Cortex MCP server to ask quick, one-off ownership questions, such as "Who owns this service?" Other user segments used the MCP to continually monitor their company's compliance with scorecards and workflows.
The two segments shaped the roadmap in two ways. Cortex smoothed the first-run experience for the engineers asking quick ownership questions, and it built out deeper scorecard and standards capabilities for the teams using the MCP to monitor compliance continuously.
RESULTS
Using AgentCat, Cortex turned its MCP from an unquantifiable but beloved feature into a measurable product differentiator. They were able to transform the hours-long messy process of preparing for every QBR into a simple workflow of opening up the AgentCat dashboard.
They then used the data and insights from AgentCat to shape their MCP roadmap and make their MCP server more appealing to all user segments and customers.
Today, Cortex can identify champions, flag churn risk, and tailor expansion conversions all from MCP usage data that they didn't previously have.
"Our plan is to make Cortex the best tool to drive operational excellence within software organizations, and that means working where users already are, such as coding agents. We're investing heavily in our MCP server and AI agent compatibility, and AgentCat is crucial to that investment," says Cristina.