Lavish is a CLI editor for HTML artifacts that enables human-agent collaboration by opening agent-generated HTML files in a local browser for interactive review and editing.
The tool addresses the friction in human-AI collaboration loops around rich HTML content. When agents generate HTML artifacts, feedback traditionally requires screenshots or lengthy descriptions of desired changes, losing the interactivity that makes HTML valuable. Lavish solves this by providing a local-first interface where users can pinpoint specific elements, select text ranges, edit embedded Mermaid diagrams as interactive whiteboards, and send annotated feedback back to the agent without leaving the editor. The approach keeps the core feedback loop entirely local with no cloud dependency, while offering optional hosted sharing through a third-party service for those who need it.
Developers should adopt this tool if they work in agent-assisted workflows that produce HTML visualizations and want to streamline the feedback cycle. It suits projects involving product planning, technical documentation, design exploration, or any scenario where agents generate interactive diagrams or layouts. The tool is designed as an AXI, meaning it runs as a simple CLI that agents can execute without setup, optimized for token efficiency through TOON output and long polling. It works across macOS, Linux, and Windows.
The project maintains active CI and release automation. Development activity shows consistent engagement with both integration testing and structured release processes.
