Know When Your Site Is Slow — Not Just Down.
FlareWarden builds a response-time baseline for every hour of the week from your monitor's own history, then flags the checks that fall far outside it — the quiet slowdowns that never show up as downtime. Cron jobs get the same treatment for run duration. No thresholds to guess.
Currently in beta.
Hour-of-week baselines learned from your monitor's own checks. Tuesday 2 PM is compared against other Tuesdays at 2 PM — not a one-size-fits-all average.
Duration drift computed from the last 50 runs. Catch the backup that quietly grows from 4 minutes to 40 — before it collides with the morning traffic.
Catch the Problems That Never Go Down
A site can answer every check and still be quietly broken. Anomaly detection surfaces the slowdowns that uptime percentages hide.
Expected-Range Band
Your response-time chart gains a learned expected range for every hour of the week, so "normal" is visible at a glance.
Watching vs. Confirmed
A single slow check shows as amber "watching." Only sustained deviations become confirmed anomalies — one-off blips never page you.
Anomaly Log
Every confirmed anomaly is recorded with its window, observed vs. expected range, and deviation — a clean record for your postmortems.
Opt-In Alerts
Email and webhook alerts on confirmed anomalies, off by default per team member. Webhooks fire the anomaly.detected event.
Mark Expected
Was that spike a planned load test? Mark it expected and it's excluded from future baselines — the model learns from your feedback.
Never Touches Uptime
Anomalies are a parallel, quieter signal. They never change your uptime percentage and never open incidents.
Up, But 3× Slower. Now You'll Know.
Every site has rhythms — busy weekday afternoons, quiet Sunday mornings. FlareWarden learns yours per hour of the week, so a response time that's perfectly normal at peak doesn't trigger noise, and one that's abnormal at 3 AM doesn't slip through.
The Backup Still "Works." It Just Takes 10× Longer.
A cron job that completes isn't necessarily healthy. Growing tables, slow disks, and runaway queries show up as duration drift long before they show up as failures. FlareWarden learns each job's typical run time from its last 50 runs and alerts when a run falls far outside it.
Learning Takes a Little Time
Good baselines come from real data, and we'd rather tell you exactly where the model stands than pretend it's magic. The chart always shows which stage you're in.
Days 0–3
LearningCollecting data. No band on the chart yet and no anomaly events — just a clear "learning" indicator so you know where the model stands.
Days 3–21
Basic RangeA working expected range appears and anomaly detection is live — it just doesn't know your weekday-vs-weekend patterns yet.
Day 21+
Full AccuracyHour-of-week baselines kick in. Busy Monday mornings and quiet Sunday nights each get their own expected range.
Already monitoring with FlareWarden? Monitors with existing history skip the wait — baselines are computed from the data we've already collected, so you get value the moment you enable it.
How It Works
Learn → Band → Flag → Report. You enable it; the model does the rest.
Learn
FlareWarden studies your monitor's own history — response times per hour of the week, or run durations for cron jobs.
Band
An expected range appears on the chart. It refreshes nightly, so "normal" keeps up with your site as it evolves.
Flag
Checks outside the range turn amber ("watching"). If the deviation persists across consecutive checks, it's confirmed and turns red.
Report
Confirmed anomalies land in the anomaly log with observed vs. expected numbers — plus optional email and webhook alerts.
On Every Plan. Even Free.
Elsewhere in this space, learned baselines are an enterprise feature gated behind $625+/mo tiers. FlareWarden ships anomaly detection on every plan. Enabling it on a monitor uses one slot from your plan's monitor pool — the same pool everything else draws from.
Each monitor you enable it on uses one extra slot — simple, predictable accounting.
Disabling frees the slot immediately. Your baseline is kept, so re-enabling is instant.
Upgrade your plan — or on any paid plan, add 10 extra monitors for $5/mo.
Frequently Asked Questions
FlareWarden learns each monitor's normal behavior from its own history — response times for every hour of the week for uptime monitors, run durations for cron jobs — and flags checks that fall far outside that range, even when the monitor never goes down. There are no thresholds to configure; the baseline comes from your real traffic patterns.
No. Anomalies are a separate, quieter signal. They never change your uptime percentage, never open incidents, and never trigger cross-region validation. Your availability numbers stay exactly as they are.
Monitors with existing history get value right away — baselines are computed from the data FlareWarden has already collected. New monitors show a basic expected range after about 3 days and reach full weekday/weekend accuracy after about 21 days of history and enough checks. The chart shows a clear learning indicator until then.
Cron duration baselines are computed from your job's last 50 completed runs using the median and a robust measure of deviation. When a run takes far longer than the job's own normal, you're alerted — useful for catching a backup that quietly balloons from 4 minutes to 40.
Yes. Each monitor has a three-position sensitivity setting: Low flags only extreme deviations, Normal is a balanced default, and High flags smaller deviations sooner. Confirmed anomalies also require the deviation to persist across consecutive checks, which keeps one-off blips from paging you.
Yes. Anomaly detection is available on every plan, including the free plan. Enabling it on a monitor uses one additional slot from your plan's monitor pool, and you can turn it off any time to free the slot.
Find Out What Your Site's Normal Looks Like.
15 free monitors, anomaly detection included. Enable it on a monitor and let the baseline build while you get on with your day.
Start Free Monitoring