Evals
Learning modulesLibraryCreate an evalCost calculatorDocsprefactor.ai
Docsprefactor.aiGet started, free
  1. Prefactor
  2. Prefactor Evals
  3. Eval library
Eval library

What do you want to check?

Describe the failure you are worried about, or the thing you want to be sure of. The library finds the evals that check it, from Prefactor, the community, and open-source projects, each with a pass and a fail from a real agent.

Try:

Submit an eval

Sent straight to Prefactor, the same way as the contact form on prefactor.ai. Approved entries are published with your name, marked as community, and join the generator's source set. Open-source evals are welcome too: give the project and a link, and the entry will carry both.

NextCreate a set for your own agentDescribe it in two fields and get five evals drawn from this library.ThenWhat does running these cost in production?Three numbers in, one number out, and what you keep for building.

Runtime evaluation for AI agents

Get started, free
On this siteLearning modulesEval libraryCreate your own evalCost calculator
Prefactorprefactor.aiDocumentationPricingBlog
ElsewhereGet Ready for AgentsAgent failures feedAgents After Dark podcast