felarof

This is cool!

We are working on agentic browser (also launched today https://news.ycombinator.com/item?id=44523409 :))

Right now we have a desktop version with ollama support, but we want to build a mobile chromium fork with local LLM support. Will check out cactus!

We would've preferred to build this as browser extension too.

But we strongly believe that for building a good agent co-pilot we need bunch of changes at Chromium C++ code level. For example, chromium has a accessibility tree for every website, but doesn't expose it as an API to chrome extension. Having access to accessibility tree would greatly improve agent execution.

We are also building bunch of changes in C++ for agents to interact with websites -- functions like click, elements with indexes. You can inject JS for doing this but it is 20-40X slower.

this is on our radar, we plan to have it ready by early next week!

still a team of 2 people, so bunch things on our plate.

We are a chromium "wrapper"

But we are much more performant than other libs (like playwright) which are written in JS, as we implement bunch of changes at chromium source code level -- for example, we are currently implementing a way to build enriched DOMtree required for agent interactions (click, input text, find element) directly at C++ level.

We also plan to expose those APIs to devs.

Thank you!

plan is to sell licenses for Enterprise-version of browser, same as other open-source projects.

We just launched an open-source alternative -- you can download today; use local LLMs with ollama and not pay $200/month.

https://news.ycombinator.com/item?id=44523409

we have an open-source alternative that is ready to download -- https://browseros.com/

for local LLM, you can download ollama and you can download 4B or 8B variants.

Reg. first part, our thinking right now is similar to other OSS projects, we will have enterprise offering and charge for that.

Thank you for your support!

Yes, in the long run, owning the underlying codebase would allow us to have more control and provide better user experience. (Very similar to cursor forking vscode)

Yeah good point. if there was a "popup" or some "notification" at OS level which asked for approval for certain actions, would that be helpful? Also, maybe we can allow users the flexibility with regards to how much automation they are comfortable with.

Haha thanks. Based on today's feedback, I guess we need to spend little more time on our naming and branding. will think about it :)

Thank you so much for this feedback! And sorry for the annoying experience. I'll debug and fix it. this still early beta version :(

We have this agent mode and chat mode both with separate tools. I think the "prompt" today is not good enough, will see if there is a better way to address it.

Regarding un-grouping, that API is currently missing in chrome. I'm currently looking to add this support.

Hmm couple of people have asked for "undo" now. will see how we can implement this. I imagine something like cursor's "restore checkpoint" would be neat.

Quick question, do you think these productivity features are critical in your day to day workflow? Any specific examples you can share? :)

spot on! These are great examples.

Hmm, there are similar use-cases which I would much prefer to offload to AI rather than me spending time on it -- "compare prices of 2tb sandisk ssd hard drive across amazon, walmart and b&h"

Good point. Our thinking so far has been to build good open-source product and then offer enterprise version as paid.

Island browser, chrome enterprise have kinda of validated the need for enterprise version of browser with VPN and DLP engine (data-loss-prevention).

I see, definitely agree that more work is needed in figuring out the right UX here. Probably open a shadow browser with OAuth sessions of only a small subset of sites required for the task?

Thanks for raising this - it's a critical concern and you're absolutely right to be cautious.

This is exactly why we're going local-first and open source. With cloud agents (like Manus.im), you're trusting a black box with your credentials. With local agents, you maintain control:

- Agents only run when you explicitly trigger them

- You see exactly what they're doing in real-time and can stop them

- You can run tasks in separate chrome user profile

- Most importantly: the code is open source, so you can audit exactly what's happening.

Thank you so much for your honest feedback. I 100% agree - this is spot on! This is exactly the vision we had.

We spend 90%+ of our time in browsers, yet they're still basically dumb windows. Having an AI assistant that remembers what you visited, clips important articles (remember Evernote web clipper?), saves highlights and makes everything semantically searchable - all running locally - would be game-changing.

Everything stays in a local PostgresDB - your history, highlights, sessions. You can ask "what was that pricing comparison from last month?" or "find my highlights about browser automation" and it just works. Plus built-in self-control features to block distracting sites when you need to focus.

Beyond search and memory, the browser can actually help you work. AI that intelligently groups your tabs ("these 15 are all Chromium research"), automation for grunt work ("compare 2TB hard drive prices across these sites"), or even "summarize all new posts in my Discord servers" - all handled locally. The browser should help us manage internet chaos, not add to it.

Would love to hear what specific workflows are painful for you!

great idea & usecase! Will look into this.Thanks!

Was huge fan of Tampermonkey back in the days.

+1. Browser definitely seems like it's on the verge of getting reinvented, either by us or someone else.

Thanks for the honest feedback!

Definitely agree there is good amount of competition here.

But we do think there is a gap in the market for open-source, community driven and privacy-first AI browser. (Something like Brave?)

Oh that's from browser use buildDOMTree.js. check it out on their github.

Not on purpose, just got compressed during production build through webpack. will get that fixed.

+1, I use AI tab grouping feature in our browser quite a lot.

Some ideas here - https://nxtscape.feedbear.com/roadmap

feel free to add new or upvote. Want to build what people want :)

Yeah, I feel LLMs can finally solve the tab overload issue. I suffer from this constantly.

I added few features which I felt would be useful - easy way to organise and group tabs - simple way to save and resume sessions with selective context.

What are your problems that you would like to see solved?

Yes, eventually we think there is more value of owning the entire stack than just be a MCP connector.

Few ideas we were thinking of: integrating a small LLM, building MCP store into browser, building a more AI friendly DOM, etc.

Even today, we use chrome's accessibility tree (a better representation of DOM for LLMs) which is not exposed via chrome extension APIs.

Yes, it works.

Hopefully in a month or two. Sorry!

Qwen3 8B works pretty well. But for complex planning and navigation tasks, big models (GPT4.1, claude 3.7) are the still the best bet. We also let you use your own API keys for the big models.