One of the most gratifying parts of building a platform is seeing all the ways that customers use it, and as we’ve rolled out Fullstory access via MCP, we’re seeing a wave of new usage patterns shared by our customers.
One of the more common patterns is a periodic review of app functionality to catch defects and points of friction—we’re calling this a “Bug Sweep”. In this article, I’ll show you how to get started with your own Bug Sweep.
What’s a “Bug Sweep” for?
You’re already using Fullstory to monitor key metrics and to root cause issues by watching sessions. Even so, a huge amount of the data’s value is still lurking below the surface: watching individual sessions reveals points of friction and bugs that don’t necessarily bubble up clearly in KPIs, and users don’t report every bug they run into. But who has time for that?
One customer writes:
Up until recently, you had to manually watch individual sessions, but Fullstory launched an MCP that lets an agent review them automatically. One person can only watch so many sessions, but using the MCP gives you a more comprehensive sweep to catch more edge cases.
I've been running Fullstory through Cursor every week on [app surface] and it's the fastest way I've found to catch bugs that dashboards miss. Since I started using this, I’ve caught 15+ bugs, which is way more than I could have caught on my own and led to big CX improvements.
Let’s get started!
If you’re a Fullstory customer, you already have access to the MCP. If you haven’t tried our MCP support out yet, start with a quick look at the MCP setup guide.
We’ll start by building our prompt. The general formula is as follows:
Find - Scan sessions for errors and frustration signals to detect potential friction. Don’t just look for events like explicit errors that might be caught by existing tooling—agents actually “watch” sessions to mine for the issues between the events.
Confirm and measure - One session with friction is a lead—is this a one-off problem, or a systemic one? Use Fullstory’s analytics to get a quantitative assessment of impact. Watch similar sessions to confirm.
Report and remember- Report what’s found, and save findings so that the next scan run re-checks them.
This builder will help you customize a starting prompt. Configure it, then copy the prompt and give it a test run!
Where this works
Most existing alerting/error monitoring relies on pre-configured definitions of bad events: exceeded metric thresholds, console errors, etc. The Bug Sweep goes one layer deeper, watching sessions at scale to find the kind of issues that customers write into support to report—or, even worse, don’t write about, suffering or abandoning in silence.
What’s next?
Today you can DIY; soon, it will be turnkey. We’re hearing about enough variants on this pattern that we're working on putting this on rails with 24/7 monitoring agents, both via MCP as well as baked into the StoryAI platform. Stay tuned for some announcements this fall.
