GitHub radar
A browser for AI agents that uses 10x less memory than Chrome
Moli is a separate browser built specifically for AI agents: it reads page structure and clicks buttons while using far less memory than Chrome, and it works with your existing Playwright and Selenium scripts unchanged.
It's an open-source browser made for AI agents, not for humans browsing the web. It reads the page's structure and only renders pixels when something actually needs them.
When it helps
Useful when your agent compares prices, collects news, or fills out forms, especially with many parallel sessions running. For visually complex sites, games, Canvas, or WebGL, stick with a real browser.
Pros
- Uses about 73 MB per site vs about 773 MB for headless Chrome
- Works with existing Playwright and Selenium scripts, no rewrite needed
- Handles many parallel agent sessions without exhausting memory
- Renders pixels only on demand — for screenshots, not every step
Cons
- Passes 81.9% of a browser benchmark vs 99.85% for real Chrome
- Not suited for Canvas, WebGL, or visually complex sites and games
How to set it up — step by step
- 1Open your coding agent (Claude Code, Codex) inside your project folder.
- 2Send the agent the text from the block below — it installs Moli and checks the setup.
- 3Check that the agent shows text fetched from example.com — that confirms it works.
- 4Point your Playwright or Selenium scripts at Moli over CDP instead of Chrome.
- 5For screenshots or heavy graphics, run Moli with layout enabled, or just use a real browser.
Text for your agent
Copy this and send it to your agent — Claude Code, Codex, any of them:
Ask your agent to install Moli from https://github.com/lexmount/moli following the README, fetch the text of the example.com test page, and show me the result.
An infrastructure tool, not a GPU load: per the README the engine renders the picture on the CPU itself, no GPU needed. A regular laptop is enough for occasional requests, and for an agent that keeps many parallel browser sessions it is worth putting this on a server, that is where the memory savings against Chrome are felt the most.
Open on GitHub▌ More finds