The Mission

SanityHarness exists to provide high-signal, agent-agnostic evaluation for AI Agents. In a sea of noise and cherry-picked benchmarks, we strive for isolation, weight, and clarity.

Sponsors & Support

Running comprehensive evals is expensive. We appreciate API credits (GPT-5, Opus, etc) or donations to sustain the harness.

Contact

mim7 on Discord
Have spare API keys? Temporary or unused keys for top-tier models (Opus 4.5, GPT-5.2) are extremely helpful for testing more agents, models and running more evals.

Please DM mim7 on Discord or contact me by email.