Moli by Lexmount

The fastest headless browser for agents and harnesses

Productivity

Developer Tools

Artificial Intelligence

Visit WebsiteSee on Product HuntTwitter ⧉

What makes Moli an agent-native browser instead of just a lightweight browser?

Nice. My agents drive a real browser all day (Gemini for images, editors over a debug port) and the slow part is always waiting for pages and uploads, not the model. Does Moli keep logged-in sessions between runs, or is every run a clean profile?

Looks amazing, great work on the launch team! Love the ergonomics of the SDK.

Before we built @Strawberry we were building something similar called dendrite SDK, very cool to see how far the tech has gotten!

Building agent automation systems, resource efficiency is everything. The 1/7 memory footprint vs Chrome is a real win - running thousands of headless instances becomes viable. Love that it's open-source too, gives more control over the execution layer.

Hey Product Hunt 👋 I’m Nico, one of the makers of Moli by Lexmount.

I wanted to share how our new Cloud Browser Agent fits alongside Moli and the Lexmount cloud platform:

• Moli is the browser engine. It’s open source and built for AI agents, providing page structure, interaction, and rendering on demand.

• Lexmount is the cloud platform. We provide managed browsers and web access APIs. Bring your own agent or automation code, and we handle the browser infrastructure.

• Cloud Browser Agent carries out the task, using our CAE protocol to understand and interact with web pages. Give it a goal and an optional starting URL. For recurring tasks, save your instructions and output format as an Agent, then create a Run for each execution.

CAE — Content, Action, Effect — connects page content, available operations, and browser-observed changes, using incremental updates to reuse stable page context. For a closer look at how CAE works and the reasoning behind its design, here’s our technical write-up: CAE Protocol.

For developers, this means decisions grounded in browser feedback, less redundant input and token overhead, and an execution trace you can inspect.

What browser task would you want to hand off first?

How does it handle websites with heavy JS or dynamic content compared with the regular chrome session ?

Is it easy to migrate an existing browser automation workflow to Moli by Lexmount?

Hi everyone! I’m Wangli, one of the makers of Moli by Lexmount and the first engineer hired for the Lexmount cloud platform.

I’d love to share how we think about the platform’s evolution.

As AI moves from generating answers to taking action, growing token usage is accompanied by another need: more agents reading websites, clicking through workflows, and completing tasks in parallel. Those agents need browsers they can rely on.
That’s where we started: reliable cloud browser infrastructure that scales with demand, supports many concurrent sessions, and lets developers pay for what they use.

As browser capacity becomes easier to scale, the next challenge moves up to the agents driving those browsers. Developers also need somewhere to run their agents and automation code. That led us to cloud browser agents and a hosted JavaScript runtime, extending the platform from browser sessions to the work happening around them.
Another direction I’m particularly excited about is connecting cloud execution with local resources.

On the broader Lexmount platform, our SDK already supports a local proxy tunnel for standard cloud browser sessions. With authorization, a cloud browser can reach an internal company portal or a VPN-accessible website through a machine on that network. The browser still runs in the cloud; the local machine provides the network connection. We explain how it works in this post.

That small capability opened up a much bigger question for us: what else could a local client make available to an agent, with the user’s explicit permission? Selected files? A camera? Other local tools?
We’re exploring how agents and browsers could work across cloud and local environments, combining cloud capacity with access to local resources while keeping users in control of what is exposed and where work runs.
That’s the future I’m excited to help build. What would you want your agents to do across your local and cloud environments?

Hi everyone! 👋 I’m Julia, COO at Lexmount.

My focus is helping developers take Moli from “this looks interesting” to something useful in their own workflows.

We’d love to hear what you’re building—whether that’s a research agent, a data collection pipeline, or browser automation—and what’s making the browser part difficult today.

If you try Moli and hit a roadblock, please share it here. Our team is around to help, and your feedback will help us decide what to improve next.

What’s one browser task you’d like to try with Moli?

Hi everyone! I’m Donough Liu, one of the makers of Moli by lexmount.

One thing we’ve been deliberate about while building Moli: making it lightweight and headless shouldn’t come at the expense of browser security.

We’ve put substantial work into browser security (including same-origin restrictions, iframe sandbox restrictions, CSP, CORS, subresource integrity, MIME-type enforcement, and Referrer Policy..). These protections matter just as much when an AI agent is browsing as when a person is. 👻

That work rarely shows up in a demo, but it’s a core part of making Moli production-ready. Real-world automation needs a browser that respects the web’s security boundaries, alongside speed and efficiency. We’re building Moli with all of those responsibilities in mind.