SPEND PROOF / FREE AUDIT
What does a successful result cost?
Compare two versions of an agent on the same tasks. Include models, tools and every retry. Spend Proof calculates cost per declared successful result and flags observed regressions.
Calculated in your browser. The field content is neither sent to the server nor stored. Do not include keys, prompts or personal data.
Success is the outcome you evaluated and labeled success=true. HTTP 200 does not establish quality. Default minimum: 30 distinct tasks per version. The result remains a candidate for a controlled trial, never deployment authorization.
Save and compare your analyses
Projects keeps aggregate results in a private workspace and compares versions of your projects. Start with 3 free saved reports, no card required. The subscription includes 10 projects: CHF 19 or EUR 19/month for 100 new reports per month, or CHF 190 or EUR 190/year for 1,200 reports per year. Renews automatically. To save an analysis, open Projects and import your measurements.