🎅 Feature request (I wish BrowserOS had...)

[shadowfax92]

Feature request: What would you love to see in BrowserOS?

This is a place to share feature ideas and requests for BrowserOS.

Drop your suggestions below! 👇

Tell us what features you'd love to see - we'll follow up if we have questions and consider them for our roadmap!

React with ❤️ to requests you'd also want!

P.S.: Join our Discord to chat with the community 👋

[callumreid]

ai model selection

[felarof99]

ai model selection

@callumreid We already have great support for bringing your own models (local models with Ollama, LMStudio or API keys to big LLM providers) and allowing you to select models!

[tekacs]

My #1 feature request (I... don't think that you have this right now?) is exposing an MCP server. Not using MCPs from inside the browser, but making it so that the browser can be - read from, controlled, etc. via MCP.

There are various Chrome extensions to do this, but they all have poor ergonomics, disconnect randomly and glitch / flake out.

If you were to add this, would gladly recommend BrowserOS to users of our product as well, since it gives us (and others) a point to hook in – but also... I would love to just... ditch many of the browser experiences and just use... BrowserOS as my browser, and ChatGPT/Claude Desktop as my AI to interact with it.

[bethebunny]

I use the Github Live Folder feature in Arc heavily. It's amazing even if it doesn't actually map well to my use case, and afaict is only implemented for Github pull requests.

• I'd love some way to configure this for services that don't need a direct integration, similar to chrome's custom search engines
  • For instance, a folder corresponding to an RSS feed that opens a new tab for items as they come in
  • Or, a URL that points to a JSON rest endpoint that lists the URLs that should be open in the folder for instance
• I wish I could configure it to not close tabs until I manually close them or the PR is closed
  • for "created by me" tabs, once they are approved I need to go find them again to merge the PR
  • some PRs I want to read/review even if approved by someone else
  • PRs that I've already approved may still have new conversation on them before merging

[atrejo3]

@callumreid We already have great support for bringing your own models (local models with Ollama, LMStudio or API keys to big LLM providers) and allowing you to select models!

The feature is there and I thank you for it, but its not obvious. You have to go into a particular tab in Chrome settings then "Add custom provider".

Before typing this comment, I couldn't find where to enter key and endpoint if I want to use an external provider. But its actually in the same place. I just never think of clicking Add custom provider as the first step to entering a URL endpoint.

Initial run of the release didn't even have an agent tab! Only "LLM chat" and "LLM hub", which are more of a distraction to the main feature this browser has.

I think I'd like to have only the agent tab with a way to just to settings from there, and an immediate way to enter a URL endpoint and optional key in these settings without having to select or add a provider. Maybe always have the form visible and the quick templates just prefills it.

****

Tell us what features you'd love to see - we'll follow up if we have questions and consider them for our roadmap!

• Use something other than chromium
• Have a forum or mailing list instead of Discord
• Use HTML, generated if needed, instead of a notion site
• Remove telemetry, including teh bitly link

None of this is necessary but indicative of how likely the project is to stay open source or suddenly change directions. That is perhaps the most important feature. Its too easy to spin up a project and community quickly and then discard it just as fast a few months later.

For actual features: easier way to debug local LLMs, maybe some command line options and some way to talk to the agent and get its response back through a server instead of typing into a text field.

[bradflaugher]

If you were to add this, would gladly recommend BrowserOS to users of our product as well, since it gives us (and others) a point to hook in – but also... I would love to just... ditch many of the browser experiences and just use... BrowserOS as my browser, and ChatGPT/Claude Desktop as my AI to interact with it.

I already love it, and would support you guys either setting up BrowserOS as an MCP server or connecting to other tools via MCP, I'd be thrilled with either feature and it would take its usefulness to the next level, right now I can automate stuff that I manually put in to the browser agent, but for some workflows I want to either push tasks to the agent (BrowserOS as an MCP server) OR pull tasks in from other places (BrowserOS Agent as host to other MCP Servers). If you implement either feature I'd jump for joy.

[isuru0766764313]

Pls add DeepSeek, GLM, Qwen, Kimi and Long Cat AI agents also.. Otherwise this is not surprise with traditional AI agents

[damngamero]

pls add a switch to switcj between ms edge interface and chrome interface

[Fixmio]

Android and iOS version would be great additions.