No magic numbers
Cost basis, authority mode, formula, versions, adjustments, and reconciliation outcome travel with the result.
Open-source cloud cost accounting
Know what your software costs. Keep the evidence.
Turn cloud billing exports into exact, reconciled service costs today, then build explainable customer and product economics on top.
The question
Cloud invoices can tell you what a provider charged. They rarely give engineering, finance, and product one trustworthy path from that charge to a business decision. Ledger builds that path without hiding the calculation.
Why Ledger
Every number is materialized from declared inputs, reconciled across boundaries, and kept close to its raw billing evidence.
Cost basis, authority mode, formula, versions, adjustments, and reconciliation outcome travel with the result.
A FOCUS-aligned core keeps financial semantics stable while namespaced extensions retain useful AWS, Azure, and GCP detail.
Superseding bills create new immutable datasets and publications. Earlier evidence and calculations remain inspectable.
Provider billing logic produces Cost Pools. Business rules decide where those already-calculated costs should flow.
One accountable pipeline
Ledger keeps ingestion, provider calculation, and business attribution as distinct responsibilities, joined by versioned contracts.
Accept immutable billing files with checksums, source metadata, and organization ownership.
Convert provider rows into a readable FOCUS-aligned financial core without losing origin.
Materialize exact service and resource costs from a declared authority and cost basis.
Prove source, normalized, calculated, and published totals agree before release.
Expose formulas, versions, evidence links, corrections, and assumptions with every result.
See the result
Scan current costs, separate gross usage from credits, drill into resources, and follow every published amount back through calculation and normalization.
Explainability is a data model
Ledger does not generate prose around an opaque total. It persists the lineage required to reproduce the number: immutable evidence, normalized fields, calculation contracts, reconciliation outcomes, and publication identity.
Review the auditability modelBring the bill you have
Every supported adapter emits the same Normalized Charge contract and publishes a capability report, so unsupported semantics cannot hide behind a successful upload.
The preferred provider-neutral path with preserved standard columns, origins, validation findings, and extensions.
Service and resource detail with Savings Plan usage, fees, and unused commitment kept explainable.
Actual-cost exports normalized with subscription, meter, resource, region, and adjustment context.
Projects, services, SKUs, resources, locations, credits, and corrections in the shared financial model.
Useful now, designed for what follows
Ledger currently establishes trustworthy provider costs and publishes immutable Cost Pools. The next domain adds business entities, measurements, and allocation rules without teaching provider calculators about customers or products.
Explore the attribution roadmap$ bin/kamal deploy INFO building ledger:3f28a9 INFO running release health checks PASS https://ledger.example.com/up $ bin/kamal app details Role Host Status web 10.0.0.12 running
Operate it without a platform team
Agent Mode runs locally with SQLite and minimal infrastructure. Production uses PostgreSQL, horizontally scalable Rails processes, background jobs, and an optional open observability stack. No Kubernetes assumptions are baked into the product.
Straight answers
No. It is an accounting substrate for producing and explaining service costs, then attributing those costs through a separate business model.
No. Financial calculations are deterministic. AI may suggest mappings, summarize reports, or recommend investigations, but it does not produce the numbers.
Not currently. Ledger is open source and self-hosted. The repository includes local development and Kamal production deployment paths.
Yes. Corrections create a new immutable import and publication while previous evidence, normalized data, calculations, and explanations remain available.
The business attribution and unit-economics layer is designed but not yet implemented. Parquet import and automated provider connectors also remain roadmap work.
Start with evidence
Run Ledger locally, upload the sample billing dataset, and inspect a reconciled service-cost report in minutes.