Scored across 20 tools
Most tools have distinct resource/action targets, and descriptions clarify when to use observe, get_html, find_elements, or screenshot. There is still some overlap around page observation and screenshots, since observe can return screenshot-only output while screenshot also captures the viewport.
The tool set is consistently snake_case under the browser namespace, mostly following verb_noun patterns like create_session, list_tabs, and save_auth_profile. A few names such as observe and screenshot are shorter verb/noun forms rather than strict verb_noun, but the convention remains readable and predictable.
Twenty tools is somewhat heavy but justified for a full browser automation server covering sessions, tabs, actions, observation, downloads, and auth profiles. Each tool has a plausible role, though the surface is near the upper end of what an agent can comfortably navigate.
The surface covers core browser automation lifecycle operations well: sessions, tabs, actions, observation, downloads, JS evaluation, and human takeover. There are minor gaps around auth-profile lifecycle, such as no obvious delete/update or explicit load operation, and tab creation is not directly exposed as its own tool.
