Connect
Deploy a read-only IAM role with one click. No credentials to rotate. No secrets to store. 90 seconds, done.
◉ Read-only◉ Multi-account◉ ~90 seconds
Live in production · 6 accounts · zero rollbacksYour AWS bill is bleeding cash.
Your AWS bill is bleeding cash.
We stop the leak — and hand the money back.
Every hour, our agent hunts down idle EC2, orphaned volumes, zombie load balancers, and oversized databases in your account — then executes the fixes while you sleep. No dashboards to babysit. No tickets to close. Just money back in your budget.
- SOC 2 in progress
- Read-only until YOU say otherwise
- See every action in real time
- 30-day rollback vault
- Conversational agent · live in production
DEMO
app.purgeops.io / dashboard
01 / Savings · trend
$0
+38.4% vs last month
1D7D30D3MYTD
Remaining idle waste · $5,120/mo
ReclaimedWaste eliminated
Auto-purged
147
Awaiting
12
Accounts
3
Live activity
Live
✓$18.40/mo
Reclaimed idle EC2 i-0a2b3c
✓$6.20/mo
Deleted orphaned EBS vol-9x8y7z
!$347/mo
Flagged oversized RDS db-prod
✓$22.20/mo
Killed zero-traffic ALB alb-legacy
Avg cost per scan
$0
What it costs us to scan one resource. What it costs YOU: nothing until we find waste.
Time to first finding
<4min
From IAM connect to your first flagged dollar.
Risk tiers
0
Low executes on autopilot. Medium and high wait for your one-click green light.
AWS coverage
0%
Every EC2, EBS, ELB, and RDS across every region. Nothing hides.
01 /Under the hood
How you go from 'zero access' to 'cash reclaimed' in under 4 minutes
Three steps. One afternoon. Then the agent takes over.
Detect
The agent hunts your account 24/7. Every idle instance. Every orphaned volume. Every zombie load balancer. Priced in real dollars, explained in plain English.
◉ Continuous◉ Risk-scored◉ AI-explained
Purge
Low-risk waste dies on autopilot. Medium and high-risk fixes wait for your one click. Nothing moves without a rule YOU set.
◉ Autonomous◉ Reversible◉ Audit-logged
The scan pipeline, in one glance
◉ Real architecture · not marketing
Architecture
The stack
Your team approves. Your AWS account gets scanned. Your waste gets purged — encrypted and audit-logged at every hop.
Your team
Approvals · audit · Slack
Your notifications
Slack · Email · webhooks
AUDIT-LOGGED
PurgeOps Cloud
Detect · Score · Approve engines
Detect engineRisk scorerApprove gate
~200ms per pass
IAM read/execute role
Least privilege · external ID
Your AWS account
EC2 · EBS · ELB · RDS · all regions
READ-ONLY SCAN · MULTI-REGION
02 /What we hunt
The 4 hidden money leaks bleeding your AWS bill dry
Every cost dashboard shows you the same 4 waste types. We're the only one that actually kills them. Here's how.
◆ Compute
The $18/month instance that costs you $18,000/year — killed on sight
CPU sitting at 0.4% for three weeks straight. Zero network traffic. A stopped cron nobody remembers. Multiply that by 40 instances and you're bleeding $8,000/month.
- CPU + network signals — flatline detection across 21 days
- Snapshot-first safety — rollback exists before anything moves
- Instance-family aware — t3.large idle ≠ r5.xlarge idle
EC2 ·
LOW RISKi-0a2b3c4d5eInstancet3.large · eu-west-1
CPU · 21d avg
0.4%Network
0 kbps in/out · 21dEst. savings
$18.40/mo✓ Snapshot verified · rollback available for 30 days
▸ Auto-execute tonight
02:14 UTCApproved◆ Storage
The 500GB volume nobody remembers attaching. Still billing. Not for long.
Detached 47 days ago. Still pulling gp3 rates. We snapshot it first — then reclaim it. Full 30-day rollback. Zero risk.
- Age-based detachment tracking — 30-day quiet period
- Tag-aware — respects purgeops:hold when engineers need one
- Snapshot vault — pre-delete snapshot kept for 30 days
EBS ·
LOW RISKvol-9x8y7z6wStatus
unattachedDetached47 days ago
Est. savings
$40.00/moSnapshot will be created before deletion · 30d retention
▸ Auto-execute tonight
02:14 UTCApproved◆ Network
The load balancer serving 0 requests for 30 days — eating $22/month of pure air
That 'we'll clean it up after the migration' ALB from Q1? It's still running. 0 requests in 30 days. $22/month straight into the void. We flag the DNS refs, wait for your OK, then delete.
- Request-count based — not just health-check traffic
- Target-group aware — checks the whole chain
- DNS record cross-check before delete
ELB ·
MEDIUM RISKalb-legacy-migrationRequests · 30d
0Route53 refs1 CNAME · legacy-app.example
Est. savings
$22.20/mo▸ DNS reference found · needs approvalReview
◆ Database
The r5.2xlarge running at 8% CPU — that's $347/month you're paying for empty racks
90 days of peak-hour data says you're using 18% CPU, 22% memory. Downsize to r5.large. Save $347/month per database. Human approval required — we never touch production without your click.
- 90-day peak-hour analysis — not simple averages
- Multi-AZ + read-replica aware
- Human gate — never auto-resizes production
RDS ·
HIGH RISKdb-prod-analyticsCurrentr5.2xlarge · Multi-AZ
Recommendedr5.large · Multi-AZ
CPU peak (90d)
18%Est. savings
$347.80/moProduction database · human approval required
▸ High risk · one-click approval belowReview
Live in production
03 /Conversational infrastructure
Ask 'why did you flag this?' — get an answer a junior engineer could act on
No cryptic error codes. No 'see documentation' dead-ends. Every finding comes with the WHY, the SINCE WHEN, and the WHAT HAPPENS IF — all written in plain English by the agent that knows YOUR accounts.
- ◉
Every flag comes with its own story
Grounded in your metrics, tags, and account context — not a generic template.
- ◉
Ask why. Ask what if. It knows your data.
"What's the rollback?" "Show me similar." The agent queries your live account data.
- ◉
Powered by Claude Sonnet 5 — deep tool-calling, not a chat wrapper
Tool-calling and structured output — handles ambiguity like an on-call engineer.
Agent
reasoning online · ready
Why did you flag i-0a2b3c?
It's a
t3.large at 0.4% avg CPU over 21 days. Estimated waste: $18.40/mo. Low-risk — snapshot verified.Schedule termination tonightHold for 30 daysShow me similarShow me similar across all accounts
Found 7 idle t3.* instances across your 3 accounts — total $142.80/mo. 4 are low-risk (snapshots verified). Want me to batch-approve those and hold the 3 medium-risk for review?Batch approve the 4Hold the 3Show me the medium-risk
Ask about a resource, rule, or account…Send ↵
Powered by Claude Sonnet 5Live now · Connect your AWS account to try itRisk routing
Decision tree
One score per finding. Low-risk dies on autopilot. Everything else waits for your click — you pick how much human gets involved.
New finding
Low risk
Auto-execute
No human touch
82% of findingsMedium risk
Batch approve
One-click daily digest
14% of findingsHigh risk
Human review
Slack thread + Loom on request
4% of findingsNothing moves without a rule
Zero rules configured means zero actions. Full stop.
Reversible where possible
EBS snapshots before delete. EC2 stop before terminate. 30-day rollback on every low-risk fix.
Every action logged
Audit-grade trail. Streamed to your feed and CloudTrail. Nothing gets lost.
Tiered approval, not all-or-nothing
You pick the line. We never do.
04 /The market gap
Every other tool hands you a to-do list. We hand you back the money.
Cost dashboards charge you $200/month to SHOW you your waste. We only get paid when we actually kill it.
Capability
Visibility tools
◉ PurgeOps
Detect waste across compute · storage · network
Yes
Yes
Executes the fix — not just recommends it
—
Autonomous + gated
Risk-tiered auto-approve
—
3 tiers
Live activity feed of every action
—
Real-time
AI-explained findings + follow-up chat
Partial · reports only
Conversational
Works on any AWS spend — even under $10k/mo (competitors won't talk to you)
— $7k/mo minimum
Any size, any bill
Reversible actions with snapshot vault
—
30-day rollback
05 /Fits your stack
Plugs into the tools you already have open
Slack pings when the agent kills waste. Terraform proposes the rollback. GitHub tracks every action. Nothing new to install.
AWSAmazon Web ServicesLIVE
SLSlackLIVE
TFTerraformLIVE
GHGitHubSOON
PDPagerDutySOON
DDDatadogSOON
Live nowQ4 roadmap
06 /Design partners
6 design partners. 30 days. $47,283 reclaimed.
Here's what happens when you let an autonomous agent do the work you've been putting off for 18 months.
$0
reclaimed · across design partners · first 30 days
0 min
median time to first finding
0 rollbacks
needed · zero incidents in beta
$0/mo
RECLAIMED · WEEK 1
The tiered approval is the thing. Auto-execute on low-risk means the savings actually happen — no more 'next sprint' tickets that die on the backlog.
4.3× ROI
FIRST 30 DAYS
We ran it against staging first, watched the activity feed for a week, then flipped autonomous mode on. Haven't looked back.
0 zombies killed
FIRST SCAN
Set up in one afternoon. By Friday it had killed twelve zombie EBS volumes we'd been ignoring since Q2.
3 hrs saved/week
OPS TIME
Vantage told us what we already knew. This actually did the work — I stopped exporting CSVs to argue about idle boxes.
1 afternoon
TIME TO CONNECT
The plain-English explanation for each finding is what got my team to trust it. Not just 'unused' — the why, the since when, the what happens if.
$0 first month
NET SAVINGS
First month: $2,140 back. Pays for itself in week one on a Growth plan, easily — and we still had a dozen medium-risk items waiting for us to click approve.
Design partners are early-access customers; quotes verified via email exchange.
07 /Pricing
Pay only when we actually reclaim money. Zero minimums. Zero risk.
Zesty won't talk to you under $7,000/month AWS spend. Vantage charges you flat to LOOK at your bill. We only get paid when the agent kills real waste — and there's no minimum, ever.
Pay-on-savings
Starter
25%of savings found
No monthly minimum. Pay 25% of what we reclaim — nothing more. If we find nothing, you owe nothing.
- 1 connected AWS account
- Daily scheduled scan
- Auto-approve on low-risk findings
- Nothing found, nothing owed
Flat rate
Growth
$149/ month
Billed monthly
- Up to 3 connected AWS accounts
- Daily scheduled scan + up to 10 on-demand scans/day
- Auto-approve low-risk findings (medium-risk rules: coming soon)
- API access
- Email support
- No fee on savings, ever
Flat rate
Scale
$299/ month
Billed monthly
- Up to 10 connected AWS accounts
- Daily scheduled scan + unlimited on-demand scans, priority queue
- Auto-approve low-risk findings (custom risk rules: coming soon)
- API access, SSO/SAML
- Dedicated onboarding + Slack channel
- No fee on savings, ever
For larger footprints
Custom
Let's talk
Tailored to your footprint
- Unlimited AWS accounts
- Negotiated scan cadence
- Dedicated infrastructure
- Custom contract & support
08 /Frequently asked
The 6 questions every technical buyer asks — answered honestly
Simple. We invoice 25% of the monthly savings we actually recorded — meaning approved or auto-executed. If we killed $0 of waste this month, your bill is $0. No base fee. No trap.
Never without your click. Low-risk means one thing: unattached volumes, snapshotted idle instances, load balancers with zero traffic AND no DNS refs. Anything else stays in your approval queue. Forever, if you want.
Real conversation. Every finding gets an AI-written explanation grounded in your actual metrics, tags, and account context. Ask why it was flagged. Ask what happens if you approve. Ask for similar findings across your accounts. Powered by Claude Sonnet 5 with deep tool-calling — the agent can query your data and propose actions in the same reply.
Read-only by default. Execute permissions are OPT-IN per action class (compute, storage, network, database) and revocable in one CloudFormation update. Everything auditable via CloudTrail.
From the Plan page, click Manage subscription — that opens the Stripe customer portal where you can cancel, switch plans, update payment method, and download invoices.
No long-term contract on self-serve tiers. Starter has no minimum spend. Growth and Scale cancel when billing goes live. Custom plans are negotiated separately.
No. AWS only, deliberately. Being 'okay' at three clouds means being great at none. Multi-cloud is on the 2027 roadmap — prioritized by real customer demand, not slideware.
Onboarding in under 4 minutes
Stop lighting money on fire.
Connect your AWS account in under 4 minutes. The agent starts hunting waste before your coffee's cold. First findings hit your feed tonight. Money back in your budget by Monday.
NO CREDIT CARD · NO SALES CALL · NO MINIMUM · CANCEL ANYTIME