Show HN: Nxtscape – an open-source agentic browser

I’m using Dia a lot for work at the moment and frankly it’s a gamechanger. granted I’m not a developer but being able to interact with an LLM that has access to the page I’m on is extremely useful:

Instead of manually hunting across half a dozen different elements, then copy/paste and retype to put something into a format I want…

I can just get Dia do it. In fact, I can create a shortcut to get it to do it the same way every single time. It’s the first time I’ve used something that actually feels like an extension of the web, instead of a new way to simply act on it at the surface level.

I think the obvious extension of that is agentic browsers. I can’t wait for this to get built to a standard where I can use it every day… But how well is it going to run on my 16GB M1 Pro?

16GB M1 Pro is good enough to run our browser! You should give it a try!

Download form https://www.nxtscape.ai/ or our github page.

Why a fox? The browser is based on Chrome. Firefox basically owns having a fox as a mascot in the browser space. Why not pick something original? The fox confusing at best, but some may say misleading. Same goes for the name.

> Is there any usecase in the future that a browser can, but an extension can't?

The only reason to use a browser over a chrome extension is to bypass security features, for example, trusted events. If a user wants the browser window to go to full screen or play a video, a physical mouse click or key press is required. Moreover, some websites do not want to be automated like ChatGPT web console and Chase.com which checks if the event was a trusted event before accepting a button click or key press. This means that a Chrome extension can not automate voice commands inferred with audio to text. However, to get a trusted event only requires the user to press a button, any button, so message or dialog prompt that says, "Press to go full screen," is all that is required. This can be down with a remote bluetooth keyboard also.

The way I see it, these limitations are in place for very, very good reasons and should not be bypassed. Moreover, there are much larger security issues using a agentic browser which is sending entire contents of a bank website or health records in a hospital patient portal to a third party server. It is possible to run OpenAI's whisper on webgpu on a Macbook Pro M3 but most text generation models over 300M will cause it to heat up enough to cook a steak. There are even bigger issues with potential prompt injection attacks from third party websites that know agentic browsers are visiting their sites.

The first step in mitigating these security vulnerabilities is preventing the automation from doing anything a Chrome extension can't already do. The second is blacklisting or opt in only allowing the agents to read and especially to write (fill in form is a write) any webpage without explicit permission. I've started to use VSCode's copilot for command line action and it works with permissions the same way such as only session only access.

I've already solved a lot of the problems associated with using a Chrome extension for agentic browser automation. I really would like to be having this conversation with people.

EDIT: I forgot the most important part. There are 3,500,000,000 Chrome users on Earth. Getting them to install a Chrome extension is much, much easier than getting them to install a new browser.

FreeBSD, Haiku, Amiga