BYOK (Bring Your Own Key)

Running replicated experiments means real API spend: hundreds or thousands of model calls per study. Under BYOK the platform orchestrates those calls but never resells them — your key, your provider account, your bill. This keeps pricing transparent (you pay published provider rates), keeps spending controls in your hands (provider-side limits apply), and avoids the lock-in of platform-denominated credits.

BYOK also matters for research governance. Institutions often require that API usage run under institutional accounts for data-handling and procurement reasons; a BYOK design satisfies this without workarounds. And because the key identifies the account, usage remains auditable in the provider's own dashboard, independent of the platform's records.

The practical companion to BYOK is up-front cost estimation: since the researcher pays per trial, a serious platform computes the expected number of trials (conditions × models × replications) and the estimated cost before dispatching anything, and offers cheap pilot runs — one trial per cell — to sanity-check a design before committing the full budget.

AIScholar uses BYOK for all subject-model trials via OpenRouter — your OpenRouter key, billed by OpenRouter, with keys encrypted at rest (AES-256-GCM), cost estimates before every run, and pilot mode for cheap design checks. The AI wizards that assist you are separate: they are included in your AIScholar plan and need no key.

• Experiment execution features
• Cost & plan questions in the FAQ

• Method Specification Prompt (MSP)A self-contained method specification precise enough for independent re-implementation.
• Factorial & fractional factorial designFull vs. fractional crossing of factor levels, aliasing trade-offs, and when to use each.
• Machine psychologyStudying LLMs as behavioral subjects with experimental methods from psychology.

Browse all 11 glossary terms