Today's AI open source · 2026-11-19
lexmount/moli
Moli is an open-source headless browser for AI agents, built in Rust.
| Repository | github.com/lexmount/moli ↗ |
|---|---|
| Main language | Rust |
| License | Apache-2.0 |
| GitHub stars · forks | ★ 24,019 · 687 forks (as of 2026-11-19) |
| Created · last update | 2026-08-10 · 2026-11-18 |
| GitHub topics | ai-agents, ai-tools, browser, browser-automation, cloud-browser, headless, headless-browser, kitesurf |
Overview
Moli is a headless browser built so AI agents can fetch web pages, extract their content, search the web and automate browser tasks. It is written in Rust and runs on Linux, macOS and Windows.
Its defining feature is a design that turns on layout and rendering only when needed. By default it works DOM-first and exports pages as Markdown or as a semantic tree that models can read easily; with the --layout option enabled, it also produces screenshots, full-page PNGs and PDFs. Besides the CLI, it can be controlled through CDP, WebDriver Classic and WebDriver BiDi, making it easy to connect to existing automation tools.
It suits developers building agents that read and work with web pages, and practitioners who need to pull page content directly as text in crawling or scraping pipelines. The project also explains how to install the repository's skills directory into an agent and use it right away.
Why it is worth a look
What stands out is that it redesigns the browser to deliver a structure agents can read first, rather than drawing a screen for humans. It appears to be an attempt to reduce the burden of launching a heavy browser when agents work with the web, by drawing pixels only when needed. Releasing it alongside a Skill that agents can install and use themselves also fits how agents are used today.
Key features
- moli fetch extracts a page as Markdown or as model-friendly semantic tree text.
- With the --layout option enabled, it produces viewport and full-page screenshots and paginated PDFs.
- --eval runs interactions such as button clicks and then exports the resulting page.
- moli serve launches an automation server controlled through CDP, WebDriver Classic and WebDriver BiDi.
- Supports profile, proxy, resource policy and tracing options.
What it is used for
- Used to fetch page content as text so AI agents can read and summarize web documents.
- Used to let agents automatically handle browser tasks that require form input or clicks.
- Used to convert pages to Markdown in web crawling and scraping pipelines.
- Used to create page screenshots or PDFs when needed.
Getting started
- On Linux or macOS, install with curl --proto '=https' --tlsv1.2 -fsSL https://github.com/lexmount/moli/releases/latest/download/moli-installer.sh | sh .
- On Windows, run powershell -ExecutionPolicy ByPass -c "irm https://github.com/lexmount/moli/releases/latest/download/moli-installer.ps1 | iex" in PowerShell.
- Try extracting a page as Markdown with moli fetch --dump markdown --wait done https://example.com .
- If you need an automation server, run moli serve; if you need screen coordinates or screenshots, launch it with moli serve --layout.
- To hand it to an agent, have the agent install the Skill from the repository's skills directory, following the prompt in the README.
What to weigh
- It is licensed under Apache-2.0, so commercial use is allowed, but conditions such as including notices and a copy of the license must be met.
- The project was created in August 2026, so CLI options and the API may change.
- The README's 'production-ready' wording and the official site's description that it 'starts faster and uses less memory than Chrome' are the developer's own claims, and the official site itself recommends Chromium for tasks where extensions or faithful screen rendering matter, so verify it on your own pages.
- Installation runs a remote script directly in the shell, so it is a good idea to review the script's contents before running it.
Related terms and companies
Sources
Written with AI assistance from the repository README and GitHub data. Star counts and other numbers are GitHub values as of the date shown; features and licensing can change, so check the repository.