Show HN: Freu CLI – Cut web agent token usage by 90% via compiled browser skills (github.com)

The Freu AI automation suite has launched its first release, introducing the Freu CLI, which dramatically reduces AI agent token consumption by up to 90%. By transitioning from a Just-In-Time (JIT) execution model to an Ahead-of-Time (AOT) compilation, Freu enables users to record web interactions that are then transformed into reusable, deterministic skill commands. This approach allows agents to bypass repetitive visual reasoning, resulting in significant cost savings during execution, as agents no longer need to interpret HTML on each run. This innovation is particularly significant for the AI/ML community as it streamlines complex workflows and enhances operational stability in enterprise settings. Freu's unique capability to compile browser sessions into structured commands, utilizing a semantic Domain-Specific Language (DSL), means that tasks can be performed more efficiently across various platforms. Additionally, the system supports self-healing selectors and smart value retrieval, ensuring that skills remain functional even if the webpage structure changes. With plans for further enhancements, including an OS-level Computer Use Agent, Freu CLI positions itself as a powerful tool for automating and optimizing web-based tasks in an increasingly data-driven world.