You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
OpenCLI's commands run as site commands. With Node.js 20+ on PATH, site update installs a pinned OpenCLI (1.8.8, about 180 sites and 1,300 commands) into ~/.chrome-use/opencli with npm install --ignore-scripts; no token is needed. A name/cmd that neither of our packs has runs through OpenCLI's own runtime, but every browser step goes through chrome-use, so it uses your current session and logins: chrome-use site hackernews/best --limit 5 --json. They are marked (opencli) in site list, site info shows their args, and the site hint lists them after ours. Our adapters win on a shared name. AGENT_BROWSER_SITES_NO_OPENCLI=1 turns this off. (#407)
site analyze [url] shows where a page's data comes from. It lists the same-site API calls the page made, the state it embeds (__NEXT_DATA__, __INITIAL_STATE__, JSON script tags) and any anti-bot vendor. It then recommends reading the site's own API from the page, then embedded state, then the DOM, the order in which they break least, and lists next steps. Do the action that loads the data first, then analyze. (#407)
site verify <name>/<cmd> [args] catches a broken adapter.--write-fixture records the shape of a good result in ~/.chrome-use/site-fixtures/ (types only, no values). Later runs fail with exit 1 when a field disappears, changes type, or a list comes back empty. Works for OpenCLI commands too. (#407)