Show HN: WebCMD, a reusable CLI command layer for browser agents (github.com)

WebCMD has been introduced as a groundbreaking tool designed to enhance browser agent automation by learning the navigational context of websites. This self-learning infrastructure allows AI agents to compile knowledge into deterministic commands, significantly reducing browser token spend by up to 90%. With WebCMD, agents no longer need to repeatedly rediscover the same sites, streamlining the automation process and driving efficiency. The tool incorporates three layers of learning: live browser control for unfamiliar sites, sitemap memory for familiar sites with unknown actions, and CLI authoring for known action spaces. Each layer decreases complexity and cost for the subsequent one, allowing future agents to utilize pre-defined commands for common tasks rather than navigating anew. This not only improves performance but also enhances the scalability of automation tasks across various applications, including research, social media monitoring, and e-commerce. WebCMD's capability to operate with authenticated sessions, APIs, and desktop applications further positions it as a versatile resource for developers and researchers in the AI/ML community.