Your Users Won't Have to Tell You
See every load, error, and interaction real users hit on web, mobile, and mini programs. Errors group into issues and go straight to on-call and AI SRE.
Sessions
JS error rate
LCP · P75
INP · P75
Sessions & errors
Top issues
TypeError: Cannot read properties of undefined (reading 'filter')
1.2k · 86 usersChunkLoadError: Loading chunk 412 failed
634 · 51 usersUnhandled Promise Rejection: Network Error
289 · 33 usersResizeObserver loop completed with undelivered notifications
176 · 29 usersSyntaxError: Unexpected token '<' in JSON at position 0
92 · 14 usersCannot read properties of undefined (reading 'filter')
src/pages/rule/add/content.tsx:322
Errors grouped into issues, stacks mapped to source
Similar errors merge into one issue, showing how many users are affected and which release introduced it. Upload source maps and minified stacks resolve to the original file, function, and line.
Cannot read properties of undefined (reading 'filter')
Events
1,204
Users
86
First seen
v2.14.0
Impact breakdown
Minified stack
Resolved with source map
src/pages/rule/add/content.tsx:322:16
Duplicates collapse
One root cause, one issue to triage
Straight to the line
File, function, and snippet in the issue
Release-aware
First-seen version makes regressions obvious
Rewind to the moment it broke
Session replay reconstructs clicks, scrolls, and navigation, with errors and slow requests marked on the timeline. User input is masked by default, and password fields are never captured.
New alert rule
Rule name
Query
View /rule/add
00:04
Click “Add condition”
00:13
Type rule name
00:22
TypeError: Cannot read properties of undefined
00:31
POST /api/rule 1.8s
00:40trace 4bf92f…Backend trace
No more guessing
Replay every step in order
Context included
Errors and requests line up with replay
Private by default
Input masked, elements excludable
From a user's error to a fix PR, in one flow
RUM issues land in your Flashduty team space and page the on-call engineer per your dispatch policy. AI SRE picks up the session, release, and source-mapped stack to investigate and propose a fix.
RUM detects
A real user hits an error; it groups into an issue
On-call routes
Dispatch policy pages the engineer on duty
AI SRE investigates
Session and source stack point to the root cause
Fix shipped
A fix is drafted and a pull request opened
Real-world story
From a real user's error to a fix PR in 8 minutes
No waiting on tickets
Handling starts before anyone reports
Reuse on-call
Team-space dispatch, or a webhook
Follow the request to the backend
Frontend requests link to backend traces
Measure speed with real user data
Not lab scores — the LCP, INP, and CLS each user actually experiences on real networks and devices. Slice by page, release, region, and device in the Explorer to find the slowest stretch.
Core Web Vitals · P75
LCP
2.1s
Good ≤ 2.5sINP
168ms
Good ≤ 200msCLS
0.04
Good ≤ 0.1LCP · by device
Desktop
1.8s
Mobile
2.7s
Tablet
2.2s
Explorer · Slowest pages
Page load waterfall
Real users, real conditions
Every network, device, and region
Drill into any page
Split by page, release, region, and device
Slow requests in view
API latency and asset loads side by side
One SDK family for every surface your users touch
Native SDKs for 8 platforms with one data model. Whether users arrive from a browser, a phone, or a mini program, their experience lands in one place.
One data model
Sessions, views, actions, resources, errors, and long tasks share the same fields on every platform.
A few lines to install
Add it via CDN or npm; collection starts on init — no per-page instrumentation.
Light on the page
The web SDK is about 50 KB gzipped, loads async, and initializes in under 8 ms.
Sampling and privacy controls
Set session sample rates and mask user input by default to keep volume and compliance in check.
FAQ
Less noise. Faster recovery.
Start free—no credit card. Book a demo to see your alerting pipeline end to end.