FinOps, run by agents.
In your cloud.
Kodly builds AI agents that run FinOps for large enterprises, inside your own cloud.
Join the waitlist| ChargePeriodStart | ServiceName | RegionId | ChargeCategory | BilledCost | EffectiveCost |
|---|---|---|---|---|---|
| 2026-09-03 | Amazon EC2 | eu-west-1 | Usage | 41,206.18 | 29,884.02 |
| 2026-09-03 | Amazon S3 | eu-west-1 | Usage | 6,118.40 | 6,118.40 |
| 2026-09-03 | Amazon VPC NAT Gateway | eu-west-1 | Usage | 9,412.77 | 9,412.77 |
| 2026-09-03 | Amazon RDS | eu-west-1 | Usage | 12,930.55 | 9,044.11 |
| 2026-09-03 | AWS Savings Plans | global | Purchase | 18,500.00 | 0.00 |
| 2026-09-03 | Azure Virtual Machines | westeurope | Usage | 22,704.90 | 17,311.06 |
| 2026-09-03 | OpenAI API | global | Usage | 7,388.12 | 7,388.12 |
- Finding
- NAT Gateway data processing in eu-west-1 rose 212% day over day.
- Root cause
- A batch job in account 4417 started reading from S3 through the NAT gateway on 02 Sep. The VPC has no S3 gateway endpoint, so every byte is billed at $0.045 per GB.
- Recommendation
- Add an S3 gateway endpoint to vpc-0b3e in eu-west-1. Route table change only. No downtime.
- Impact
- About $6,400 a day at current volume.
- Autonomy
- Recommend only. Waiting for approval.
Customers
Enterprises that trust Kodly to run their autonomous FinOps practice.
The problem
Cloud cost is a data problem that enterprises still solve by hand. Analysts pull billing exports, reconcile commitments, chase tags, explain anomalies, and repeat it next month. The bill keeps growing. Now it includes SaaS and AI spend that no legacy tool was built to read.
General-purpose models don't fix this. They know cloud pricing from blog posts, hallucinate SKUs, misread a billing export, and have no idea what you committed to.
What we do
Kodly agents are different by construction.
- Models
-
They run on open-weight models we fine-tune on the substrate of FinOps: billing schemas, FOCUS, pricing and commitment mechanics, allocation logic, unit economics across AWS, Azure, GCP, SaaS, and AI spend.
- Harness
-
They run in a harness built for the job: connectors to your billing and usage data, evaluation suites that measure correctness on real cost questions, memory of your environment, guardrails on any action that changes spend, and a full audit trail.
- Coverage
-
The agents cover the whole FinOps workload.
- Anomaly detection and root cause
- Commitment and rate optimization
- Allocation, tagging, and showback
- Forecasting and budget tracking
- Autonomy
-
You decide, per task, how much autonomy the agents get.
- Deployment
-
Everything deploys inside your cloud. Your billing data never leaves your account. No SaaS, no shared tenant.
Early access
Kodly is independent, founded in 2026, and in production with early enterprise customers. Early access is open.