How Meilisearch uses LogRocket’s MCP to improve free trial experience and feature adoption

“LogRocket was able to show us that there was a decent percentage of our customers who couldn’t get through our registration form. There was a translation on the page. It ended up crashing the page. It’s something that we maybe would have seen in some of our notifications, but not often enough for us to be able to really look into it and prioritize what the fix was going to be.

But being able to see it at that scaled level, being able to step back and say, ‘Okay, well, this is happening to, you know, X percentage of our potential customers,’ then we’re able to say, ‘Okay, let’s prioritize this. Let’s get a fix in place.’ Which we did, and that helped more people go through our registration process.”

– Gillian McAuliffe, Head of Experience and Engineering Operations, Meilisearch

This post is adapted from a conversation between Gillian and the LogRocket team. It has been edited for length and clarity.

Gillian McAuliffe is the Head of Experience and Engineering Operations at Meilisearch, the unified search and AI retrieval platform.

Gillian and her team were looking to improve two key aspects of the Meilisearch experience: free trial experience and adoption of newly shipped features.

The Meilisearch product team effectively has to track two separate, but connected, product experiences: Meilisearch Engine (an API-only interaction) and Meilisearch Cloud, which is the managed and hosted version with a full UI.

A user might set up their project in the UI, then switch to the API to index documents or run searches. Connecting those two sides of the experience required manual cross-referencing.

When users submitted a customer support ticket, Meilisearch would watch LogRocket sessions and cross-check them with data from API usage. This worked, but it became cumbersome at scale.

To close these gaps, Gillian built two new tools. Both tap into insights from Galileo AI via the LogRocket MCP, which makes LogRocket’s session replay data queryable by AI agents and internal tools, like the ones Gillian built.

A weekly report that combines LogRocket sessions with Meilisearch API usage data to show how the free trial experience is landing, and
Feature Watch, an internal Slack app that automates recurring check-ins for newly shipped features.

She shared how these tools came together, and how they’re helping Meilisearch create a system for continuous improvement of their product.

New to the LogRocket MCP? Check out our launch post for a deeper dive.

Can you tell us a little bit about Meilisearch?

Meilisearch is a super fast and developer-friendly search engine. Whether it’s powering AI search, RAG, or keyword search that’s lightning quick, Meilisearch simplifies how teams build and ship their search. Most teams use it through Meilisearch Cloud.

Editor’s note: Check out LogRocket’s complete guide on getting started with Meilisearch.

Meilisearch users can interact with the product in multiple ways. Does that add a wrinkle when you’re trying to triage user issues?

Meilisearch Cloud is effectively two products.

Meilisearch Engine is almost embedded into the cloud; it’s the data retrieval side of things. It’s API-only interaction. It’s our open-source product, and you can host it yourself.

The other side of things is Meilisearch Cloud, which is effectively the managed and hosted version of that Meilisearch Engine. But that effectively gives us two different products that we need to watch, and we need to be able to piece information together in order to put together the full picture of our customer experience.

Previously, we were using LogRocket mostly in a support context, so that would be on an individual basis. We would be able to find some user sessions and pair it together with API information from Meilisearch to get that full picture on an individual basis for support tickets that were coming through.

That works at an individual level. The sessions are really good; we can see what’s happening from the customer perspective.

But when we try to do that at scale, it becomes a little more problematic. So for us, it’s very hard to be able to piece together information from both the API and from the user session side.

How does the LogRocket MCP fit in?

On a daily basis, there’s an automation that will check for any customers that are three days into their free trial. We’ll gather some information from our database and from the registration form: the information that they’ve given us. We pair this with the task information from their Meilisearch instance. That’s the API side of things.

We send that off in a prompt to Galileo asking basically, “How has this customer’s experience been up until this point? Tell us if there’s anything going well, going badly, and here’s a bit of extra information to try and fill out that picture.”

And so Galileo is obviously able to take all of that information, take all the session data afterwards, and combine it all for us into one kind of overall picture, which is super important.

We save that into a Meilisearch instance, and then from the Meilisearch instance, we’re able to create multiple internal dashboards that really show how different things are performing, from the information that we’ve got about the free trial.

Anything specific this has helped you fix or improve?

As part of the registration for our forms, LogRocket was able to show us that there was a decent percentage of our customers who couldn’t get through our registration form. There was a translation on the page. It ended up crashing the page. It’s something that we maybe would have seen in some of our notifications, but not often enough for us to be able to really look into it and prioritize what the fix was going to be.

But being able to see it, you know, at that scaled level, being able to step back and say, “Okay, well, this is happening to, you know, X percentage of our potential customers,” then we’re able to say, “Okay, let’s prioritize this. Let’s get a fix in place,” which we did, and that helped more people go through our registration process.

You also built a Slack app called Feature Watch. How does that work?

We’re shipping features really fast here, and we don’t wanna just ship and forget, right?

We wanna be able to monitor things that are out in the wild, and the MCP and Galileo have really helped us make that process of feedback simple to stick to.

We’ve created a Slack application called Feature Watch. Anyone in the company can create a feature watch. It’s effectively a scheduled task that sends a prompt to Galileo that’s really specific about this feature to try and get an understanding of how people are using it, and the feedback for it in terms of when it’s out there in production.

So again, just trying to understand and iterate on the feature development.

What’s been the biggest benefit of what you’ve built?

All of this information that we get back from Galileo gets saved into a Meilisearch instance as well, and we’re creating lots of UI and internal dashboards based on that data. It really helps us be able to see how we’re improving over time.

I think using the MCP is really helping us to operationalize our mindset of continuous improvement. That’s really important here at Meilisearch.

The post How Meilisearch uses LogRocket’s MCP to improve free trial experience and feature adoption appeared first on LogRocket Blog.