A realistic web environment for building autonomous agents.

Continuous and scalable web agent evaluation in evolving environments.

Evaluating multimodal agents on realistic visual web tasks.

Benchmarking LLM agents on consequential real-world tasks in a simulated company.