Stop the bill before the bill stops you.
Cloud Watchdog detects dangerous cloud spend in minutes using your provider's usage metrics, reconciles dollars with the cost APIs once they catch up, and lets you opt-in to safe auto-stop on dev/staging only. Production stays locked by default. AWS is supported in alpha; Azure and Google Cloud are next.
10-min setup · no long-lived cloud credentials stored · cancel any auto-stop in 5 min
Lambda invocations 32× over 7-day baseline
cw-test-fn · us-east-1 · env=dev
4,812 invocations in 5 min · budget hit at $42.18
Suggested
Throttle to 0
Auto-runs in
04:38
Example alert · not a live message
5 min
Detection window via CloudWatch
0
Production resources auto-stopped, ever
$19
Starter / month, founding price locked
10 min
Setup via CloudFormation template
Why native cloud-billing alerts fail startups
You don't need another dashboard. You need a circuit breaker.
Every major cloud's native billing tool has the same five structural gaps for teams under 50. Cloud Watchdog closes them one by one.
How it works
Detection in minutes. Action only after you say so.
Connect your cloud in 10 minutes
Deploy a small role template in your account (AWS today; Azure & GCP soon). Cloud Watchdog assumes that role with a per-account secret. No long-lived credentials stored.
Poll usage metrics every 1–5 min
Lambda, EC2, NAT, ECS, RDS, ELB on AWS today; equivalents on Azure/GCP next. Deterministic rules detect spikes against your 7-day baseline.
Slack alert with a Cancel button
Every action sends a Slack message with full context and a 5-minute cancel window before any auto-stop runs.
Find and kill idle waste every 6 hours
Unattached EBS volumes, unused Elastic IPs, idle NAT gateways, zero-connection RDS, low-CPU EC2, orphan snapshots. Each finding shows the exact monthly $ you save by removing it.
How alerts actually arrive
Slack + Email — not a dashboard you forget to open.
Every alert is dispatched to both channels. Slack messages include a Cancel button so you can call off an auto-stop in 5 minutes; emails carry the same context for whoever's reading on their phone at 11 pm.
Slack
#alerts-aws
Block-kit message with severity emoji, rule + resource details, and the 5-min Cancel + Throttle now buttons inline.
alerts@cloudwatchdog.online
Branded HTML email with the same severity coloring as Slack. Signed SPF + DKIM so it never lands in promotions.
What gets protected
Four guards, day-one.
Detect spikes in real-time, kill idle waste continuously. Both halves of the problem instead of just the alert half.
Lambda runaway guard
Invocations, errors, duration, concurrency. Throttle to 0 on tagged dev functions before the recursion compounds.
10 ms
Slack-to-throttle path
EC2 dev stop guard
Stop tagged dev/staging instances when usage spikes. Production-tagged resources are refused at the IAM layer.
0 prod
Resources ever stopped
NAT egress watcher
Bytes processed and outbound traffic spikes — the silent six-figure category that AWS Budgets can't isolate.
<5 min
From spike to alert
Idle resource sweeper
Detect unattached EBS volumes, unused Elastic IPs, idle NAT/RDS/EC2, and orphan snapshots every 6 hours. Each row shows exact monthly $ savings.
6 types
Waste categories scanned
Built for humans, not consoles
What CloudWatch makes hard, we make obvious.
Six concrete things the AWS Console + CloudWatch Alarms either don't do, charge extra for, or bury four clicks deep.
Unit-aware thresholds
70%, not 70
Pick CPU and type 70 — the form shows %. Pick NetworkOut and type 10 — the form shows MB. The rule list reads back "NetworkOut ≥ 10 MB" instead of CloudWatch's bare "10485760". One less mental conversion every time you read an alert.
Auto-suggest threshold
P95 × 1.2, in one click
CloudWatch makes you guess the threshold value. We measure the resource's last 7 days, compute P50/P95/P99, and propose a sensible number. One click fills in the input. Refine from there.
Full metric library
44 metrics across 7 services
EC2 (8 metrics including CPU credits + metadata-no-token), EBS (queue length, R/W throughput, idle time), RDS, Lambda, S3, NAT, ALB — every metric AWS Console shows on a per-resource graph, plus the rare ones nobody surfaces but matter for cost (S3 BucketSizeBytes, NAT BytesOutToDestination).
Multi-resource scope
One rule, many instances
Pick "all current + future EC2" so new instances get covered automatically — or hand-pick a subset by checkbox. CloudWatch forces one alarm per resource; you don't have to recreate the same rule 80 times.
Resource detail graphs
Every metric, one click
Click any resource → see the full AWS Console-style graph grid: every metric for that service, last 24 h, with a "Set alert on this metric" CTA right beside each graph. No tab-hopping into CloudWatch.
Suggested rules from inventory
Zero-config onboarding
Have 8 Lambdas? Suggested: runaway-invocation guard. Have a NAT gateway? Suggested: egress-spike alert. Cloud Watchdog reads your inventory and proposes the rules you'd otherwise spend an afternoon authoring.
Safety rails
Guard-rails baked in, not bolted on.
No long-lived credentials
Short-lived role assumption + per-account secret on every customer connection.
Read-only mode available
Run alert-only forever — no write permissions if you don’t want them.
Open-source role templates
Audit our CloudFormation (and equivalent Azure / Terraform) in public before you deploy.
No destructive actions
v0 only throttles, stops, or sets desired-count to 0 — all reversible.
No long-lived credentials
Short-lived role assumption + per-account secret on every customer connection.
Read-only mode available
Run alert-only forever — no write permissions if you don’t want them.
Open-source role templates
Audit our CloudFormation (and equivalent Azure / Terraform) in public before you deploy.
No destructive actions
v0 only throttles, stops, or sets desired-count to 0 — all reversible.
Starter
Most popular$29$19/ month
Founder pricing — locks in for life.
3 cloud accounts, 10 circuit breakers, 25 alert rules, idle-waste scanner, Slack + email.
Choose StarterTeam
Coming soon$79/ month
5 cloud accounts, unlimited breakers, SMS + per-env tagging.
Notify me at launchScale
Coming soon$299/ month
10 cloud accounts, team dashboard, audit log export, SLA.
Notify me at launchBuilt by a founder, not a company
Stuck on setup, hit a bug, or want a feature? Reach out — I'll reply.
Cloud Watchdog is built and run by a single founder. If anything in the onboarding, alerts, or billing flow feels off — or you just want to chat about your cloud bill — drop me a line.