codeburn

Based on the available evidence, codeburn is best understood as an AI coding cost observability tool. It is not a code generator, not an IDE copilot itself, and not primarily a model gateway. A better analogy is a local analytics layer for AI coding spend, helping developers audit where tokens and money go across tools, models, projects, and tasks.

Its practical value is consolidating usage records that are otherwise scattered across different AI coding products. The GitHub repo title explicitly says it tracks usage and cost across many tools and agents with breakdowns by model, project, and task. The Zhihu posts repeatedly mention live monitoring, reports, provider filters, and exports like PDF/CSV/JSON. In terms of evidence strength, the official repository is the strongest proof of what the product is; the articles are useful as secondary walkthroughs, but they do not yet provide broad independent proof of how well it works in varied real-world setups.

On cost and setup, the evidence supports that it is free, local, and open source, but does not support any hard claim about guaranteed savings or ROI.