lexmount/moli

Moli is a headless browser for AI agents that prioritizes structure extraction and on-demand rendering over continuous visual state maintenance.

The tool addresses the mismatch between what browser automation tasks actually require and what traditional headless browsers provide. Most automation workloads need page structure and DOM access rather than continuously rendered pixels. Moli inverts the typical architecture by treating the native DOM and style state as the primary source of truth, triggering layout and paint operations only when explicitly needed. This structure-first approach eliminates the overhead of maintaining a fully rendered visual state while still supporting JavaScript execution, DOM manipulation, CSS processing, networking, storage, and standard automation protocols.

Moli suits teams building AI agents that need to fetch pages, extract content, perform web searches, or automate browser tasks at scale. It works through multiple interfaces: a CLI for direct page extraction and rendering, and an automation server supporting CDP, WebDriver Classic, and WebDriver BiDi protocols, allowing integration with tools like Playwright. The tool runs on Linux, macOS, and Windows. Choose it when resource efficiency and fast structure extraction matter more than pixel-perfect rendering, or when you need both capabilities without the overhead of maintaining continuous visual rendering.

The project shows active development with regular commits addressing core functionality and protocol support. Work spans multiple areas including rendering pipeline improvements, protocol implementation refinement, and cross-platform compatibility. The maintainers engage with the community through a Discord channel and respond to issues and feature requests.