Benchmark report
Method and results for Codna against Cursor on 87 matched bug-fix cases.
Frequently asked
On 87 matched bug-fix cases against Cursor, Codna used 5× fewer tokens and ran 1.7× faster. Both verified 87 of 87 fixes. Codna averaged about $0.02 of model spend per fix.
Total tokens, wall-clock time, model cost and verified fixes per case, averaged over 87 matched cases run on identical checkouts.
Codna maps the repository without a model, for zero tokens, and hands the agent an evidence bundle scoped to the issue. It prints the raw-to-bundle size on every run.
The repository map is built from import patterns, so it is not tied to one language. codna impact narrows a diff to the tests it can affect. On-device recall covers Python, JavaScript, TypeScript, Go, Rust, Java, C, C++, C#, PHP and Ruby.
Codna locates the affected code with its dependency and blast-radius graph, fixes from the evidence bundle, and reports root cause, confidence, blast radius and regression risk. On GitHub, it opens a pull request that explains itself. You merge.
Understanding runs on your machine. Only the evidence bundle or the diff reaches your model provider, with your key. Set privacy.egress to fail-closed in codna.yaml and Codna runs your tests only under kernel-level network denial.