Browser Use documents that screenshots go to the agent's own model as rendered, with use_vision=False or judge=False as the controls. The judge receives up to four images in total.
By [RuntimeWire Staff](https://runtimewire.com/author/runtimewire-staff)
· Published
Co-founders Magnus Muller and Gregor Zunic built Browser Use, the open-source framework that helps AI agents operate websites through a browser. Browser Use, which announced a $17 million seed round led by Felicis Ventures in March 2025, released version 1.0.5 on November 14th, 2026. The 1.0.5 release, published by Sarath Menon at 08:24 UTC on November 14th, adds a default post-run judge that receives browser screenshots exactly as captured, while only the accompanying text is redacted. The related pull request was also merged on November 14th, 2026.…