BrowserKitten 1.2.0: a Chrome sidebar agent shipped as an unpacked extension

BrowserKitten is a Chrome sidebar agent that reads pages you are already logged into, performs clicks and form controls, and saves results as documents, webpages, or files. Version 1.2.0 is distributed as a loadable unpacked extension. According to the README, there is no Node.js install, Python runtime, or build step required for normal use. Installation follows Chrome’s standard unpacked workflow: download the source ZIP, extract it to a folder you intend to keep, enable Developer Mode at chrome://extensions, and load the folder that directly contains manifest.json, icons/, and src/. The minimum Chrome version listed is 135. That folder must stay in place because Chrome reads extension files from it directly, and updates are applied by overwriting the same folder and reloading the extension. Two separate model configurations The design separates reasoning from page execution. An inference model handles task understanding, page analysis, planning, and tool calls, and must support tool calling. A second configuration, labeled System1, handles clicks, fills, and selection controls using an OpenAI-compatible /chat/completions endpoint. The two are configured independently; providing only an inference key will not enable page actions, and the README documents a NEED_JEV error for that case. System1 includes a probe that sends a small non-operating request to check the endpoint and response format, with an explicit caveat that a successful probe does not validate all websites or visual inputs. The sidebar also includes a point-selection control for targeting part of a page, and optional settings for image generation, transcription, and additional search or fetching. Users can supply their own API keys and Base URLs, or sign in with a Browser Kitten account. The account backend is not included in this repository, and available models, quotas, and plans are shown after login. For the unpacked build, the README states no Browser Kitten login or purchase is required, while external