SHAREPI / HARNESS PERFORMANCE
Our agents. Measured.
Loading recorded results…
Loading performance graphs…
A GOOD PLACE TO START
Built for the work we do.
THE BENCHMARK LIBRARY
Find your next test.
Choose a capability. Look under the hood.
Build a clearer picture of your agent.
A curated catalog of published benchmarks, not an exhaustive registry. “Needs integration” means the benchmark is available upstream but has no runner connected here. Scoring labels describe the evaluator; agent runs can still vary. How to read these labels ↗
FROM CAPABILITY TO CONFIDENCE
A score is a start.
The evidence is the point.
Start with a repeatable task. Inspect what passed, what failed, and what it cost. Then make the next run better.
Open AppWorld runner ↗CONNECTED RUNNER / APPWORLD
Put your agent to work.
Everyday Agent Benchmark · simulated apps, verifiable outcomes.