Hey, I am trying to connect to the browser on my o...
# general
t
Hey, I am trying to connect to the browser on my openhands, and I tried to look for a tools section, but it is not working. Please, can anyone offer some assistance? I connected it to my GitHub repository, but was not able to view the changes in the browser when I ran a prompt; the browser failed when i put it in not sure why. Is there a connector I am missing?
o
A few things I’d check first: 1. Is the app actually running inside the OpenHands workspace? For example, the dev server should be started with something like
npm run dev
or the project’s equivalent command. 2. Is the server binding to
0.0.0.0
, not only
localhost
? Some sandbox/browser setups can’t reach a server bound only to localhost. 3. Check the terminal logs for the port, then open that exact forwarded/browser URL. 4. If the prompt changed files but didn’t start the app, ask OpenHands explicitly: “start the dev server, tell me the port, and open the browser preview.” If you paste the framework or the error message, people can probably narrow it down.
t
thank you so much, I will check these and get back in
👍 1
Would this also be applied to having the terminal up and running? Because I ran some prompts but do not see the commands the agent ran
I am also curious what model do you run on your setup?
o
Yes, similar idea. The terminal has to be actively running in the workspace. Also, depending on the setup, the agent may not leave every command visible in the terminal panel you’re looking at, so it’s worth checking the task logs or execution history too. If you want to force it, you can prompt it with something like: “Show me the exact commands you ran, then run the dev server in the visible terminal and leave it open.” For my setup, I use Codex through the desktop app.
t
Thank you so much for the help. I am a newbie, so pardon my haphazardness. I was too excited to try it out.
o
No worries at all. That is completely normal when you’re first testing these tools. The main thing I’d suggest is to keep each task very small: 1. ask it to start the app 2. ask it to show the exact command it ran 3. confirm the localhost URL 4. then ask it to make one change and refresh That makes it much easier to tell whether the issue is the app, the terminal, the browser, or the agent. Good luck with it. The first few runs are always a little messy.
❤️ 1