Demo
See it in action
Netway identifies every dollar of avoidable network spend in your AWS account — and gives your team the exact command to fix it.
Deploys in 5 minutes · No agents · No code changes
The waste is real. The data exists. But nothing connects the bill to the decision that caused it.
AWS tells you how much you spent on Data Transfer. It doesn't tell you which workload, which team, or which architecture decision caused it. VPC Flow Logs exist — but nobody reads raw log files.
Engineering says the architecture is correct. Finance sees the bill trending up. The engineer who chose a private subnet didn't know they were creating a $500/month NAT invoice. Nobody owns the gap between them.
Every new service, every new region, every new ML training job adds to the baseline. Without detection, network cost grows linearly with your engineering team — and nobody notices until the bill is already large.
No agents. No code changes. No access to your application.
One CloudFormation stack. Lambda function, Athena workgroup, S3 bucket — all created automatically.
The agent is open source — audit it on GitHub → 5 minutesNetway enables VPC Flow Logs on the VPCs you choose automatically. Traffic data flows to S3 in real time.
AutomaticLambda queries Athena across 30 days of traffic, enriches flows, then ships aggregated summaries to Netway. Detectors run server-side — each looking for a specific cost pattern.
Weekly scanFindings appear on your dashboard with a named resource, a monthly dollar amount, and the exact fix command.
Per finding
14 days free. No credit card required. Your first scan runs automatically after deploy.
Netway detects 14 categories of avoidable network spend — from the most common to ML-specific patterns.
High-volume data leaving your VPC directly to the internet at $0.09/GB. CloudFront reduces this to under a cent per gigabyte.
GPU training jobs writing model checkpoints to S3 through NAT. The data volumes are 10x larger — so is the waste.
Workloads reading from S3 buckets in another region pay $0.02/GB in cross-region transfer fees plus added latency.
Multi-GPU distributed training with nodes in different AZs. Every gradient synchronisation is a cross-AZ transfer charge.
Your workloads are paying internet transfer rates to access your own S3 data. A free VPC endpoint eliminates this entirely.
Model weights loaded from S3 on every inference cold start via NAT. A VPC endpoint cuts both the cost and the cold start latency.
Application servers in one availability zone querying a database in another. Every query is charged at $0.01/GB each way.
Calls to SSM, CloudWatch, Secrets Manager, and STS going through your NAT Gateway — for traffic that never needs to touch the internet.
Instances routing to a NAT Gateway in a different AZ pay both NAT processing and cross-AZ transfer fees on every outbound byte.