Open-source carbon and energy libraries for engineers

Engineering teams increasingly embed emissions estimates directly into their own products: a website carbon calculator, a CI pipeline that reports the cost of a build, a dashboard that shows emissions per model or per customer. Open-source libraries make this possible without a commercial contract, but they shift a set of obligations onto the team that adopts them: the licence, the maintenance of the project, the data behind the calculation and the reproducibility of the number that eventually reaches a report.

The libraries differ in scope. CO2.js converts bytes or page views into estimated emissions for web and digital services. CodeCarbon measures energy and emissions from compute workloads. BoaviztAPI models the life-cycle impacts of hardware. The SCI specification, published by the Green Software Foundation, provides the method that ties energy, location-based carbon intensity and embodied emissions into a single rate. Each library depends on external datasets for grid intensity or embodied factors, and those datasets change.

The common failures are engineering failures rather than modelling failures. A licence that is incompatible with a hosted product, a dependency that has not been released in years, a hard-coded national grid average that cannot be replaced with a regional or hourly dataset, or an output that cannot be reproduced because the factor version was never pinned. Method documentation is often thin, and units are easy to misread.

Evaluate libraries on licence compatibility, maintenance activity, data source replaceability, integration cost and reproducibility. A library that cannot be audited is not cheaper than a vendor, it just moves the work.