APIEval-20

APIEval-20 is an AI benchmark from KushoAI that measures how well API-testing agents can find bugs when given only a JSON schema and a single sample payload. It runs an agent's generated test suite against live reference APIs that contain planted bugs, then scores bug-detection accuracy, coverage, and efficiency objectively. The benchmark spans 20 scenarios across 7 domains, so results reflect performance on varied real-world API shapes rather than a single case. For developers and QA teams building or evaluating AI test-generation agents, APIEval-20 provides a repeatable yardstick to compare approaches and see where an agent misses edge cases. It's positioned as a research-grade benchmark for the API-testing-agent space rather than a tool you point at your own production APIs.

APIEval-20 is used by developers and QA teams to benchmark AI API-testing agents on bug detection, coverage, and efficiency using only a schema and a sample payload.