3 early MCP insights and learnings from our customers

I spend my days partnering with companies that use Fullstory to its fullest extent, across teams and use cases., Over the last few months, I've seen the same mindset shift across my customers: MCP & AI first.

It starts when a team connects Fullstory to an LLM through MCP. Behavioral data stops being something you go and check, and becomes something you can ask questions about, and increasingly, something your systems act on without waiting on a human.

If you're looking to get set up with MCP, start with Fullstory's MCP docs.

How customers use Fullstory MCP

Here's what three of our early MCP customers are actually doing in their own words.

1. Talking to their analytics

At one of the nation’s largest automotive retailers, MCP changed how teams interact with behavioral data. A senior UX design manager put it plainly:

"[MCP] lets us not just be a recipient of data anymore, but interrogate it. … I can look at Fullstory, I can look at everything, and contextualize that into one answer."

Instead of opening a dashboard and hunting, he pulls behavioral data, survey results, and financial data into a single conversation and gets one answer back. He runs it a few times a week to monitor funnel performance and quickly identify any issues to his executive team.

The second-order effect surprised them more: who could suddenly do this.
As a lead product designer on the team said:

"It's democratized data and access. I'm able to get a ton of information that is not normally at the designer's hands … and can get solutions out very easily."

Yet another Fullstory customer brings Fullstory data together with other analytics and back-end monitoring tools via MCP to set daily priorities.  For this engineering team, their focus is on quantifying issues to help prioritize their focus. Through this process, the team gets a regular report on errors, outages, and negative behavioral swings, driven by the severity of impact. Thanks to this approach, they have a data-driven north star that guides their team’s work, far beyond what one human could find in a dashboard.

That's the quiet story of MCP. It doesn't just make analysis faster, it widens who gets to act on what the data says.

2. Catching problems before their customers do

Equals Group, a London-based financial technology payments company, is using MCP for something more autonomous. Their VP of technology, Richard Coombes, described spotting trouble without waiting for a complaint:

"Now if there's an issue, we can usually detect it without the customer contacting us. There are signals, a rage click on the front end, or a payment taking a long time in our data lake."

Once a signal fires, MCP pulls the rest of the picture together, logs and past tickets and recent releases, and turns it into a decision:

"We use the MCP to pull in all the logs, then decide: this looks like a customer having an issue. We can solve it straight away, or reach out and say we know you're having a challenge. For us, that's a huge differentiator."

This is the proactive loop in practice. Behavioral signals feed a judgment, and the judgment drives action while the customer is still in the moment, before they've quietly given up.

3. Closing the support-to-engineering loop automatically

The engineering team at The Last Unicorn wired MCP directly into how they build. Their founder and CEO, Zac Sheffer, walked me through it:

"We set up an automation so that every time a new ticket gets created, it will automatically review the ticket and try to find if there's a customer record associated with it."

He tried to break it and couldn't:

"I tried to trick it. This email actually isn't the right email, but it knew enough information that it was able to narrow down to the right email, find the Fullstory session, and then tag that onto the ticket automatically."

From there it gets genuinely surprising. Their agents pull the session through MCP, generate comparable synthetic data, and build a browser test that replays exactly what the user did. The result is a reproducible test attached to every customer ticket:

"Literally within a couple of minutes I can go from picking up a ticket to confirming that it's all fixed, before me as an engineer even has to look at it. It basically brings the amount of work that I have to do down so, so low."

That's the reactive loop, fully closed. A real user's struggle becomes a reproduced bug, a proposed fix, and a verified result, mostly before an engineer has to touch it.

But none of these can work without ground truth

Here's the thread running through all three use cases.

Fullstory MCP is the connection, but what travels through it must be complete and trustworthy, or the outputs fall apart. The same UX design manager, talking about the LLM his team works through, had the sharpest way of putting it:

"It's like your smartest intern. It's really, really good at its job, but it doesn't know anything about what you want."

The context you feed it is the whole game. That's why the teams getting the most out of MCP are the ones who were already capturing complete behavioral data, not a tagged fraction of it. When the record is complete, you can finally answer the question that actually matters. The same team's Fullstory administrator put it this way:

"Fullstory is going to help us tell the why behind any numbers we're seeing … and keep a pulse on whether the transformations we're doing are working."

Where this goes

Three companies, three very different problems, one direction.

Analytics is moving from something a few analysts check to something anyone can interrogate, and something our systems can act on, proactively when they can and reactively when they must. MCP is how the behavioral data gets there. The complete, real-time record of what your users actually do is what makes any of it trustworthy.

This is early, and the best use cases are still being invented by customers like these. If you're building on Fullstory MCP, I'd genuinely like to hear what you're doing with it.

We've created an MCP prompt library to help you envision what's possible with Fullstory MCP, and if you want to submit an idea, success story, or prompt you've found valuable, you can do so there.