Find your AWS waste in seconds. A read-only CLI that reports idle and orphaned resources with dollar estimates — the Trusted Advisor cost checks you shouldn't need a Support plan for.
$ awswaste scan --profile prod
Account 123456789012 · regions: us-east-1, eu-west-1
┌──────────────────┬───────────────────┬───────────┬─────────┬──────┬────────────────────────────────────────┐
│ CHECK │ RESOURCE │ REGION │ $/MO │ CONF │ DETAIL │
├──────────────────┼───────────────────┼───────────┼─────────┼──────┼────────────────────────────────────────┤
│ ebs-unattached │ vol-0a1b2c3d4e5f │ us-east-1 │ $400.00 │ High │ 5000 GB gp3, available since 2025-11-02│
│ nat-idle │ nat-07f1a… │ us-east-1 │ $32.85 │ Med │ 0 bytes moved over 30d (threshold 1MB) │
│ eip-unassociated │ eipalloc-99887766 │ eu-west-1 │ $3.65 │ High │ Elastic IP 52.10.10.10 not associated │
└──────────────────┴───────────────────┴───────────┴─────────┴──────┴────────────────────────────────────────┘
Estimated monthly waste: $436.50 (3 findings)
prices: estimates, updated 2026-06 · fix commands in --format json (never executed)
# Go install (works today)
go install github.com/angga-22/awswaste/cmd/awswaste@latest
# Homebrew (from the first tagged release)
brew install --cask angga-22/tap/awswaste
awswaste scan --profile prodawswaste uses the standard AWS credential chain — the same profiles/SSO sessions
your aws CLI already uses. No passwords, no keys baked into the tool.
Common flags:
| Flag | Purpose |
|---|---|
--profile <name> |
AWS shared-config profile (default: env/default chain) |
--region <list> |
Comma-separated regions to scan (default: all enabled regions) |
--format table|json|markdown |
Output format (json and markdown include the fix commands) |
--days <n> |
Age/idle window in days (default 90) |
--idle-bytes <n> |
Idle threshold for NAT traffic over the window (default 1 MiB) |
--min-savings <n> |
Hide findings below this monthly $ amount |
--only / --skip |
Run or skip specific checks, e.g. --only ebs-unattached,eip-unassociated |
--prices <file.yaml> |
Override the embedded price table |
--exit-code |
Exit 3 when unsuppressed findings exist — gate a CI job or cron alert on waste |
--config <file.yaml> |
Config file (default: ./.awswaste.yaml if present); flags always win |
--write-baseline |
Accept all current findings into .awswaste-baseline.yaml; future scans gate on new waste |
--show-suppressed |
List suppressed findings instead of the one-line summary |
--org --role-name <role> |
Scan every account in your AWS Organization (docs) |
--accounts <ids> --role-name <role> |
Scan an explicit list of accounts |
--role-arn <arn> |
Scan one account through an assumed role |
awswaste scan --org --role-name awswasteReadOnlyOne command: accounts are discovered via AWS Organizations, a read-only role
is assumed in each, findings come back grouped per account (worst first) with
a fleet-wide total. Unreachable accounts are skipped with a note, never fatal.
Role deployment (one StackSet / Terraform module) and the hub/spoke policies
are in docs/multi-account.md.
An account with 40 known findings is red forever — nobody can gate on that. awswaste separates the debt you've accepted from the waste someone just added:
awswaste scan --write-baseline # accept today's findings (.awswaste-baseline.yaml, commit it)
awswaste scan --exit-code # exit 0 — until someone adds NEW waste (then 3)Human-authored suppressions (with reasons, and optional expiry) live in
.awswaste.yaml, which can also hold default flags:
regions: [us-east-1, eu-west-1]
min-savings: 1
suppress:
- check: ebs-unattached
resource: vol-0abc123
reason: "DR volume — keep until Q3 migration"
expires: 2026-09-30 # finding reappears after this date
- check: nat-idle
account: "111122223333" # optional narrowing; empty fields match anything
region: us-east-1Suppressed findings never disappear silently — every report carries one honest
line (12 finding(s) suppressed (baseline) — $36.08/mo), and --show-suppressed
lists them. Exit codes are locked: 0 clean or all suppressed, 1 execution
error, 3 unsuppressed findings with --exit-code. A ready-to-copy GitHub
Actions workflow is in docs/ci.md.
| Check | What it flags | Confidence |
|---|---|---|
ebs-unattached |
EBS volumes in the available state (attached to nothing) |
High |
ebs-gp2-gp3 |
In-use gp2 volumes — gp3 is ~20% cheaper, migrates online, deletes nothing | High |
eip-unassociated |
Elastic IPs with no association (idle EIPs bill hourly) | High |
ebs-snapshot-orphaned |
Self-owned snapshots older than --days and not referenced by any AMI |
High |
ec2-stopped-ebs |
Instances stopped > --days whose EBS volumes keep billing |
Med-high |
nat-idle |
Available NAT gateways moving < --idle-bytes over the window |
Med |
elb-idle |
ALB/NLB load balancers with ≈0 traffic over the window | Med |
logs-no-retention |
Log groups that never expire and already store data (savings is a ceiling) | High |
ami-unused |
Self-owned AMIs older than --days referenced by no instance, launch template, or ASG |
Med |
rds-idle |
DB instances with 0 connections over the window (stopped ones auto-restart after 7 days!) | Med |
ec2-old-generation |
Running prev-gen instances with a same-size, strictly-cheaper x86 successor (t2→t3, m4→m5, …) | High |
Skip a resource with the tag awswaste:ignore=true.
Deferred (scoped out on purpose, not forgotten): EC2 right-sizing (needs
memory metrics for trustworthy calls), S3 lifecycle gaps (needs object-age
analysis), Aurora Serverless idle detection, classic ELBs and ALB
target-health inspection, and continuous monitoring.
The full v2 plan is in docs/design/v2.md.
awswaste only calls Describe* / List* / Get* APIs — nothing it does can
mutate a resource. The fix hints are printed (in --format json), never executed.
Build trust before you build automation.
The full least-privilege policy is in docs/iam-policy.json:
{
"Version": "2012-10-17",
"Statement": [{
"Sid": "awswasteReadOnly",
"Effect": "Allow",
"Action": [
"sts:GetCallerIdentity",
"ec2:DescribeRegions", "ec2:DescribeVolumes", "ec2:DescribeAddresses",
"ec2:DescribeSnapshots", "ec2:DescribeImages", "ec2:DescribeInstances",
"ec2:DescribeNatGateways",
"ec2:DescribeLaunchTemplates", "ec2:DescribeLaunchTemplateVersions",
"autoscaling:DescribeLaunchConfigurations",
"elasticloadbalancing:DescribeLoadBalancers", "elasticloadbalancing:DescribeTags",
"cloudwatch:GetMetricData",
"logs:DescribeLogGroups",
"rds:DescribeDBInstances"
],
"Resource": "*"
}]
}Every action awswaste calls is in that list — nothing more. Resource: "*" is
unavoidable for Describe APIs, but every action is read-only. If your
credentials only cover some regions (e.g. an SCP or session policy restricts
them), awswaste collapses the rest to one note: N region(s) skipped line
rather than failing.
| Tool | Niche | Why awswaste |
|---|---|---|
| AWS Trusted Advisor | Cost checks behind a Support plan | No paywall, runs locally |
| AWS Compute Optimizer | Right-sizing only | Covers orphaned/idle waste too |
| komiser | Dashboard you host | Zero-setup CLI |
| Infracost | Pre-deploy Terraform estimates | Scans what's actually running |
| cloud-custodian | Policy engine | Single command, ranked $ savings + fixes |
The wedge: a zero-setup local CLI that ranks waste by dollars and hands you the fix command.
Scale: multi-account / AWS Organizations— shipped:--org.Guardrail: config file, suppression baseline, CI gate— shipped:.awswaste.yaml+--write-baseline. GitHub Action repo still to come.Checks: RDS idle, CloudWatch Logs retention, unused AMIs, old-generation EC2— shipped.Accuracy: price table generated from the AWS Pricing API—tools/genprices+ monthly refresh workflow shipped (needs apricing:GetProductsrole to run; embedded table remains hand-estimated until the first generated snapshot lands).
Savings are estimates, not invoices. awswaste ships a small embedded unit-price table (region-approximate, on-demand USD) and labels every report accordingly. Verify against the AWS Pricing page for your region before acting.
One known upper bound: AWS bills EBS snapshots on incremental stored bytes,
which the API doesn't expose — ebs-snapshot-orphaned estimates from the full
volume size, so its $/MO is a ceiling, not a midpoint.
awswaste is an independent open-source project and is not affiliated with, endorsed by, or sponsored by Amazon Web Services, Inc. "AWS" and all related marks are trademarks of Amazon.com, Inc. or its affiliates.
MIT