Just an update if you want to use google antigravity oauth for openclaw. Dont do it your account gets banned. :)

How can you add OAuth for openAI?

14 hours ago, Naidu said:

How can you add OAuth for openAI?

I didn't have this issue, but saw that there was a solved issue on github. https://github.com/jdhill777/openclaw-unraid/issues/8

Anyone know how to get the agents to create cron jobs? When I try it says that it can't because it doesn't have access to the openclaw command.

I haven't seen anywhere on how to add in the openclaw cli to an existing container.

Thanks for adding,

of course i just installed it and need to try.

anyoneget the same issue ?

my answers are empty ( using gemini with api key)

Ive got the same issue, with a fresh install and but using Claude api key. I did try changing to host mode to bridge and custom IP, which made no difference. Does the container need to run in Admin/Root mode? Am i missing something in post install process from the Unraid store, Is there a container setup/install script to run?

12 hours ago, Curtis777 said:

Thanks for adding,

of course i just installed it and need to try.

anyoneget the same issue ?

my answers are empty ( using gemini with api key)

What local AI models (ollama), are you guys using?
I just mine configured and am starting off testing: olmo-3:7b-instruct

I also connected Telegram and the agent was confused by the metadata from the telegram prompt, maybe since its only a 7b model, i added some info in the soul.md file to not respond about the metadata, just the message.
(Not openclaw) Some bench marking analysis i did on my local models for a 16gb gpu using claude: (note claude misprinted some model names in the artifact from the spreadsheet, eg qwen2.5 is qwen3, gemma2 is gemma3 etc. https://claude.ai/public/artifacts/21b88a74-5087-402f-8a93-a833218a6529

I'm testing olmo3 since it had the highest tokens per second and full accuracy of my test questions, as well as cabability: Tools.

But i'm not sure if i want to use a model with vision, or thinking also?

Something that can fully fit in vram, so for a 16gb gpu, anything less than 12-14gb to allow context, so roughly 12-20B parameters roughly, or 7-8B to allow more context.

I mainly want to connect this to Home Assistant, and not yet looking to have it to anything too major until i can be sure i have the security locked down.

same issue for me too. Also using claude api key : chatbot answering empty responses. Perhaps something to setup in the "model" tab of the setup panel in the webui of openclaw but i am too limited to successfully do something working in it ! my bad

11 hours ago, Nosbus said:

Ive got the same issue, with a fresh install and but using Claude api key. I did try changing to host mode to bridge and custom IP, which made no difference. Does the container need to run in Admin/Root mode? Am i missing something in post install process from the Unraid store, Is there a container setup/install script to run?

same issue for me too. Also using claude api key. (bot answering empty responses). Perhaps something to setup in the "model" tab of the setup panel in the webui of openclaw but i am too limited to successfully do something working in it ! my bad

Installed from CA, executed the Homebrew command:

Having a similar problem not being able to install skills from the UI. Everything is marked as blocked and "Missing bin:*"

Install from CLI during onboarding fails for some skills. Those that are installed well still come up as blocked in UI

Any ideas?

Anyone getting 2026-02-13T13:01:09.873-05:00 Gateway failed to start: Error: tailscale serve/funnel requires gateway bind=loopback (127.0.0.1)?

I've had a very strange issue with this docker twice now over the past few days.

Having installed and trying to set up, I've not been able to successfully get replies from the chat/browser, and in trying to get the most basic of replies through the Anthropic API, I've had a major issue that required a restart, both times.

The docker/program will hang, followed by all of my unraid 'shares' disappearing, and things slowly starting to break in other dockers as they attempt to access the /mnt/user/ share. Strangely, I can still browse my files via the 'Main' tab, but /mnt/user/ in particular (possibly others) was not visible/working. In both instances, uninstalling openclaw, and performing a full reboot, fixed the shares.

I otherwise run dozens of other dockers 24/7, many with internet access (via reverse proxy), and the server is showing no other errors, not disk/ram errors. And given the issue was repeatable, it does lead me to think this openclaw docker is somehow to blame.

A) What could cause the openclaw docker to do this?
and
B) is it me doing something wrong, or something not quite right with the docker, or openclaw itself?

If it wasn't repeatable, I'd be thinking some kind of RAM issue, but as above, uninstalling openclaw & rebooting after has fixed it twice now.

Edit:

On 2/12/2026 at 9:19 AM, Curtis777 said:

Thanks for adding,

of course i just installed it and need to try.

anyoneget the same issue ?

my answers are empty ( using gemini with api key)

22 hours ago, Nosbus said:

Ive got the same issue, with a fresh install and but using Claude api key. I did try changing to host mode to bridge and custom IP, which made no difference. Does the container need to run in Admin/Root mode? Am i missing something in post install process from the Unraid store, Is there a container setup/install script to run?

Same issue, I couldn't get a reply, using anthropic/claude either

Edited February 13Feb 13 by Jazmodo

On 2/3/2026 at 11:32 PM, xleejohnx said:

I got everything up and running but i having an issue on getting telegram to work.

I need to run openclaw pairing approve telegram with the code and i cannot figure out how to make it work

anyone get telegram working?

I had this same issue (and wish the maintainer would update the readme.md to indicate steps required to actually make this work.)

I was able to get the bot to respond after editing the JSON and manually enabling the telegram under the plugins section:

"plugins": {

"entries": {

"telegram": {

"enabled": true

}

}

}

After that I still couldn't figure out how to approve the pairing code through the UI. To my surprise I just asked in the openclaw chat dialog how to do it, and it went ahead and approved my pairing request and also fixed some other settings that it said were misconfigured. I then asked it to make additional setup recommendations and it walked me through a few questions and started recommending skills. So far this is pretty cool!

Since its inside the container how does it work ? And how does it gets the job done ?

15 hours ago, Jazmodo said:

I've had a very strange issue with this docker twice now over the past few days.

Having installed and trying to set up, I've not been able to successfully get replies from the chat/browser, and in trying to get the most basic of replies through the Anthropic API, I've had a major issue that required a restart, both times.

The docker/program will hang, followed by all of my unraid 'shares' disappearing, and things slowly starting to break in other dockers as they attempt to access the /mnt/user/ share. Strangely, I can still browse my files via the 'Main' tab, but /mnt/user/ in particular (possibly others) was not visible/working. In both instances, uninstalling openclaw, and performing a full reboot, fixed the shares.

I otherwise run dozens of other dockers 24/7, many with internet access (via reverse proxy), and the server is showing no other errors, not disk/ram errors. And given the issue was repeatable, it does lead me to think this openclaw docker is somehow to blame.

A) What could cause the openclaw docker to do this?
and
B) is it me doing something wrong, or something not quite right with the docker, or openclaw itself?

If it wasn't repeatable, I'd be thinking some kind of RAM issue, but as above, uninstalling openclaw & rebooting after has fixed it twice now.

Edit:

Same issue, I couldn't get a reply, using anthropic/claude either

you are probably getting rate limited

16 hours ago, Fffrank said:

I had this same issue (and wish the maintainer would update the readme.md to indicate steps required to actually make this work.)

I was able to get the bot to respond after editing the JSON and manually enabling the telegram under the plugins section:

"plugins": {

"entries": {

"telegram": {

"enabled": true

}

}

}

After that I still couldn't figure out how to approve the pairing code through the UI. To my surprise I just asked in the openclaw chat dialog how to do it, and it went ahead and approved my pairing request and also fixed some other settings that it said were misconfigured. I then asked it to make additional setup recommendations and it walked me through a few questions and started recommending skills. So far this is pretty cool!

Thanks, this helped me out too.

Is anyone else having an issue with their OpenClaw where it will be functional but the agent tab in UI doesnt actually display the model being used? My dropdown is also empty. Is this a bug? Do I have something misconfigured maybe? Also I'm trying to figure out if there is a way to run Ollama with it to have a fully local setup. I know i'm using Claude Opus at the moment but the only way I can actually see that is through terminal.

Anyone run across this error yet?

Error: missing scope: operator.read

8 hours ago, jgmeeks said:

Anyone run across this error yet?

Error: missing scope: operator.read

I just started messing with this yesterday to get familiar with it and had it working. I started a new session, had some issues and now I am stuck with that same error and can't get past it.

Yup, same operator read error over here. Tried to work with another AI to fix the errors shown in the logs, didn't have any success. Seems this issue came with the latest update as my docker container updated last night.

yes right now the issue is known. Here is a link to where it can be fixed.

https://github.com/openclaw/openclaw/pull/17296

Mostly just downgrade back to the previous version. I have no idea how so i hope someone to help with that!

37 minutes ago, razorline said:

yes right now the issue is known. Here is a link to where it can be fixed.

https://github.com/openclaw/openclaw/pull/17296

Mostly just downgrade back to the previous version. I have no idea how so i hope someone to help with that!

Edit the container. Under the repository field change openclaw:latest to openclaw:2026.2.13 in order to use the image released 2 days ago.

Edited February 15Feb 15 by robodan

On 2/5/2026 at 6:49 PM, Specthildur said:

has anyone managed to get browser functions to work? I would like to use my local searxng instance, but I don't know how to get it's own function to work, I would like to avoid braves api one.

Did you find a solution for this? I also try to use searnxg because Brave API is not free anymore or at least not without credit card information.

Newbie question: If Openclaw runs in a docker container, doesn't that mean any tools it installs in the course of fulfilling tasks will be loss upon restart of the container? Would it be better to run Openclaw in a Virtual Machine so that the tools it installs persist?

5 hours ago, gentlemoon said:

Newbie question: If Openclaw runs in a docker container, doesn't that mean any tools it installs in the course of fulfilling tasks will be loss upon restart of the container? Would it be better to run Openclaw in a Virtual Machine so that the tools it installs persist?

Openclaw stores the persistent data into the appdata folder. That way, the data is not only persistent during restart but also after updates.

On 2/6/2026 at 1:22 AM, razorline said:

I am trying to set up gog for google and the auth url it wants me to use doesn't work since its not from the same instance, is there a work around for this?

Did you get this working? I tried this trough open claw directly, but it adds parameters like credentialsPath to the config, which are no longer supported. I tried
gog auth add [email protected] --services gmail,calendar,drive,contacts,sheets,docs
but that generates a link that points back to local host and a port that is not accessible when running in a docker.

I was able to get around the redirect by adding --manual at the end and it seems that it saves the credentials in the keyring now. But openclaw still refuses to send emails. There must be some additional configuration in openclaw to get this working.

Edited February 16Feb 16 by TylerDurden182730