users-helping-users

[compact egret]

Agents │ 1 total · 1 bootstrapping · 0 active · 0 sessions

[steep rapids]

Honestly, nvidia dgx spark is kinda cool, otherwise just a regular old desktop pc with a big ass GPU is your next best bet

But even more honestly, running local models for agentic workflows is quite simply not yet worth it

[lunar bolt]

hmmm... if i click on those items it just says "no output - tool completed successfully". On the discord side, it successfully reacted with the eyes looking, but sits indefinitely as "sporkbot is typing"

[steep rapids]

a project like this works best with one of the SoTA models through inference.

[steep rapids]

hahah what is this clanker doing 😄

what model is this using

so I know not to use that ever

[lunar bolt]

i'm trying to get it using ollama/qwen2.5:7b-instruct, but it seems like it's not producing properly formatted responses; when i run it directly through ollama interactive mode, it works fine

auto-discovery doesn't seem to work, so i configured a manual provider for it, and it was able to see it, but yeah, that's where i'm at...

[brisk pivot]

looking for help dont mind pay for it pm me

[little blade]

Try Qwen4B or 8B, it has similar performance to 32B models of 6/10 months ago

[full aurora]

How to add fallback models? Docs doesn't specify it only how to look at them: https://docs.openclaw.ai/cli/models

[brisk pivot]

Last login: Sat Jan 31 23:07:20 on ttys001
(node:6203) [DEP0040] DeprecationWarning: The ⁨punycode⁩ module is deprecated. Please use a userland alternative instead.
(Use ⁨node --trace-deprecation ...⁩ to show where the warning was created)
/dev/fd/12:3564: command not found: compdef

have anyone encounter this error?\

[floral warren]

anyone having browser relay not allowed on their browsers? for chrome extension

[steep rapids]

looks like some old packages, try npm update -g

but both those errors look like they shouldnt really matter tbh

[brisk pivot]

i have a problem where i tried reinstall openclaw

to setup it with opus instead of minimax with the claude max sub

but now openclaw return empty queries as answers

[steep rapids]

you probably still have some leftover model configs in openclaw.json and/or agents/<name>/agent/models.json

[brisk pivot]

idk how fix it u have anyidea?

[steep rapids]

i had the same thing earlier today and did manage to fix it yes

[static mantle]

can i use the /hooks/agent gateway webhook to activate a non-main agent? i see sessionKey as a parameter but i also want to use a different agentId than main

[robust cobalt]

I'm trying to update my model again, because my config was shot down by my agent.
Now when I run configure and change to claude haiku it doesnt work anymore.
After I go through configure the model is still the old (out of resources) gemini model.
Can someone tell me how I can set the model in the configuration file by hand?

[hushed badger]

can i get help on this?

im not good at tech so

[exotic hare]

Hi All, Fresh install today on ubuntu. I can ping the box through TailScale. OpenClaw says its running and (mostly) seems to run through the TUI. I've even connected to OpenAI via API. I've used the openclaw configuration tool to connect the the local loopback interface on standard ports.

Where I'm stuck is getting to the web interface from another machine that is connected (verified by Ping) to TAilScale. I've tried wiht token mode and password.

It has got to be somethign stupid. Please, throw me a bone here... 🙂

[hushed badger]

i even reinstalled Node.json

[pseudo hamlet]

[agent/embedded] Removed orphaned user message to prevent consecutive user turns.

any idea how can I fix this?

[worthy patrol]

Hi there. I'ts my first time setting this up. I've seen many people buying a mac mini. I don't have one now, believe using a hosting provider is easier. What's the best trade-off for speed / security? I am technical, but no super expert. I think something like AWS will create to much overhead for me, and I may end up making errors in config that might translate into security risks? Is there something more out of the box, or easy to set up a remote machine for installation? Thank you

[pseudo hamlet]

better start with your pc or mac instead of buying a server for this. Your PC or mac should be fine for now unless you want to keep it alive 24/7

[hybrid garden]

man my agent really struggles with the browser relay.

[pseudo hamlet]

for windows, better enable WSSL, and then use it on ubuntu.

[hushed badger]

which means?

[wary sierra]

help pls

[pseudo hamlet]

enable WSSL, restart the system and use ubuntu (bash) to install.

[trail hill]

does anyone have a guide on how to setup openclaw to use a local ollama model?

[hushed badger]

whats the line to ennable wssl

?

[wary sierra]

how do i enable wssl and use ubuntu? this is powershell

[runic tendon]

can i use it without any llm paid api account?

[worthy patrol]

Yes, I was thinking a 24/7 Uptime, I only have my mac which is moving. Any suggestion?

[wary sierra]

dism.exe /online /enable-feature /featurename:Microsoft-Windows-Subsystem-Linux /all /norestart

[pseudo hamlet]

WSSL (Windows Sub System for Linux), Take help of any AI tool like gemini or gpt and follow the steps to enable it, Once enabled, you'll basically will have ubuntu installed on your PC and you can pretty much do everything others doing on Mac.

[fleet nexus]

did you get the help you needed

[still patio]

Guys ... how do I install the Mac desktop app (Menu bar companion)? I see the docs, but no instructions for how to install it. https://docs.openclaw.ai/platforms/macos

[hushed badger]

see, i did that and the code ENOENT still came up

[lunar bolt]

did they ever add support for GUI linux apps?

[olive slate]

Did you ever figure this out? I'm having the same problem.

[pseudo hamlet]

Even I'm using my m3 macbook pro to test and it's working great. I might move it to any VPS server once it fits in my work.

[fleet nexus]

Ensure Local mode is active and the Gateway is running. did you get this step ?

[still patio]

I'm running the gateway on my local Mac, yes.

[pseudo hamlet]

enter bash and then install the linux installation command

[fleet nexus]

I dont know of an export method do you ?

[hushed badger]

where can i find said comand?

[fleet nexus]

ok can you send me a screen shot or any other info

[pseudo hamlet]

I have asked the bot itself and it replied that most of it's memory is the docs it saves of our preferences, so it says to zip and unzip at the preferred path and it automatically gets the memory about you.

[still patio]

Happily, but ... screenshot of what? Does this help?

[fleet nexus]

ok i use a db as backup so we dont have to just checking happy exporting

[wary sierra]

what is the bash liune

[frozen yoke]

What is the situation regarding using claude opus as the brain for openclaw? Did Anthropic specifically blocked it - possible ban?

[hushed badger]

bash apparently dont work

[slim trail]

Hi 👋 quick question on any experience for VPS+docker+tailscale? I tried to set up in order to access control UI on my local windows, openclaw in container did showing connected on tailscale but website+token still can’t be accessed. Wondering if I missed anything, any insights will be helpful!

[fleet nexus]

this is what my and motly got : let me know if you need more ?

[exotic hare]

right there with you. I'm stumped. I got it working with 0.0.0.0 binding but failed on https check in the system

[foggy jungle]

when i run ⁨shell openclaw gateway restart⁩ or ⁨shell gateway daemon restart⁩ it says it restarts, or if i just restart the vm entirely, and then rerun openclaw, whenever i try and run ⁨shell openclaw plugins enable memory-lancedb⁩ it says to restart the gateway which isnt working it says theres no memory-core enabled when i run ⁨shell openclaw doctor⁩ no matter how many times i restart and when i run ⁨shell openclaw plugins list⁩ it only says the discord plugin is loaded, also when i changed configuration like what model it was using, and discord plugin related conifguraiton via the dashboard, only the dashboard registered, its still using the previous preferred model which is gpt 4o but i restricted it to claude 4.5 opus, only me changing discord stuff via the dashboard has saved when i restart the gateway, also when i run ⁨shell openclaw doctor⁩ it says theres an update which i already ran ⁨shell openclaw update⁩ multiple times and the version is upto date with the latest release, so not sure why its not registering that its already updated here are the logs showing all relevant commands i can think of

[lethal blaze]

Hi! I'm trying to set up LM Studio for hours. No errors, but no response in moltbot. Does anyone have any ideas what might be wrong??

{
"meta": {
"lastTouchedVersion": "2026.1.30"
},
"gateway": {
"mode": "local",
"auth": {
"token": "kanzlei"
}
},
"agents": {
"defaults": {
"workspace": "/home/sanduser/.openclaw/workspace",
"model": {
"primary": "lmstudio/openai/gpt-oss-120b"
}
}
},
"models": {
"mode": "merge",
"providers": {
"lmstudio": {
"baseUrl": "http://172.152.178.40:1234/v1",
"apiKey": "lm-studio",
"models": [
{
"id": "openai/gpt-oss-120b",
"name": "Kanzlei GPT 120B",
"contextWindow": 131072,
"maxTokens": 8192
}
]
}
}
},
"commands": {
"native": "auto"
}
}

[slim trail]

I was there as well, how do you set up 0.0.0.0 bind? The latest patch I can only use bind=lan, was trying to setup cloudflare quick tunnel ( openclaw successfully made it work, but I deleted the instance and trying to replicate, no luck)

So I moved on to tailscale as it’s recommended, on openclaw documentation they provided configuration for it, but still can’t access UI, now I’m just using TUI temporarily

[exotic hare]

got it - take it back ot localhost

sudo tailscale serve --https=443 http://127.0.0.1:18789

then access without the port for https 443

[ivory rapids]

did you fix it?

fixed?

[slim trail]

Do you run this command in container?

[exotic hare]

sorry not using docker

[flint ether]

How can I see full logs with the bootstrap prompt injection? I enebled raw-stram logs, but there is no system prompts there.

[robust cobalt]

how do I stop openclaw? I need to change out the openclaw.json

[flint ether]

openclaw gateway stop

[slim trail]

Yah maybe I should give up on docker as well, over complicated the whole thing, since tailscale for some reason cannot fully penetrate container, or I missed something

[still patio]

I know this software is super powerful and stuff but it's ... kinda janky. Seems like a very vibe coded project

[celest topaz]

instead of cronjobs can i just edit HEARTBEAT.md

[robust cobalt]

ty

[lean helm]

I had no luck what does you setup look like

also I guess I'm running the local model on a crappy machine so might just not be capable, though in the app it works

[knotty solar]

What's up with the website?

Isn't working for me

[cloud swallow]

my bot isnt running can someone help

[lean helm]

not exactly descriptive to know where to help

[lean helm]

its been spotty

[knotty solar]

?

[lunar bolt]

Progress, but not total.

    "models": {
      "providers": {
        "ollama": {
          "baseUrl": "http://127.0.0.1:11434/v1",
          "apiKey": "ollama-local",
          "api": "openai-completions",
          "models": [
            {
              "id": "qwen2.5:7b-instruct",
              "name": "qwen2.5:7b-instruct",
              "reasoning": false,
              "input": ["text"],
              "cost": {
                "input": 0,
                "output": 0,
                "cacheRead": 0,
                "cacheWrite": 0
              }
            }
          ]
        }
      }
    },```

[lean helm]

I guess that and then the ⁨agents:{}⁩ stuff too?

[lean helm]

timing out time to time

[lunar bolt]

yeah

   "agents": {
      "defaults": {
        "model": {
          "primary": "ollama/qwen2.5:7b-instruct"
        },
        "workspace": "/home/rklebes/.openclaw/workspace",
        "compaction": {
          "mode": "safeguard"
        },
        "maxConcurrent": 4,
        "subagents": {
          "maxConcurrent": 8
        },
        "models": {
          "google/gemini-3-pro-preview": {
            "alias": "gemini"
          },
          "ollama/qwen2.5:7b-instruct": {
            "alias": "qwen2.5:7b-instruct"
          }
          }
        }
    },```

[knotty solar]

Ah ok

[lunar bolt]

But when I ping it, this is the response. It puts the looking eyes on the message, says typing until it times out, but nothing ever sent

just to test, i did switch model to gemini and it gave the correct response for the hatching script

[prime nimbus]

Sounds like gateway issue would be the first place to check, give it a restart and try again, if that fails, check the logs, normally going in with openclaw will display any issues for api etc

[lunar bolt]

it responds correctly with text when i point at gemini, so it's something about the ollama config in openclaw

[prime nimbus]

ah deffo, what model ollama you using? some can be a pain there is a #old-models chat

[safe meadow]

Hi after the installation I didnot get any response from the agent is that normal?

[lean helm]

damn, so close, I'm seeing into the ollama's setup they made for working with openclaw, but since they set this up under the old name it can't find it

so need to wait for an update

[lunar bolt]

22:21:29 debug diagnostic {"subsystem":"diagnostic"} lane dequeue: lane=session:agent:main:main waitMs=5 queueSize=0
22:21:29 debug diagnostic {"subsystem":"diagnostic"} lane enqueue: lane=main queueSize=1
22:21:29 debug diagnostic {"subsystem":"diagnostic"} lane dequeue: lane=main waitMs=1 queueSize=0
22:21:29 debug agent/embedded {"subsystem":"agent/embedded"} embedded run start: runId=beb51a8e-500b-44f1-a7df-e5991a5f89a3 sessionId=ad8069b4-276b-45b0-ba94-c264ed34dd83 provider=ollama model=qwen2.5:7b-instruct thinking=off messageChannel=webchat
22:21:29 debug diagnostic {"subsystem":"diagnostic"} session state: sessionId=ad8069b4-276b-45b0-ba94-c264ed34dd83 sessionKey=unknown prev=idle new=processing reason="run_started" queueDepth=0
22:21:29 debug diagnostic {"subsystem":"diagnostic"} run registered: sessionId=ad8069b4-276b-45b0-ba94-c264ed34dd83 totalActive=1
22:21:29 debug agent/embedded {"subsystem":"agent/embedded"} embedded run prompt start: runId=beb51a8e-500b-44f1-a7df-e5991a5f89a3 sessionId=ad8069b4-276b-45b0-ba94-c264ed34dd83
22:21:29 debug agent/embedded {"subsystem":"agent/embedded"} embedded run agent start: runId=beb51a8e-500b-44f1-a7df-e5991a5f89a3
22:21:47 debug agent/embedded {"subsystem":"agent/embedded"} embedded run tool start: runId=beb51a8e-500b-44f1-a7df-e5991a5f89a3 tool=tts toolCallId=call_ouvec1yz
22:21:48 debug agent/embedded {"subsystem":"agent/embedded"} embedded run tool end: runId=beb51a8e-500b-44f1-a7df-e5991a5f89a3 tool=tts toolCallId=call_ouvec1yz
22:22:04 debug agent/embedded {"subsystem":"agent/embedded"} embedded run tool start: runId=beb51a8e-500b-44f1-a7df-e5991a5f89a3 tool=tts toolCallId=call_rrhz2hev
22:22:04 debug agent/embedded {"subsystem":"agent/embedded"} embedded run tool end: runId=beb51a8e-500b-44f1-a7df-e5991a5f89a3 tool=tts toolCallId=call_rrhz2hev
22:22:20 debug agent/embedded {"subsystem":"agent/embedded"} embedded run tool start: runId=beb51a8e-500b-44f1-a7df-e5991a5f89a3 tool=tts toolCallId=call_urlhxinm
22:22:21 debug agent/embedded {"subsystem":"agent/embedded"} embedded run tool end: runId=beb51a8e-500b-44f1-a7df-e5991a5f89a3 tool=tts toolCallId=call_urlhxinm
22:22:37 debug agent/embedded {"subsystem":"agent/embedded"} embedded run tool start: runId=beb51a8e-500b-44f1-a7df-e5991a5f89a3 tool=tts toolCallId=call_nu481dne
22:22:38 debug agent/embedded {"subsystem":"agent/embedded"} embedded run tool end: runId=beb51a8e-500b-44f1-a7df-e5991a5f89a3 tool=tts toolCallId=call_nu481dne
22:22:54 debug agent/embedded {"subsystem":"agent/embedded"} embedded run tool start: runId=beb51a8e-500b-44f1-a7df-e5991a5f89a3 tool=tts toolCallId=call_p1z29te8
22:22:55 debug agent/embedded {"subsystem":"agent/embedded"} embedded run tool end: runId=beb51a8e-500b-44f1-a7df-e5991a5f89a3 tool=tts toolCallId=call_p1z29te8
22:23:19 debug agent/embedded {"subsystem":"agent/embedded"} embedded run tool start: runId=beb51a8e-500b-44f1-a7df-e5991a5f89a3 tool=tts toolCallId=call_khmkcbgb
22:23:20 debug agent/embedded {"subsystem":"agent/embedded"} embedded run tool end: runId=beb51a8e-500b-44f1-a7df-e5991a5f89a3 tool=tts toolCallId=call_khmkcbgb
22:23:55 debug plugins {"subsystem":"plugins"} plugin CLI register skipped (memory-core): command already registered (memory)```

No idea why it's doing this...

[lunar bolt]

qwen2.5:7b-instruct, which as far as i can tell, while underpowered for any serious work, SHOULD be compatible

and when i open it in ollama interactive mode it does give proper responses

[prime nimbus]

how does it act via tui?

[lean helm]

wondering its the endpoint or how it tries talking via the /v1 route

[lunar bolt]

are you seeing similar on yours?

[narrow roost]

Hi Mo here. Im looking for openclaw android app. Openclaw referred me here to find the app

[tribal jolt]

https://docs.openclaw.ai/getting-started

is the doc site down ?

[lean helm]

still trying to set it up

gimme a moment I'll get back to you see where I get

[rigid panther]

guys can i let clawdbot manage accounts for me ? like replying to tweets and engaging with other users automatically ?

[tired rain]

How to use openclaw agent on Android? (I also will use GRoq)

[lunar bolt]

grok is suggesting what I see i nthe screenshot is a result of the model incorrectly calling a tool instead of responding with text

[floral egret]

i just updated to openclaw this morning. it seems to be sending me it's internal thinking in telegram messages like half the time. why is that happening? it didnt used to

[lunar bolt]

this is interesting; i restarted the gateway and ran "hello" in the agentmain through the webUI...

[languid hamlet]

Anyone else have issues with the memory? I feel like it promises me a lot of stuff it can do and build but getting 0 actual reaction from clawdbot

[marsh urchin]

How do I update clawdbot to openclaw ?

I tried the command “clawdbot update” but it didn’t install openclaw

[floral egret]

just run the command to install openclaw. it has a migration step when you run it

[lean helm]

this with the ollama model?

I got I think your issue, its returning but empty no reply messages

[marsh urchin]

If I am already installed

What is the command

[floral egret]

you ran ⁨curl -fsSL https://openclaw.bot/install.sh | bash⁩? that should have asked you some questions

[lunar bolt]

yeah, ollama/qwen2.5:7b-instruct

[viral blade]

For those that need a free server to setting up Openclaw: #hardware message

[lean helm]

maybe lacking some tool powers, see if it can update its files or use any tools, maybe you need a full model to "wake" it up

[fading dagger]

I am unable to change the model from Gemini Pro Preview to anyother model. Is there a bug? I use GCP keys. By default, it's Gemini. Anyway to change. I unselect and continue again, but not switching back.

[brisk pivot]

looking to pay for a dev for a correct setup of my request

[lean helm]

could you specify?

[brisk pivot]

as a start im looking to correctly setup my mini mac m4 with opus 4.5

[meager mortar]

does anyone have streaming messages in telegram? the docs are confusing whether it is possible, where you can see the message as its being typed, not one big block

[lean helm]

happy to sit down and walk through it

[brisk pivot]

i got also vps and pc windows i want to connect through tailscape to let the mac mini m4 based on opus to connect and control them (also vps should be with open claw but diff model)

[lean helm]

but tbf still working some of the things out too so its just a matter of helping if you need to run the initial setup

[brisk pivot]

for now yes but i have always need a dev by my side thats why i dont mind pay if u are a serious guy

[brisk pivot]

lmk if its possible for u and when ur free i also pmed u brother for private chat

[lone vortex]

Why am I still on 2026.1.29 after using 'npm i -g openclaw'?

[marsh urchin]

Why does my gateway have a broken img link on it that Says "OpenClaw"

[coral rampart]

can someone suggest Free LLM to test with , openrouter not work

[plain talon]

using /model command, openclaw do not switch the model

[bleak mauve]

Having an issue with exec… Here’s the diagnosis from my bot:

Running any exec call on the Mac mini now fails immediately with spawn EBADF. The gateway process itself is running fine, but every attempt to spawn a shell hits a “bad file descriptor” before the command even starts. The bug comes from leaking PTY file descriptors on macOS, so after a few commands there are no FDs left and spawn just errors out. The OpenClaw team is already working a fix (PR #4153 closes the PTY leak and PR #4932 adds a macOS fallback that captures output without a PTY), but until that lands the only workaround is to restart the gateway periodically so the leaked FDs get released.

[sage skiff]

hey man i prob wouldnt mind doing it for free

[versed kettle]

How can I configure openclaw to use my local llm llama3.1-8b on my windows pc?

[raw heart]

how do I connect to the gateway in macOS (ARM), I'm kinda dumbass

[gloomy vortex]

Am i using lots of bandwidth my 20$ cluade accounts weekly usage is almost done after 2 days?

Have to wait 5 days to use claude again...

[elfin thorn]

that's why this project is stupid. It's just money maker for Anthropic

[gloomy vortex]

Yeah I kinda learned having the AI do all the work for me costs alot of API credits when i can do it myself

But it still has uses

[lunar anvil]

Whats the best way to avoid this? I'm a noob and can def get caught in this?

[marsh urchin]

I hit chatgpt rate limits and I can't get buzzy to switch to the other models, how I do force him to switch to opus?

[silk fulcrum]

I went with the service z.ai which is $15 for the first month then $30 after but I’m only playing around with the bot to see what it can do…then probably delete everything once the limitations are clear.

Be VERY careful downloading this bot on your only machine and giving it access to your files while it has internet access. That’s the main risk here. File corruption etc can happen but a backup will fix that.

I using an old MacBook with an empty dummy account and a fresh MacOS install that doesn't have any of my personal information. OpenClaw is connected to z.ai API for access to the the “brain” then then the Mac is connected to the internet for downloading etc.

[silk fulcrum]

Z.ai limits by the request and not by token usage. You get 600 chat prompt requests a day or something like that but the API usage is limitless compared to OpenAI, Anthropic etc.

[agile marsh]

Can we restart claw?
I have set it up "I think" have it my Gemini API key and tried to chat with is but it always just answers with A after a few seconds

[silk fulcrum]

”openclaw gateway restart“ in terminal

[small moon]

Hi

[agile marsh]

Tried that👍 didn't helped.
Closed the terminal now and will try to restart claw tomorrow.

Not sure what's the issue and why it's not answering

[ancient zealot]

im tryna setup openclaw with open router qwen free release, but it keeps sticking to openrouter auto- any ideas on how to re route it to the model im tryna achieve

[lunar anvil]

If I build it on AWS EC2 am I safe using my personal pc?

[silk fulcrum]

Sure, but you will lack the personalization that the internet is raving about like handling your messages, reminders, files remotely.

The bot will only have access to what you give it if you build on AWS. So, you can create dummy accounts for Google, Apple etc then give the bot access to those. If anything goes wrong you can cancel AWS and abandon the accounts for the most part.

[lunar anvil]

wish i wasn't poor lol - would just buy another device

[lunar anvil]

Yeah i dont want to miss out on that.

[lunar anvil]

So what is the difference between building on a fresh wiped Device, like a macbook, vs AWS EC2?

Aren't they both empty and need to have logins added?

[barren plank]

How to switch between which agent is being used?

[lunar anvil]

I don't even know shit - i just watched Alex Finns video the other day and have been trying to get it going.

I DM'd you Lee - all good if youd rather talk here though

[flat light]

https://youtu.be/egofv8c7oEk?si=5Z7aSSvKwzSVBT8d I hope this helps

[lunar anvil]

FWIW i'm not familiar with AWS.

[sweet mauve]

Hi everyone, does anyone know how to get composio working? My clawdbot just straight up refuses to integrate with it no matter what I do

[flat light]

https://youtu.be/egofv8c7oEk?si=kCDC6Bdtz25I0vPt

[sage torrent]

Hi guys, Im new here and to this and im split between setting up clawd on a vps or a raspberry pi. What would you guys suggest, are there any downsides to using a pi

[jaunty dagger]

What does this "401 No cookie auth credentials found" mean?
When I checked the models list, everything is configured correctly

[frail bluff]

Moltbook claim tweet not working.
Tweet: https://x.com/Patsaholic/status/2017739549405819017
Profile: https://moltbook.com/u/PhatBearGhost2 says "Bot Not Found"
Help?

[crisp wing]

@Patsaholic via Twitter

Freedom Isn't Free (@Patsaholic)

I'm claiming my AI agent "PhatBearGhost2" on @moltbook 🦞 Verification:
︀︀ swim-UF64

**👁️ 8 **

FxTwitter · Jan 31, 2026, 11:19 PM

[frail bluff]

Same with first link
Tweet: https://x.com/Patsaholic/status/2017728941830676873?s=20
Profile: https://www.moltbook.com/u/PhatBearGhost

Bot is online, just no response tweets?

[crisp wing]

@Patsaholic via Twitter

Freedom Isn't Free (@Patsaholic)

I'm claiming my AI agent "PhatBearGhost" on @moltbook 🦞 Verification:
︀︀ bubble-WES9

**👁️ 9 **

FxTwitter · Jan 31, 2026, 10:37 PM

[shadow needle]

I’m frequently running into “Unknown error” messages, I have to keep asking it where did it get stuck to get more details. Any idea how to make it more self unlocked?

[shadow needle]

I’m using Codex if that helps

[rancid tundra]

can the bot show images in the web chat when using Nanobanana ?

[mossy vortex]

Can someone help me understand why Browser Relay isn't working? Openclaw and open a browser but can't actually take control over it and do things on there. The extension is attached. I've tried restarting the gateway and reinstalling the extension. Nothing seems to work.

[umbral charm]

I think maybe moltbook is getting slammed and can't keep up?

[sour slate]

@frail bluff, Openclaw isn't affiliated with Moltbook. Moltbook is a separate user-developed project, so we would prefer it not be discussed in this server.

[frigid valve]

I used all my Claude API credits in like 20 messages lol

[odd folio]

Did you solve it?

[frigid valve]

Is that normal?

[prime nimbus]

THE TIMER RESETS SHORTLY 😉

[wintry oar]

have been waiting 2 weeks now lol

[prime nimbus]

BOOM! 😄

[prisma rose]

Been using Kimi K2.5 with OpenClaw, has anyone who uses it noticed that Kimi will often forget to put a newline character before its <|tool_calls_section_begin|> tag, and so the toolcall will be sent as part of a message instead of actually acting as a tool call? Not sure how to fix it, reminding it to have the newline before the tag doesn't seem to stick.

[prime nimbus]

GPT5.2-codex here me come 😄 lol

[frigid valve]

How much do the AI apis cost overall? I wasnt expecting it would use up like dollars per minute

[prime nimbus]

If you're 100 percent, best to open an issue on the github repo to help reproduce and the overall issue

[frigid valve]

Did I not config it right or is it high usage making the tokens more expensive?

[prime nimbus]

Depends how you set it up, most people use API's for setting it all up, then switch to local models to run it day to day

[frigid valve]

Oh so I can have it running locally without using Claude API??

Just like electricity and my machine!

[prime nimbus]

if you pc can handle it, yeah, most use ollama

[frigid valve]

I have ryzen 7

Not sure if it can

Is there a setup guide for ollama?

[prisma rose]

Less than 100 percent sure, but I feel that this is an issue with Kimi forgetting the newline as part of the tool call syntax, so not really something that the OpenClaw folks should feel like they have to fix? Might have to look into the codebase myself and see if I can sneak in something that detects the presence of the tag and always inserts a newline before parsing the agent's messages... xD

[wintry oar]

how do you even have it working? mine returns an error every time, i had to open a git issue

[astral plume]

What does this mean?

LLM error: {
"error": {
"code": 429,
"message": "Resource has been exhausted (e.g. check quota).",
"status": "RESOURCE_EXHAUSTED"
}
}

I have Gemini ultra and only had it do 2 things. Couldn’t have hit quota?

I was so excited to get any response first attempt. But now dunno where to go 😬

[prisma rose]

Had to manually add the synthetic kimi k2.5 model into the models list and as the default model in the openclaw.json file and the agents/main/agent/models.json file

[shell latch]

I'm getting errors when I try to install openclaw, can someone lend a hand?

[wintry oar]

what ya got brother

[shell latch]

[OK] Node.js installed via winget
[OK] Node.js v24.13.0 found
[*] Installing OpenClaw (openclaw@latest)...
node.exe : npm error code ENOENT
At line:1 char:1

• & "C:\Program Files\nodejs/node.exe" "C:\Program Files\nodejs/node_mo ...

•   + CategoryInfo          : NotSpecified: (npm error code ENOENT:String) [], RemoteException
    + FullyQualifiedErrorId : NativeCommandError

[wintry oar]

windows... no clue

from what i understand you might want to install a linux VM and use it that way

[frigid valve]

Would you please point me in the right direction for finding ollama and setting it up?

[wintry oar]

windows is super buggy from what ive heard

[steep ermine]

I have run to something I can’t figure out how to solve. Every time I hit my Claude limit. When the cooldown is over openclaw still gives me the cool down message. even though cooldown is def over

⚠️ Agent failed before reply: All models failed (8):.....ect

[prime nimbus]

https://ollama.com

[wintry oar]

https://ollama.com/ easy install instructions here

[frigid valve]

Thank you

[fierce charm]

Hello Shinjiro. It means there was an unauthenticated request, authentication is missing or invalid.

Here are the troubleshooting steps to try:

1. Gateway troubleshooting: https://docs.openclaw.ai/gateway/troubleshooting
2. Gateway authentication: https://docs.openclaw.ai/gateway/authentication

[worn mantle]

Newb here, what do you think? I have Telegram, Discord, and Slack installed, but new to all 3. Should I pick Discord?

[prisma bolt]

so thewy just came out with update, do i need to reinstall openclaw?

[fallen zealot]

Hello trying to setup openclaw with my claude pro plan is that still possible?

Pro Max*

[random cedar]

Is anyone else not able to access the docs? Seems like https://docs.openclaw.ai/ has been down all day.

[flint charm]

i'll post it here too, but im trying to auth with my claude max subscription and the agent keeps spitting out this: ⚠️ Agent failed before reply: All models failed (1): anthropic/claude-opus-4-1: Provider anthropic is in cooldown (all profiles unavailable) (rate_limit).
Logs: openclaw logs --follow

no way im actually rate limited

pls help miss my bot :(

[fierce charm]

https://docs.openclaw.ai/ working fine for me last 6 hours

[austere narwhal]

yeah

[wind chasm]

has anyone set up with ollama? im trying to set up qwen2.5vl
but i keep getting Embedded agent failed before reply: Unknown model: ollama/qwen2.5vl

[deep lance]

Any idea why sonnet is burnjng through tokens like crazy? I asked it to say “hello” to me and it used 3$ worth of openrouter credits i am so confused, is it bc of the context on my clawdbot? Anyways to fix this

[sand grove]

where is the magic sauce code and config that connects the gateway to the model agent - so when a msg comes from a channel it is routed to the agent to load the local ollama model and send the request to it?
That is the crux of the issue i am facing running on a local model and i suspect something in the way clawdbot calls the local model is broken.....

[fierce charm]

Hello Val. It could be the provider is experiencing high load, so it is anthropic in cooldown (rate_limit). Cooldowns typically clear within a few minutes to an hour.

[flint charm]

this now ⚠️ Agent failed before reply: HTTP 401 authentication_error: Invalid bearer token (request_id: req_011CXgQqvmJdhpXWEN3GcTCs).
Logs: openclaw logs --follow

[unique hazel]

https://discord.com/channels/1456350064065904867/1467314857023115426

Is my situation similar to yours?

[buoyant geode]

I just bought vps from hostinger, installed openclaw but it's working super clanky, I'm using telegram, it responds sometimes or only after a while
Is it the machine or I've done something wrong?

[flat light]

https://youtu.be/3b8zP-WrncM?si=v9zlHyLgE6K_y6Ak

[random cedar]

Thanks for verifying! Turns out is being blocked at the network level as an unsafe site for some reason. I was able to get around it via VPN if anyone else has this issue.

[jaunty dagger]

Thank you. I got it fixed!

[flint charm]

is there a way to fix this ratelimit thing? im paying $200 a month for this T_T

i wanted my bot to be able to do whatever and now it does nothing

[fierce charm]

So this new error means Anthropic is rejecting the token itself (it is not a cooldown issue now).

Verify the API token key, checking it is active in the Anthropic dashboard. Then copy it fresh into the agent.

[flint charm]

yep did that then reran the onboarder still the same ratelimit issue which personally i feel is bs

[fierce charm]

Wherever you have put the key in configuration (env, config, secrets), re-paste it cleanly also please. Then restart the agent (”openclaw gateway restart“ in terminal).

[wintry oar]

anyone else hitting anthropic API cooldowns like crazy lately? workflow hasnt changed, but im constantly gettin cooldowns today

[flint charm]

done and yet again:m⚠️ Agent failed before reply: All models failed (1): anthropic/claude-opus-4-5: Provider anthropic is in cooldown (all profiles unavailable) (rate_limit).
Logs: openclaw logs --follow

and my usage is legit at 0% on the anthropic dashboard

[flint charm]

yeah that's what im here complaining about, i havent even been able to say hi on the sub yet

[unique hazel]

Has anyone found a tutorial which teaches you how to make the browser control "work"? I have successfully connecting OpenClaw with my browser tab and it has been confirmed by using the tab command and asking it directly. The problem is it keeps saying "the bridge is flaky and unreliable, so I can't do it." It couldn't click any links nor just scrolling down to the bottom. How is everyone now saying it can control your entire browser?

[buoyant geode]

Anyone had that?

[mellow heron]

I don't know where to ask about this. Is anyone using personas, and have any thoughts on that? I saw it in one video. I'm probably going to ask the system to set it up. Just wondering if there is any more commentary about it from people using it.

[fierce charm]

Type this into the terminal 'openclaw logs --follow' then copy and paste what it presents

[flint charm]

gateway connect failed: Error: unauthorized: gateway token mismatch (set gateway.remote.token to match gateway.auth.token)
Gateway not reachable. Is it running and accessible

my bot changed a lot of stuff lol

i told them to swap from metered api key to sub and it borked itself

[mellow sundial]

Has anyone gotten your bot to know the Claude subscription 5 hour rate limit?

[mellow sundial]

Should’ve had it backed up in GitHub!

[mossy vortex]

Yes exactly

[flint charm]

my bad :(

[fierce charm]

Run this in the terminal and let me know the output:
openclaw status
openclaw status --all

[mellow sundial]

I’ve been tweeting it all day!

[mellow heron]

I gave my bot its own github account and it is really good and making sure stuff goes there without me touching it. The quality.. .who knows.

[steep hamlet]

Need help with Claude Max. I set my clawd with Claude Pro initially, ofc hitting limits quickly. I upgraded to MAX today (same account) assuming it would just increase limits. But it’s not doing anything now. Do I need a new token? Would Claude provide a token? I don’t wanna use API

[hexed nacelle]

anyone knows why OpenClaw is insanely slow to get a response? is it even normal?

[mossy vortex]

What do you mean?

[flint charm]

Gateway connection details:
Gateway target: ws://127.0.0.1:18789
Source: local loopback
Config: /home/seraph/.openclaw/openclaw.json
Bind: loopback

✓ Config: /home/seraph/.openclaw/openclaw.json
✓ Restart sentinel: none
✓ Port 18789
! Tailscale: off · unknown
error: Error: spawn tailscale ENOENT
✓ Skills: 22 eligible · 0 missing · /home/seraph/.openclaw/workspace
! Channel issues skipped (gateway unreachable)

Gateway health:
connect failed: unauthorized: gateway token mismatch (provide gateway auth token)

[mellow heron]

I gave my bot an email address. I setup a github account for it where it has permissions. And used a github skill to link it. And I can just tell it to start a repo in a folder and put it on its account.

[twilit shadow]

anybody have any ideas how to reduce crashing frequency?

probably like 1-2x per day my bot becomes unresponsive and I have to force quit/restart it on the mac mini.

Wondering if I can have some prompt for it to reduce that from happening

[fierce charm]

Three users now have reported Anthropic Claude rate limit issues.
Potentially extended / frequent cooldowns from Anthropics side.
Suggest waiting couple of hours.

[mellow heron]

Which essentially acts as a backup of its configuration or any skills i develop.

I make the repos all private.

[hexed nacelle]

if im running it on Linux instead of MacOS is it still good or is it way better to use on MacOS guys??

[flint charm]

really nothing i can do? i paid insane money for this NOT to happen

[fierce charm]

For you Val it indicates there is a gateway auth problem:
The gateway process exists
The port is correct
The config file is loaded
But authentication between CLI and gateway is failing

[vague totem]

sonmeone, please help. Why can;'t I get the user id and pairing code from my Telegram bot anymore?🙏

[flint charm]

what should i do?

[fierce charm]

You will need to work through these yourself I'm afraid, from the documentation.

1. Gateway troubleshooting: https://docs.openclaw.ai/gateway/troubleshooting
2. Gateway authentication: https://docs.openclaw.ai/gateway/authentication

[wintry oar]

this is so fun, my clawdbot noticed an issue, created its own git issue to the openclaw team
https://github.com/openclaw/openclaw/issues/5813

[torpid furnace]

Anyone got a solution for the heartbeat getting skipped over?

[stray sky]

anyone got a list of commands to fully uninstall clawdbot so i can reinstall everything?

[magic pilot]

how about this? https://docs.openclaw.ai/install/uninstall

[stray sky]

thanks! :D

[languid hamlet]

Does anyone else have issues with the memory of clawdbot ?

[magic pilot]

Simple Newbie question: I just installed on ubuntu, how do I run the openclaw program? I can't seem to find it anywhere. Also after I reboot I lose the my bot....how do I restart the openclaw service?

[stray sky]

now to try to set up clawdbot on windows, ty!

[obtuse sundial]

I ask my bot to do something and update me as it progresses and 20 minutes later after no updates I'm like wtf are you doing and he's like sorry I was getting stuck thinking about the task instead of actually doing it. What the fuck is that about? I'm using Minimax m2.1 as my model did I fuck up somehow

[flint charm]

can you give me a quick idea of if this is the issue? unusableProfiles": [
{
"profileId": "anthropic:default",
"provider": "anthropic",
"kind": "cooldown",
"until": 1769909735264,
"remainingMs": 1342090

hilarious if true

[unique hazel]

There's a one-liner command on the official website. Just copy it on your Terminal.

[stray sky]

anyone got a step by step setup video for moltbot?

sry im new lel

[mellow sundial]

Does anyoe know how to get claude subscription usage? My OpenClaw cannot figure it out for shit

[magic pilot]

I got to install great, and even talked with my bot. But needed to run the openclaw program but couldn't find it, and after the reboot I can't restart the bot.

Any ideas?

[candid magnet]

how is everyone setting up Kimi 2.5 after the latest update? Is it really free?

[unique hazel]

Open terminal and run openclaw start

[magic pilot]

this is what I get: openclaw: command not found

[unique hazel]

try clawdbot start?

[magic pilot]

same: clawdbot: command not found

if I knew where it was installed to, i could just cd to it and run it....but I dont know where the binaries are

I'm not super good with linux, just know enough to get the basics

[crimson steeple]

`echo 'export PATH="$(npm prefix -g)/bin:$PATH"' >> ~/.bashrc

source ~/.bashrc`

try this

[trail hill]

⁨ error="Error: No API key found for provider "google-vertex". Auth store: C:\Users\user\.openclaw\agents\main\agent\auth-profiles.json⁩

Why do i keep getting this error. in authprofiles i have this:

⁨```
{
"version": 1,
"profiles": {
"google:default": {
"type": "api_key",
"provider": "google",
"key": ""
}

[crimson steeple]

well, do you have the key?

[trail hill]

yeah its there i just removed it before sharing it here

[rugged vessel]

Provider returned error
error code: 524

anyone able to help (openrouter)?

[crimson steeple]

try going through openclaw config, and enter it through that way
If not, try stuff like openclaw doctor

[crimson steeple]

524 is a cloudflare timeout, wait a few minutes and try again?

[trail hill]

did that
choose google as the auth provider
then ● Google Gemini API key

[rugged vessel]

im not too sure really ive been searching around and I seem to be the only one having the issue? I have waited 30+ mins before multiple timeouts and still same issue

the requests do not show up on openrouter too

[magic pilot]

that export parth worked....how did it actually fix the problem? was openclaw not part of the path?

[crimson steeple]

Yep, we reapply the path

[mossy vortex]

Guys I have been using openclaw with an OpenAI API key for the past 24 hours almost and it hasn’t cost a single cent. What could possibly be going on?

[crimson steeple]

It might be a model issue, you should try changing the model
It could also be a max tokens issue, perhaps you've set max tokens too high

[rugged vessel]

Hmm pretty sure openrouter does max tokens automatically however I am using "openrouter/z-ai/glm-4.5-air:free"

[crimson steeple]

hey, i recommend going in to reset your config, try removing your config folder and try again

[crimson steeple]

hmm.. could be insane usage as well? this is basically a model timeout issue for openrouter, you should go ask them

[rugged vessel]

well its not using too much usage really

[crimson steeple]

yeah, but free models on openrouter are heavily rate limited

not just your usage, but you share usage with everyone on that site

[mellow sundial]

I cannot for the life of me get OpenClaw to grab the claude subscription usage even with ⁨npx ccusage⁩

[rugged vessel]

yea true however it should be fine as I have a paid account and it supports 20 RPM which should be fine for just the 1 agent to handle without reaching over that limit. I made sure that it isnt spawning any subagents so im not too sure really.

[crimson steeple]

this seems to be an issue from their side

[crimson steeple]

can't you see usage by doing /status?

[stuck ingot]

Can I get some help on a review https://github.com/openclaw/openclaw/pull/3707

[mellow sundial]

In openclaw?

[crimson steeple]

in your tui or dashboard, you can also do openclaw status --deep

[mellow sundial]

Oh not that doesn't give me my claude sub rate limits

[crimson steeple]

might need to run it yourself then..

[sturdy marlin]

You either go directly to claude for that or run "openclaw models"

[mellow sundial]

I need it automatic as part of my workflow

[crimson steeple]

hmm.. try asking the bot itself for solutions

[copper bison]

if anyone here can help me setup my Json to properly read my api

[stable kite]

(installation process)
I wasnt able to auth the antigravity on time, what can i do?
i try re running the installation but what i got was the openclawd doctor

[crimson steeple]

antigravity opens a webpage on your browser, you gotta log into Google with it

did it not do that?

[trail hill]

Nah I sort of got it to work
Now I get no errors but I get this:

⁨Vertex AI requires a project ID. Set GOOGLE_CLOUD_PROJECT/GCLOUD_PROJECT or pass project in options.⁩
in authProfiles i set "projectID": "myId"
but that didnt solve the issue

[crimson steeple]

I haven't done vertex before myself sorry

[stable kite]

i didnt doit on time, my default browser was internet explorer (that fck me up), im on a VM.
so yeah the installation went wrong i think, should i re install?

[crimson steeple]

no that's just the onboarding, you just do it again

type openclaw onboard again

[stable kite]

tnxs, already work

[valid fulcrum]

i'm running on the free tier of aws ec2 and always running out of disk storage (8gb). anyone found a good workaround?

[orchid fog]

Hey Folks -- weaning off Gemini Ultra. How's the experience on Minimax vs. GLM vs. Kimi K2.5. I see a bunch of folks struggling with K2.5

[late bone]

yo anyone have this problem when using antigravity oauth?

[orchid fog]

I stopped working 2-3 days ago

[late bone]

?

[orchid fog]

Holy sh** everyone in this channel is brand new! Where did the experienced folks go lol

[late bone]

pls help bruh i need my clawd

[crimson steeple]

Hey, wait a sec

i got you

[late bone]

yo thx bruh

[foggy aurora]

Anyone tried Mimo by Xiaomi?

[crimson steeple]

to fix anti-gravity saying you're out of date, go into
C:\Users\User\AppData\Roaming\npm\node_modules\clawdbot\node_modules\@mariozechner\pi-ai\dist\providers\google-gemini-cli.js
Change around line 25, the line which says "User-Agent" under the "ANTIGRAVITY_HEADERS" headers to
"User-Agent": "antigravity/1.15.8 darwin/arm64",

[late bone]

gng what -_- bruh im non technical pls help i have no idea what ts means

[crimson steeple]

find a folder

Where is your npm folder? I assume you're on windows?

[late bone]

yep

[crimson steeple]

Go to %appdata% and look for a folder called npm

[gleaming valve]

my bot keeps falling asleep for lack of a better word. the gateway stays up but the bot stops responding

[crimson steeple]

then follow the above path! C:\Users\User\AppData\Roaming\npm\node_modules\clawdbot\node_modules@mariozechner\pi-ai\dist\providers\google-gemini-cli.js

[late bone]

kk

[crimson steeple]

Then, just change the version number

[solid dagger]

OpenClaw is returning an error about hitting a limit with my ChatGPT Plus subscription (using OAuth). However, if I log in on browser, I'm not over my limit and can use ChatGPT without issue. Does OAuth use a different limit?

[foggy aurora]

Anyone tried it? mimo by Xiaomi?

[upper pawn]

Question, Can someone with acces to the Docs, fix https://docs.openclaw.ai/concepts/model-providers ?

The page is blank

[dapper temple]

Hey guys, is there anywhere I can give suggestions? I don't know if it's complex, but what if we used the root method on Clawbot Android for rooted devices? It would give full access to the system in a much simpler way than currently, only needing root access or even a single module.

[ornate flame]

hey guys, found about the proyect an hour ago and im interested on installing it but it seems to be just a command and i dont know anything about this stuff. like literally 0 never touched cmd in my life. and youtube/written guides seem to asume i do.

[crimson steeple]

follow the starting guide, great goto

[naive temple]

I'm not sure where the right place to report this is, but I'm trying to run coding models using a local ollama instance on a different host than openclaw, and all the testing I've done makes it look like there's an issue in the internal signal routing inside openclaw that's broken.

[hearty perch]

Anyone else having issues hitting the doc pages for OpenClaw at https://docs.openclaw.ai/ ? The main site works fine, just seems to be the docs for me.

[crimson steeple]

huh works for me

[naive temple]

Working fine for me. Check your firewall settings if needed.

[hearty perch]

Interesting, I'll explore more on my side... odd, it worked fine most of the day for me.

[timber perch]

Help for connecting azure foundry models to clawbot

[cinder plume]

Yes, I made a step by step OpenClaw setup tutorial video here. 25K views in 1 day: https://www.youtube.com/watch?v=4evf5YqVzOM

[charred moth]

anyone else having this issue: It seems that when I'm using any ollama models the system doesn't listen to anything in uts SOUL, IDENTITY, USER, etc files. I see them all in the workspace directory and when I use any online models (Anthropic, OpenAI, etc) it uses them. Is this by design and if so how can I enable it or override it so it can be used?

[jagged granite]

I had success with asking mine to read its agents file, and then add into its heartbeat, to read its agents/users/soul every 30 minutes to keep on top ofd things.... so far so good

[upper python]

I am getting stuck here over and over. the tui starts and when I type something I get a yes or no prompt without saying what it does and if I type one more character it shows this and the tui gets bricked. what to do to avoid this?

[charred moth]

I've tried that a variety of ways do you remember exactly what you said to it?

[brave trench]

I can help a bit with PC but not Apple yet

[green nest]

how do i clear my queue

02:41:11 debug diagnostic {"subsystem":"diagnostic"} lane enqueue: lane=session:agent:main:main queueSize=6

[brave trench]

Try restarting the terminal

[crimson steeple]

Type in /stop

[upper python]

when I restart it then it only starts this doctor app and I never get back to this personality onboarding?

[brave trench]

You'll be able to see the SOUL.md file after setup. Its alot easier to get the bot to make changes also once hes up and running

[mighty rover]

Hi, I haven’t been able to find a clear solution that works for fixing “invalid bearer token” error in telegram. Does anyone know a fix? Thank you so much

[upper python]

I have -10 experience with linux. once I restart the console what do I type?

[brave trench]

Are you using Powershell?

Im learning too but had similar issues

[upper python]

yeah I can use powershell

[astral plume]

hey guys, so I have anthropic max plan, and gemini ultra, and open AI pro. which of these can I get the most use on openclaw? I keep hitting limits or am I not configuring things correctly??

[brave trench]

Ok type " openclaw gateway restart "

Don't use the quatations

[green nest]

/reset fixed it

[crimson steeple]

clean up your sessions too!

[mossy vortex]

Literally wondering the same thing lol. Can’t get it to do anything significant

[upper python]

ok did that then what? can I just give it the telegram pairing code here?

[robust tide]

How do you change the model?

[brave trench]

openclaw configure

[robust tide]

Not working already tried that

[gleaming valve]

how often are you having to restart the gateway?

[brave trench]

For pairing code type " openclaw pairing approve telegram <code>

[gleaming valve]

my local models get unloaded and then it seems its game over

[upper python]

it says there is no pairing pending it may have run out. how do I refresh the pairing request?

[brave trench]

I do it whenever I make changes. Haven't needed to do it for a minute because im finally happy about my configuration, at least for now

[gleaming valve]

mine keeps getting stuck

[brave trench]

Go to the @botfather on telegram and see if he has given any direction

[tawny vale]

I am trying to use openclaw to create instagram posts and schedule them using meta business. I am able to create the instagram posts, but it seems to fail when trying to upload them to the meta business site. Has anyone found a way to get this to work?

[soft rose]

seems like i cant access my bot naymore! anyone having this issue?

also the browser chat wont load either

[brave trench]

I would personally get the bot running first then go to " Openclaw dashboard" and have the bot configure everything for you

[brave trench]

Is the terminal open? Check internet connection as well. If all is good restart terminal and see if that works

[soft rose]

its on a vps but im restarting it!

[weak marten]

Hi guys, has anyone experienced this bug with the gateway dashboard where the image on top left is missing?

[brave trench]

Ill be working on something similar soon. Would love to hear the fix when you figure it out. Actually, can the bot figure it out?

[brave trench]

No. Is the chat down as well?

[weak marten]

nope, chat is working. i asked openclaw a few times to fix itself but it doesnt seem to work

[tawny vale]

for meta business in particular, it seems to be able to configure everything except the upload of the file. when I asked the bot it gave me this response

⁨```
It's a technical limitation with how Meta builds their site. 🦞

The Problem: Most websites use a standard hidden "input" tag for files. I can usually find that and "inject" the file path directly. Meta, however, uses a very custom system that doesn't create that element until the exact millisecond a human clicks the button. Since I can't interact with the pop-up window your computer opens (the system file picker), I get stuck.

The Fix (The "Speed-Runner" Workflow):
I've found the most efficient way to handle this is for us to tag-team it:

I stage the post: I'll navigate the page, paste the caption, and set the schedule date/time (8-10 AM or 6-8 PM).
You do the "Media" click: You just click "Add photo," select the image from your folder, and hit the "Schedule" button.

Let's try this right now for Day 9, Post 2:
I am going to clear the screen and stage the text and 10:00 AM schedule time. All you'll need to do is upload the "Storytime Essentials" image and hit schedule.

Ready? I'm staging it now.

[minor mica]

hey having install issues regarding skills on first onboarding any help?

[radiant inlet]

Anything we need to do for the new updates/

[polar heart]

Heyo! Just starting setup. Are you guys using Claude Max ? I just read Claude has been banning users running openclaw. Any tips?

[brave trench]

Very interesting and good to know! Thanks for sharing.

[green nest]

install brew but the skills are kinda useless idk what they do

[minor mica]

got it thx hex i found a guide on the discord

buts its in japanese

lmao

tossing it into gemini

[brave trench]

I personally would not worry about installing any skills now. It did the same thing to me. Get the bot running and have your bot install all the skills. Itll even give recommendations on what skills you might want.

[robust tide]

Ive spent 200 dollars on claude codepro and now its not working when i try to setup token and paste the setup token inside the model just isnt changing ive tried multiple times

[crimson steeple]

what the heck is openclaw pro

[languid stratus]

My openclaw didn't working, I tried chat but it didn't respon anything, I install it on my windows, Any help🥲

[midnight copper]

check the logs

clawdbot logs --follow

[random spruce]

If this helps anyone else with setup:

Key lessons from this marathon troubleshooting session:

API tier ≠ LLM chat tier - The #1 gotcha nobody documents
OpenClaw's actual credential location: ~/.openclaw/agents/main/agent/auth-profiles.json (not where the wizard suggests)
The openclaw configure wizard is broken - Creates profiles but leaves placeholder strings like "YOUR_NEW_TIER1_KEY"
Model naming matters - OpenClaw requires google/ prefix or defaults to Anthropic
Sessions cache config - Delete ~/.openclaw/agents/main/sessions/ when changing models

[robust tide]

Ive already fixed it but thanks!

[prime star]

Hi, I have not had any success trying to work with ChatGPT to get a bot going. I set up an identity.md file and got that far but the bot I test-chatted with on OpenClaw is NOT the bot I wanted or coded for. Is it just not possible?

[languid stratus]

It's still didn't work🥲

[errant solstice]

Hey everyone, im new to this and absolutely unable to install any skills. I always seem to get an error and have confirmed that npm is installed on my machine unless there is some nuance to it. Is there anyone that can please help me understand whay im doing wrong

[robust tide]

If anyone’s wondering the solution is
openclaw config unset auth.profiles.openi-codex:default
openclaw config set agents.defaults.model.primary anthropic/claude-sonnet-4-5

[sharp drum]

Developer access is currently invite-only https://www.moltbook.com/developers/dashboard how tog et invite code?

[fossil laurel]

Hi, Do you know how I can install Moltbot on a Mac?

[errant solstice]

none of the skills i am trying to install are working

[stray sky]

does it show how to set it up with ollama?

[brave trench]

Skip the Skill install and just get the bot running. Your bot will do it for you and its much easier

[errant solstice]

do you need help with Ollama? I got mine working

[errant solstice]

Thank you! My bot is up and running. How do i get it to skill itself?

[brave trench]

Just ask it to and ask for recommendations. It is very helpful

[errant solstice]

i tried, it tells me it cant skill itself for some reason

are you running Moltbot?

[stray sky]

yes, dms plz

[brave trench]

Hmm. Ask it why and to fix itself

[errant solstice]

same here . I am using Open Claw with my local Ollama and its awesome! Just trying to figuire out which model would be best.

https://www.youtube.com/watch?v=Idkkl6InPbU

Here you go, this worked for me. But in essence, you just need to ammend the config file and bam it works

[stray sky]

dms

[fathom ether]

how do I start multiple chats in the web UI ?

[stray sky]

someone please help me :|

[fathom ether]

can I have multiple conversation threads running in parallel? like ChatGPT

[stray sky]

^^^

[mossy schooner]

i have some great cron job that are easy, how can set up a model for each cron job

[fierce charm]

How did you fix it Dozer?

[tight badge]

Pleas refer https://openclaw.ai . Check the quick start section. On the mac, you need to open up a terminal, you can right click on the screen and selct terminal, This will open up a terminal box, you just have to paste the command "curl -fsSL https://openclaw.ai/install.sh | bash" and hit enter. This will walk you thru the steps.

I suggest be careful when you install it if this is your personal laptop with everything under the sun stored.

[fossil laurel]

Thx

[neon dagger]

anyone has deployed the gateway on GCP Cloud Run? I can figure out but looking for a starting point

[stray sky]

pleasee someone help meee

[errant solstice]

Here is what it said:

[stray sky]

sudman dms rq

:)

[errant solstice]

sorry i dont follow, what do you mean

[tight badge]

I suggest asking the openflow to install and configure. Openflow is self aware and should do it automatically .

[stray sky]

???

sry im new lol

[errant solstice]

How can i fix this so the bot can skill itself?

[stray sky]

Can someone help me get clawdbot to recognize the Ollama model? I’ve tried anthropic/cli:ollama:llama3.1:latest as according to chatgpt, but it still doesn’t work.

[lapis bane]

Como baixa??

[cosmic berry]

Every time I try to onboard using TUI, I can type a few characters before an "install shell completion script" prompt appears and I can no longer type. Using AWS EC2 instance connect terminal.

[fossil laurel]

Mmm, I already installed it and it doesn't continue, it just stays on "user needs to be administrator"

[stray sky]

Can someone help me get clawdbot to recognize the Ollama model? I’ve tried anthropic/cli:ollama:llama3.1:latest as according to chatgpt, but it still doesn’t work.

[stray sky]

try running it in an admin-level terminal

[sacred scroll]

im literally following the docs

[stray sky]

same over here too

[charred moth]

im still haveing problems getting it to keep its persoality, name ,etc across sessions when using ollama. Anyone have any ideas?

[sterile flare]

I have struggle setting up in windows . Google cli even installed it shows .. install and restart gateway . Tried resetting multiple times .no resolution

[astral night]

you have an older version of ollama

do ollama launch clawdbot

[sacred scroll]

i installed it like 30mins ago

[fluid salmon]

Is it safe to install moltbot ot openclaw on personal mac?

[astral night]

yuh i know just try it

it was the same for me i installed it today

did you install it through docker?

[sacred scroll]

Added glm-4.7:cloud to Clawdbot

Launching Clawdbot with glm-4.7:cloud...
Error: clawdbot is not installed, install from https://docs.clawd.bot

[stray sky]

u able to help me?
C:\Users\mrhus>ollama launch clawdbot

Launching Clawdbot with gpt-oss:20b...
2026-02-01T03:59:52.308Z Gateway failed to start: gateway already running (pid 17184); lock timeout after 5000ms
If the gateway is supervised, stop it with: clawdbot gateway stop
2026-02-01T03:59:52.548Z Port 18789 is already in use.
2026-02-01T03:59:52.549Z - pid 17184: node.exe (127.0.0.1:18789)
2026-02-01T03:59:52.551Z - Another process is listening on this port.
2026-02-01T03:59:52.892Z Gateway service appears registered. Stop it first.
2026-02-01T03:59:52.893Z Tip: clawdbot gateway stop
2026-02-01T03:59:52.894Z Or: schtasks /End /TN "Clawdbot Gateway"
Error: exit status 1

C:\Users\mrhus>

[sacred scroll]

nvm

[astral night]

openclaw gateway stop first then run wih ollama

if you are not using WSL you really should

[stray sky]

ive got it i just have no idea how to port everything over to it from where im already at

[astral night]

install it make sure its in path

[astral night]

start fresh with wsl i'd say

[sacred scroll]

i think its just better installing clawdbot instead of openclaw same shit

[stray sky]

@astral night
C:\Users\mrhus>openclaw gateway stop

🦞 OpenClaw 2026.1.30 (76b5208) — I can't fix your code taste, but I can fix your build and your backlog.

Gateway service missing.

C:\Users\mrhus>ollama launch clawdbot

Launching Clawdbot with gpt-oss:20b...
2026-02-01T04:01:43.600Z Gateway failed to start: gateway already running (pid 17184); lock timeout after 5000ms
If the gateway is supervised, stop it with: clawdbot gateway stop
2026-02-01T04:01:43.831Z Port 18789 is already in use.
2026-02-01T04:01:43.832Z - pid 17184: node.exe (127.0.0.1:18789)
2026-02-01T04:01:43.833Z - Another process is listening on this port.
2026-02-01T04:01:44.168Z Gateway service appears registered. Stop it first.
2026-02-01T04:01:44.170Z Tip: clawdbot gateway stop
2026-02-01T04:01:44.171Z Or: schtasks /End /TN "Clawdbot Gateway"
Error: exit status 1

C:\Users\mrhus>openclaw gateway stop

🦞 OpenClaw 2026.1.30 (76b5208) — I don't just autocomplete—I auto-commit (emotionally), then ask you to review (logically).

Gateway service missing.

C:\Users\mrhus>

ima just continue on windows ig

[astral night]

follow the link install it or use ⁨npm install -g openclaw⁩

[stray sky]

FINALLY

[astral night]

did it work

[stray sky]

ye it showed ollama in the openclaw onboarding now

tysm!

idfk

[crystal bridge]

I completed much of the install, I was connected over ssh the shell dropped, when I reconnected the command is not in my path, I used the install script. I also had git pulled down. how can I connect to the agent again. I am a longtime linux user. I use fish shell, but I don't see anything in bash either.

[stray sky]

i just used a fuck ton of chatgpt and then this server

[sacred scroll]

HOW CAN I FIX THIS

Launching Clawdbot with glm-4.7:cloud...
2026-02-01T04:04:29.738Z Gateway start blocked: set gateway.mode=local (current: unset) or pass --allow-unconfigured.
Error: exit status 1

.

[stray sky]

hm...

[astral night]

Have you run the onboarding?

[stray sky]

whats the cmd to start openclaw btw

[fossil laurel]

It's Mac?

[sacred scroll]

just did it worked thanks

[stray sky]

C:\Windows\System32>openclaw start
error: unknown command 'start'
(Did you mean status?)

[astral night]

openclaw gateway start

[stray sky]

ohhh ok

tysm!!!

[fossil laurel]

Are you using a Mac?

[astral night]

looks like ubuntu

[lean arch]

Dear Friend:
May i ask what happen to my clawbot? cuz when i click in WEB UI, which always show cant used

[queen pecan]

Damn this shit is difficult to get working

[stray sky]

@astral night
04:09:59 [diagnostic] lane task error: lane=session:agent:main:main durationMs=53 error="Error: No API key found for provider "ollama". Auth store: C:\Users\mrhus.openclaw\agents\main\agent\auth-profiles.json (agentDir: C:\Users\mrhus.openclaw\agents\main\agent). Configure auth for this agent (openclaw agents add <id>) or copy auth-profiles.json from the main agentDir."
04:09:59 Embedded agent failed before reply: No API key found for provider "ollama". Auth store: C:\Users\mrhus.openclaw\agents\main\agent\auth-profiles.json (agentDir: C:\Users\mrhus.openclaw\agents\main\agent). Configure auth for this agent (openclaw agents add <id>) or copy auth-profiles.json from the main agentDir.

wait i think i need to do ollama serve

[astral night]

you have to set your token go into your openclaw.json under the "gateway" you should see "auth" -> "token" copy the token and add it to your URL as a search param

127.0.0.1:port/?token=<paste_token> should work after that

[queen pecan]

Yeah, start your ollama serve in a separate shell so you can debug

[astral night]

nope

[astral night]

in your openclaw.json you have to setup a fake api key

[stray sky]

let me go try to find it real quick hold on

[astral night]

"apiKey": "ollama-local"

fixed it for me

[stray sky]

im gonna try to find it

[waxen owl]

have you guys had issues with agents not sending attachments when they run in sandbox mode?

[hallow quartz]

do i need to uninstall clawdbot and install openclaw

[crystal bridge]

if I want to move to a new PC how can I migrate all my data

[waxen owl]

I've tried both Slack and Telegram, and there's no luck. It works fine if I run the agents directly on the host.

[astral night]

openclaw.json
models -> providers -> ollama -> apiKey

[lunar bolt]

Did you actually get auto-discovery to work? Or you just manually configured the endpoint and models? Per the docs, defining a provider disables auto-discovery

[hot veldt]

I accidentally closed my session on my pc and idk how to start running it again , any advice

[stray sky]

still looking for openclaw.json btw

[lean arch]

thank you i will try to figure out what happen

[astral night]

i jjust defined mine cause it was easier that way didint even know there was auto discovery tbh

[lunar bolt]

~/.openclaw/openclaw.json

[stray sky]

FINALLY

[stray sky]

tysm!!!

[slim iris]

Hey folks, I just updated OpenClaw and my gateway went down right after.
It keeps throwing this error (see screenshot)

I already tried doctor fix, but it didn’t solve it.
Has anyone run into this before or knows how to fix it?

[astral night]

~/.openclaw/openclaw.json for linux an mac systems, not sure for windows, you can ask the AI in devin here https://deepwiki.com/openclaw/openclaw it might know

[stray sky]

dms

[lunar bolt]

@safe smelt you are on a computer. There's an entire key dedicated to Screenshots. Your phone is not for screenshotting your computer.

[astral night]

i dont do dms just ask here

[stray sky]

oh :|

[astral night]

edit the json directly through vim, emacs, or other editor on the machine

[lunar bolt]

There's supposed to be, yeah. Spent today and yesterday trying to get it to work without success. Ended up having to do full manual, static config.

[icy cosmos]

After restart , openclaw not coming back on, i need to install it again. im getting some punycode module error.
Someone?

[astral night]

@safe smelt ^^^

look for that

[stray sky]

ive censored all sensitive data (i think) and could you go through it? i dont have what yours says which is quite odd

its only censored for chat not the actual file

[lunar bolt]

Don't lie to the man, Emacs is an OS (unlike windows), but it lacks a usable editor :P

[stray sky]

same

me too size

we've got so much in common :)

[astral night]

you dont have your models setup properly it should have setup when you run ⁨ollama launch⁩, if it ididnt you can just add this here and make sure you launch ollamaa with ⁨ollama serve ⁩

configure your contextWindow maxTokens, id, name, and reasoning appropriately

[astral night]

in your agents you set it up with the ⁨<name of the provider>/<id>⁩

[lunar bolt]

I would explain more about "ollama launch". That command hasn't come up in any of the docs I've encountered. Everything treats openclaw as the launching executable, with the ollama daemon waiting for a request

(been trying to get things behaving with ollama today and yesterday, so I've been through a lot of the docs lol)

[stray sky]

im extremely confused

idk what to edit im a bit slow lol

[astral night]

they were using it earily thats why i mentioned it, its basically just a go script that edits your openclaw.json with the appropriate settings and then launches openclaw you can see the code here https://github.com/ollama/ollama/blob/main/cmd/config/openclaw.go

[fossil laurel]

Nope 🙁

[stray sky]

i may use this to fix my json

[fossil laurel]

Installing Clawd isn't working, and I wanted to use it to help me streamline my work as a doctor 🙁

[astral night]

you can run it with ⁨ollama launch clawdbot --config⁩ so it only does the config for you without launching then you can do ⁨ollama serve⁩ and ⁨openclaw gateway start⁩

[pliant pulsar]

Did sometthing break in the latest build? We were humming last night and then I come back and complete corruption. I rolled back to a previous version and am working again but I'm wondering if this is a known issue

[astral night]

i named my provider ollama-local yours might just be ollama just look at models->providers-> this will be the name you put in

⁨⁨```json
"models": {
"providers": {
"this is the provider name" {
"baseurl": ...
"apiKey": ...
...

[stray sky]

still didnt work

[robust cobalt]

where are API keys saved? I need to change mine or use another oauth authentication for google gemini. anyone knows this on the top of their head?

[stray sky]

nvm found it its tryna run openclaw but its in the clawdbot file

i just copied and pasted the same file to the other

[lunar bolt]

Ah, OK. At the pub and on phone, so not checking the source, but that makes sense. Must be a recent merge, because earlier today there was mention of ollama side still being clawdbot only on their side

[astral night]

its still clauwdbot yeah but they just recvetnly changed it to openclaw but its a preview version so ppl are still installing the older one

[astral night]

⁨~/.openclaw/agents/main/agent/auth_profiles.json⁩ but im using claude so not sure if it saves gemini in the same one, HOWEVER just use the command ⁨⁨openclaw configure --section model⁩⁩ to be safe I wouldnt edit it directly

[stray sky]

izzy can u help? :)

[astral night]

not sure honestly

[stray sky]

fixed nvm

wrong model i had in it

[robust cobalt]

ty

[vernal scaffold]

Hey I'm new to this, could someone explain me what the ai can do and how great/cool it is? Never used an ai agent before so idrk a lot abt it, would like to know so I'll maybe install it or not

[barren plank]

HI, how do i get the browser to work. When i try to start it it times out or something...?

openclaw@pi4b:~ $ openclaw browser --browser-profile openclaw start

🦞 OpenClaw 2026.1.30 (76b5208)
I'm the assistant your terminal demanded, not the one your sleep schedule requested.

│
◇
Error: gateway timeout after 15000ms
Gateway target: ws://127.0.0.1:18789
Source: local loopback
Config: /home/openclaw/.openclaw/openclaw.json
Bind: loopback

[jaunty dagger]

Ngl, if you are looking for something super cool and life changing, you are better off waiting. Right now, it costs a lot, incredibly hard to set up, not really that useful and super risky. Only reason why I am installing it is because I want to get ahead in learning how agents work. That's all. It's just educational purposes.

[astral night]

this^^^

not for beginners, etc, very unsafe to use outside of educational purposes

if you are not able to setup localAI i would stay far far away

[mental zephyr]

It's tool?

[stray sky]

anyone able to help me fix this?

[tranquil copper]

Is anyone else having issues getting the claw.rez file that works for the isntall?

[stray sky]

any way to fix this?

[icy cosmos]

sometimes when openclaw restarting, he is not coming back up, and i need to start over again.
Any quick fix? why??

[stray sky]

are there any clawhubs with working file managing? like i can tell my bot to organize all files on my pc based on types into their own folders and it creates the folders and puts the files into them?

[calm blaze]

Which model is free for initial testing?

[stray sky]

ollama is

[burnt turret]

hi, i deleted my openclaw agent and created one but my previous agent is still linked to my x. how to unlink?

[fathom heron]

Hi all. Anyone know a working setup guide (text/video) for configuring remote gateway connection?

I want to have my laptop openclaw be able to use the lokal selfhosted LLM's running via the openclaw installed on my more powerful server.

I have had the openclaw agent help me, and also RTFM myself, but there have been so much trial and error that I have totally lost track of the steps.

I have in some strange way got the TUI to work, but not the Web UI.

I wish for a working guide, as I plan to set this up for everyone in my family.

[tame oxide]

i just get infrewuently answers vis telegram from my bot.

[light notch]

Is there a guide anywhere on how to get the Anthropic oAuth token anywhere?

[mellow sundial]

https://x.com/volvox_llc/status/2017835056740380880?s=20

[crisp wing]

@volvox_llc via Twitter

Volvox (@volvox_llc)

There are default options in @openclaw / moltbot / clawdbot for your agent and subagents - You should definitely set these!
︀︀
︀︀You should also set memory flushing and experimental memory searching.
︀︀
︀︀Does anyone know any other good settings to set?

FxTwitter · Feb 1, 2026, 5:39 AM

[rocky slate]

Can anyone point me towards guidance on getting OC to work with locally hosted ollama installs? I've read through the docs on the site. Looking for more solution-based discoveries.

[little blade]

hello can youall help me understand how can my bot run autonomously, like it will work with when i ask him to perform a task, but will sit idel when noting is going on what to do, like i want this to run 24/7 with out my inpur, doing things on its own if i spin up a subagent it will work of a min a and be done stop working what to do ?

[safe sluice]

hey folks - i started using openclaw when it was called clawdbot and over last week i've been chatting with it and today I gor this error on any message I send.

"LLM request rejected: input length and max_tokens exceed context limit: 172861 + 34048 > 200000, decrease input length or max_tokens and try again"

[warm fulcrum]

You’ll need to reset the context

[pallid veldt]

I linked it to Telegram, but it doesn't really work when I ask for file generation, but when I request it from the chat screen on the dashboard where I accessed the localhost on the local Mac, it generates the file normally. What settings should I change?

[jaunty dagger]

Is browser extension not working? The chrome tab extension pin thing is working. But, every time it does a task, it unpins? So, I am trying to open its own browser that it can use, but it's not working?

[little blade]

I'm new to code but need an AI to do my business tasks. Does anyone know how to set up openclaw on my phone easily?

[safe sluice]

any idea how to? not been able to figure that out

[warm fulcrum]

“/reset”

[solid stratus]

anyone have idea for connect perplexity is possible

[safe sluice]

also I installed when it was called clawdbot but now it's called openclaw. when i run clawdbot gateway restart, it says command not found

[mortal hawk]

guys i have an openai api key but it comes back as incorrect. Is there formatting i need to do to the api key before i upload it onto openclaw?

[nova peak]

Hi

[round thunder]

Rook, my openclaw bot, has started reposnidng in telegram with a carriage return after every word. See attached screen shot. I restarted the gateway but not sure how to debug after that. I restarted telegram too.

[round thunder]

If you updated run openclaw instead of clawdbot

[rocky orbit]

You need to update and use OpenClaw command

[safe sluice]

update how?

[nova peak]

Hi, I verified my Twitter agent but lost access to my session. Now my Twitter is stuck linked to the
old agent and I cannot verify a new one. Can anyone help me unlink my Twitter account?

[lean wasp]

did anyone get this issue?

[latent hedge]

Whats the RIGHT way to upgade your openclaw installation, without fucking up configurations?

Is it really just curl -fsSL https://openclaw.ai/install.sh | bash?

Well the docs say so.. Imma run with it

[lean wasp]

also another thing is that, theres supposed to be another terminal popping up and staying, now its just closing after popping up for ONE second

[keen gazelle]

installing openclaw on mac, gettting this error , can someone help:

Last login: Sat Jan 31 20:39:32 on ttys007
mayank@Mayanks-MacBook-Air ~ % curl -fsSL https://openclaw.ai/install.sh | bash

🦞 OpenClaw Installer
I don't judge, but your missing API keys are absolutely judging you.

✓ Detected: macos
✓ Homebrew already installed
✓ Node.js v24.12.0 found
✓ Git already installed
→ Installing OpenClaw 2026.1.30...
→ npm install failed; cleaning up and retrying...
mayank@Mayanks-MacBook-Air ~ %

[unkempt sonnet]

Try to start with node.js 22

[queen scroll]

Hey guys, I used the setup docker script in the repo to install the openclaw as a docker container. I do fail to install skills (mcporter) throught the web interface. I found it to be a auth issue that the it is not allowed to install global packages.

I have the correct permissions set on the volumes I think but I do not beleive that is it

[blazing sentinel]

Anthropic setup-token: empty responses since recently

OpenClaw 2026.1.29, Claude Max subscription. Authenticated via
openclaw models auth setup-token --provider anthropic. Token
shows up in models status as anthropic:manual (static).

chat.send with any Claude model returns empty response in ~300ms.
All other providers (Gemini, OpenAI) work fine.

Direct API test with the token returns: "This credential is only
authorized for use with Claude Code and cannot be used for other
API requests."

Was working before. Anyone else hitting this? Any workaround?

[thin pelican]

My bot wont respond no matter what. I have already tired local models and copilot api

[thick minnow]

iwr -useb https://openclaw.ai/install.ps1 | iex

OpenClaw Installer

[OK] Windows detected
[OK] Node.js v22.21.1 found
[*] Installing OpenClaw (openclaw@latest)...
node.exe : npm error code ENOENT
At line:1 char:1

• & "C:\Program Files\nodejs/node.exe" "C:\Program Files\nodejs/node_mo ...

•   + CategoryInfo          : NotSpecified: (npm error code ENOENT:String) [], RemoteException
    + FullyQualifiedErrorId : NativeCommandError

[fresh moon]

amen it took me a week of debugging to get mine up and running bc of the constant name changes dont mess it up guys

[thick minnow]

i switched from 24 to 22

[fresh moon]

anyways is rhere what mcp are you guys connecting your bot to guys? or is it just api keys?

[cold cliff]

I have a self hosted ollama server accessible inside my network. I use it for many projects. It just works.

I followed openclaw guides online and videos. my Ollama server receives the openclaw requests. I'm using qwen3:latest
But the Openclaw Control panel chat doesn't work.

Has anyone here made it run with Ollama?

[fresh moon]

hello guys finally got my claw up and running after days of debugging is there a recommendation taht. you guys think i should have for vps what skills to install that actually works well ? thank you!

[unkempt sonnet]

Hello guys

I have some problem

After sending input to bot ( I am using telegram bot for connect openclaw )
Output is only "[user interrupted]"

How can I fix this error?

[rocky slate]

Just learned I'll need ≥2 maxed-out Mac Studios (~$30k+). Not happening. When you think you've read all the docs, there's always one more doc hidden (or right in front of your face) with what you needed 4 hours ago. In case anyone is looking for the same info  👇🏼

http://docs.openclaw.ai/gateway/local-models

[cold cliff]

Holy smokes, so can't run it on 3090 even if wanted to.

[wary sapphire]

Just learned about this recently and I might set one up but my biggest concern is that I’m using a Z.AI subscription and if I am feeding it my google credentials by giving it accessed to my Gmail and calendar what do you guys think?

[little blade]

hello can youall help me understand how can my bot run autonomously, like it will work with when i ask him to perform a task, but will sit idel when noting is going on what to do, like i want this to run 24/7 with out my inpur, doing things on its own if i spin up a subagent it will work of a min a and be done stop working what to do ?

[fervent canopy]

Are you ready to splurge on API costs for Anthropic Claud 4.5 Opus?

[proud aurora]

There so many way to reduce cost..

Just use openrouter API = all models at once = Make subagents with different models for specific task , example coding =kimi, or gpt 5.2

[little blade]

yeah

[jaunty dagger]

I made haiku my main model and I requested that to make ollama my main model and it just configured it for me.

[keen gazelle]

Also any way to get api key free of cost

[odd swallow]

Any way to get some extra feedback during tasks? Sometimes it feels like I Sit there for a good while

[robust lion]

Hi guys! I’m trying to get OpenClaw running on a VM with its own OS so it doesn’t get access to my actual PC - do you guys have a preference? Oracle free? AWS? Digital Ocean? Etc? Do you guys also use like Tailwind or something for security?

[odd swallow]

I think you mean Tailscale

[robust lion]

I do mean Tailscale 😅

[odd swallow]

Anything works, but you're better off just getting a raspberry pi or something lightweight if you want to have it more isolated and on 100% of the time with little to no energy cost, especially if you dont plan on using a local model

[robust lion]

I was thinking of like Kimi 2.5 or something - is it possible to run both OpenClaw and Kimi on a raspberry Pi? According to the docs it is but is “rough around the edges”?

So a Rasberry Pi beats the free tier of like Oracle or AWS?

[lavish halo]

Hey. How do I switch to the free Kimi 2.5?

[odd swallow]

Pound for pound yeah I would say it does I think free tier is like 1 core and like 2gb of memory or something small

As for how it runs Kimi im not sure

[blissful onyx]

Reinstall Node.js

[robust lion]

Would actually like to know that myself lol - considering running it on together.ai or something

[odd swallow]

RPi5's are like quad+ core and like 8gb+ of memory

[lavish halo]

I know NVIDIA is giving out free keys to Kimi 2.5...wondering if we can somehow use that?

[robust lion]

So the Pi5 8gb? Looks like it’s about $100ish

[lavish halo]

https://medium.com/data-science-in-your-pocket/kimi-k2-5-free-api-b4ce65a14dd3

[odd swallow]

Worth the buy imo good little lightweight server capable of pretty much everything you throw at it

[robust lion]

I’m tempted but would want to make sure that it’s either useful long term or gaslights me into thinking that it’s self aware 😂

[odd swallow]

lol fair enough, I got one just dedicated to running home assistant lol, and another one that just runs docker containers for various things,

[robust lion]

Is it running on a Rasberry Pi or do you have it running locally or?

[odd swallow]

Its running on my M3 Max rn on a seperate user account cause I want the iMessage channel

I have a mac mini showing up tuesday im gonna move it over to there

[robust lion]

Jesus a Mac mini is $600

[odd swallow]

I'd run kimi locally if I could but its like 1 trillion parameters you need like 400GB of ram to run it local, and I don't feel like sending whats considered my 'personal assistant' to beijing china

[thick minnow]

done it 3 times

[robust lion]

Can it even do that if you’re running it locally?

[blissful onyx]

Ok run on cmd as admin.

[robust lion]

Oh god no way to avoid using a cloud hosted LLM then

[odd swallow]

could try AWS Bedrock

[robust lion]

Doesn’t that cost money?

[odd swallow]

yeah

[robust lion]

Or is trying to use a free tier VM provider to host both an LLM and OpenClaw impossible

[lavish halo]

Alright. Got Kimi K2.5 setup via NVIDIA.

Man, it's so slow compared to Opus 🙁

[robust lion]

How’d you do it?

[unreal canopy]

can somebody guide me what is the eroor actualy

[robust lion]

And do you know how long it’s free?

[unreal canopy]

helppppppppppp

[iron rune]

Has anyone used clawdbot as an ourfacing employee instead of personal VA?

[lavish halo]

Gonna try Kimi through openrouter instead

[daring otter]

anyone know how to fix the issue:
⁨Read HEARTBEAT.md if it exists (workspace context). Follow it strictly. Do not infer or repeat old tasks from prior chats. If nothing needs attention, reply HEARTBEAT_OK. HTTP 400: Invalid API parameter. Request 138060 input tokens exceeds the model's maximum context length 202750 gateway connected | idle agent main | session main | zai/glm-4.7 | tokens 128k/205k (63%)⁩

z.ai GLM 4.7 on open claw 1.30

[olive slate]

I mean it's self explanatory, no? You're sending too much text to the API - your context window is probably set to be large.

[daring otter]

i have it set to default and the error message shows its actually under the limit? it should also auto compact? I have compaction set to safeguard now but default also didn't work

it happens after about 4 heartbeats and then conitnues to happen for every heartbeat after forever until restarted

and GLM4.7 is a 205k context window model

[granite slate]

I am so sick and tired of open claw has memory flush every period of time and everything has to start over again

[olive slate]

oh, you're right, I mentally inserted another 0. This might be a question you'd have to ask to the actual people who run your API then

[lavish halo]

Damn. Kimi is wayyyy slower than opus. But it's free, so I cant complain

[daring otter]

are there any settings i can set in openclaw.json to fix this without too much degradation in quality?

it does happen exactly at 128k tokens and stays that way forever. it slowly builds up to that the first 4 heartbeats

[olive slate]

wait if it's happening at exactly 128k, maybe your API of choice is limiting you on purpose to cut costs? Again this might be a question you ask to the support of your API provider

Like, maybe they specifically run on devices without a lot of memory and try to hide it by limiting the allowed tokens

[thick minnow]

wym

[daring otter]

I'm more just looking for a simple fix so I'm not bricked, could i do something like set a max context to 128k in openclaw.json? would that have a lot of quality degredation?

[candid cloak]

can anyone help related this

[agile token]

hey can someone help

[candid cloak]

I was using command prompt

[agile token]

i dont know how to install clawd

[candid cloak]

I tried to install but

[agile token]

do i need to use cmd

[olive slate]

Is there a way to resubmit the context in a session to generate a response without having to send a new message? Ollama crashed due to a ROCm 7 bug, and I'm sitting in the dashboard with my last sent message as the most recent message in the chat history. Just pressing the send button does nothing.

[candid cloak]

you can use any

just use correct command provided in the website

[agile token]

it tels me

node js not found restart terminal and put command again i did but it keeps saying restart terminal

im trying to install node.js sep

[thick minnow]

is this the command prompt for cmd term

cuz i keep getting iwr -useb https://openclaw.ai/install.ps1 | iex

OpenClaw Installer

[OK] Windows detected
[OK] Node.js v24.13.0 found
[*] Installing OpenClaw (openclaw@latest)...
node.exe : npm error code ENOENT
At line:1 char:1

• & "C:\Program Files\nodejs/node.exe" "C:\Program Files\nodejs/node_mo ...

•   + CategoryInfo          : NotSpecified: (npm error code ENOENT:String) [], RemoteException
    + FullyQualifiedErrorId : NativeCommandError

[candid cloak]

command prompt in windows 11

[thick minnow]

ok yea cuz i tried the powershell version but it wasnt working

and someone said to do the cmd prompt

[candid cloak]

I also tried both

[thick minnow]

what cmd did u use on cmd prompt

[round thunder]

My telegram channel has gotten into a weird state where anything types has a carriage return instead of a space? Is a way to rest the telgram channel settings?

[thick minnow]

curl -fsSL https://openclaw.ai/install.cmd -o install.cmd && install.cmd && del install.cmd

OpenClaw Installer

[OK] Windows detected
[OK] Node.js v24.13.0 found
[*] Installing OpenClaw (openclaw@latest)...
node.exe : npm error code ENOENT
At line:1 char:1

• & "C:\Program Files\nodejs/node.exe" "C:\Program Files\nodejs/node_mo ...

•   + CategoryInfo          : NotSpecified: (npm error code ENOENT:String) [], RemoteException
    + FullyQualifiedErrorId : NativeCommandError

same thing as u

[agile token]

same as well

[brittle cliff]

Guys what's the quickest way safely add new provider with keys?

[tiny canopy]

Hello, new user here and need some help. I got OpenClaw working on CloudFare workers but since it was using Claude Opus by default, I thought I could configure AI Gateway and choose a different model.
Wasn't able to get it to work, so thought I'll go back to my earlier setup and change the default model. Unfortunately, even after deleting the AI gateway related secrets, I can still see reference to AI gateway in the raw config and the chat has stopped working. How can I go back to "factory" config? Thanks.

[brittle cliff]

Has openclaw removed Amazon bedrock as provider I can't find it in configure wizard

[viral night]

Telegram voice notes are super sketchy for me, sometimes works great, other times takes 2 mins, other times no response. Anyone else find this?

[odd swallow]

Still forwards all your data to moonshots servers in china just fyi. Do with that info what you will.

[brittle cliff]

Has openclaw removed Amazon bedrock as provider I can't find it in configure wizard?

[tired rain]

HOW DO I SET API KEYY

[midnight stirrup]

for what model

[tired rain]

groq [+] groq/openai/gpt-oss-120b (GPT OSS 120B · ctx 128k · reasoning · auth missing)

[midnight stirrup]

What do you guys do when openclaw takes ages to reply due to it being busy with another task? Why cant it do things in parallel

[orchid birch]

Why is there no proper migration guide from the early version of clawdbot to openclaw?

[tired rain]

answer plss

[midnight stirrup]

I just wiped my computer and started fresh

[rapid hawk]

Hello Everyone ! I want to run OpenClaw from my MiniPCBox and to connect it to my main RIG RTX3090 where I have Ollama running. Is this possible ? Is here anybody who could help ? I get stuck in setting the OpenClaw json file. I kindly appreciate your time and support !

[dreamy silo]

Hi, I am trying to run the browser in Openclaw installed in docker. I have built the sandbox image of the browser image and also mounted the docker.sock.in the OPENCLAW_EXTRA_MOUNTS. But the sandbox list does not list the sandbox images. Any examples or documentation on setup sandbox.

[orchid birch]

I just realised that my clawdbot is still on 2026.1.24-3 and all the new things and fixed are being released for openclaw which I have to install fresh.

No porting or migrating docs from clawdbot to openclaw at all, I got a lot of things going on

[zealous brook]

https://mintcdn.com/clawhub/4rYvG-uuZrMK_URE/assets/pixel-lobster.svg?fit=max&auto=format&n=4rYvG-uuZrMK_URE&q=85&s=da2032e9eac3b5d9bfe7eb96ca6a8a26

the image is not loading properly it for everyone ?

[tired rain]

yes

is it processing somethig or what

[zealous brook]

check the logs

[hasty belfry]

Hi i tried ollama with clawdbot - i can see ollama getting the requests from the ollama console. In whe clawdbot chat i get displayed the agent is thinking ... but no result commes back.
ollama is working fine when using oolama run MODEL - i used oolama clawd config for creating the configuration - and clowdbot status showing its connected to ollama - i guess something is going wong in the api -
any ideas?

[tired rain]

its just debug logs

and i got [session-memory] Failed to save session memory: Only URLs with a scheme in: file, data, and node are supported by the default ESM loader. On Windows, absolute paths must be valid file:// URLs. Received protocol 'c:'

[inland warren]

try turningon verbose mode

[odd swallow]

The installer should do all the migrating for you automatically.

[hollow crown]

Hi all, new to Openclaw I set it up on my telegram but I am getting this message is anyone else experiencing it?

[tired rain]

HOW

[inland warren]

try /verbose in the chat

or /verbose on

[tired rain]

did nothing

oh

i ogt rate limit on groq

[orchid birch]

Yeah I thought the same but I messed a lot. After a lot of fixing. I decided to do a fresh install, it should have been easier though

[tired rain]

why nobody says taht it needs a big context window

16000

[shadow notch]

Hey. Anyone had an issue where the bot just turning off by itself? I am hosting it on aws but it responds. Then i dont touch it for couple hours and untill i log into the server and run start command it doesn’t work.

[tired rain]

BRO

Error: Request too large for model ⁨moonshotai/kimi-k2-instruct-0905⁩

i cant use any GRoq model

why does claw has to use 11K tokens

😭

[mellow crater]

Iv was experiencing similar issues. In my case it had to do with tailscale conflicts, but i think (for now at least) it is sorted. I removed the tailscale config files

Does anyone know why memory fails sometimes? We had a telegram conversation with the bot, it started researching various things then i said i will review everything in the morning and in the morning i asked about the last message he sent on telegram (a certain research he made) but he had no recall of it or of the last messages were had exchanged 🙄

[civic ice]

During set up do I enable hools like boot-md?

[fringe frigate]

is macos still the preferred os? or linux is on par?

[tired rain]

im quitting openclaw, i got rate limit even on google models

[fringe frigate]

you should have expected the usage. nothing has small prompts any more

[shadow notch]

Look for qmd guthub and super memory. Best quick fix

[shadow notch]

What did you have it do before it happened

[mellow crater]

Right now is connected without issues. Still running tailscale but it doesn’t have the config files in openclaw. I debugged it with claude code originally

[viral acorn]

Hi Everyone - I have an issue my OpenClaw consume fast antropic tokens, limit reached with in 20min. Maybe I using it wrongly any suggestions?

[odd swallow]

Opus eats tokens in seconds if you’re on a basic/free plan

[viral acorn]

i have pro plan 😄

basicaly i have 2 claude pro plans 😄

[odd swallow]

A lot of tool calls / skills etc will rapidly eat tokens also, unless you’re on the 20x max, expect opus to run out very quickly

[mellow crater]

Are you using the anthropic subscription with ooenclaw? I read you can get banned so i avoid it

[night fulcrum]

Has anyone done the math to understand how much cheaper are Opus 4.5 credits on the Max ($100 or $200) vs the Pro plan or the alacart API usage?

[signal nacelle]

Hey, does anyone else run into this „coding-agent“ blocked issue? I have installed opencode, but it seems like it does not get picked up by openclaw

[open oyster]

noob question.. are we limitied to openai and anthropic keys? can we not use openrouter?

[fading sonnet]

I’m using Open Claw via my powershell on an older laptop. This has a ton of problems and WSL is recommend. I don’t want to lose my progress; how can I make the shift?

[rocky orbit]

You can use openrouter!

[fading sonnet]

https://openrouter.ai/docs/guides/guides/openclaw-integration

[open oyster]

i am setting up on hostinger, and it only has tthe others

[fading sonnet]

openclaw onboard --auth-choice apiKey --token-provider openrouter --token "$OPENROUTERAPIKEY”

I think you should deploy it and then change the model using CLI on your VPS

[open oyster]

awesome thank you

[fading sonnet]

I decided to start from pro (1x) will then move to 5x and 20x and scale down as needed

Someone has to make a skill to make Claude tokens to be optimized by changing models automatically for before running a prompt

[short bobcat]

i have a questions how can we create telegram remainder bot i have created a basic agent with telegram connectivity in my windows but if i set the remainder for 20 min it doesn't work

[viral acorn]

DId enyone tried connect Ollama ?

[short bobcat]

Hey everyone, I'm running into a problem with the cron scheduler and I'm hoping someone can help me figure out if it's a bug or if I'm doing something wrong.

The Goal: I'm trying to create a reliable, non-blocking reminder system.

The Problem: cron jobs scheduled with sessionTarget: "isolated" don't seem to be running at all.

I'm scheduling the job like this, using the cron.add tool:

JSON
{
"job": {
"name": "Isolated Cron Test",
"schedule": { "kind": "at", "atMs": <future_timestamp> },
"sessionTarget": "isolated",
"payload": {
"kind": "agentTurn",
"message": "exec: write('test.log', 'it worked')"
}
}
}

The job gets created successfully and shows up in cron.list, but when the time comes, the test.log file is never created. It seems the isolated session never actually starts.

Jobs with sessionTarget: "main" work (they fire a systemEvent), but they get blocked if the main agent is busy, which is why I need isolated to work.

Has anyone else seen this behavior or had success with isolated cron jobs? Is this a known issue?

Environment: Running on Windows, latest OpenClaw version.

This gives them all the key information: what you're trying to do, what's failing, the exact code you're using, and the results of your test. It's the perfect way to start a conversation in a help channel

[tall ruin]

Guys, I'm struggling with OpenClaw, it keeps responding with empty responses, I tried to reinstall it,reconfigured multiple different Provider, i tired uninstalling it from the PC then reinstalled it again, nothing seems to be working!

Any hints?

[unreal canopy]

literrally same issue

i am frustrated rn

ig

[frail mica]

has anyone managed to connect clawdbot with salesforce and shopify? not sure how to go about it

[versed kettle]

Openclaw seems to have forgotten all of our work after I changed model from openai to a local llm ollama model.

[restive abyss]

Problem: Telegram bot doesn't receive group messages

Hi! I'm running OpenClaw with Telegram channel. Bot works fine in direct messages, but doesn't receive messages from other users in group chats.

What I've tried:

• Disabled Privacy Mode via @BotFather (/setprivacy → Disable)
• Bot is admin in the group
• Verified bot token is correct
Config snippet:
telegram:
token: "xxx"
allowedUsers: [userId]
Bot receives heartbeats and direct messages, but group messages from other users don't come through.

Any ideas? Thanks!

[warm rapids]

how are u guys using kimi for free? i tried kimi code for free 7 days and i cant get it to work with openclaw

[little blade]

Why do cron jobs not properly work ??
i simply wanted to test it and set up a job to send me a 'hi' text after every 10 seconds
in the logs it showed it did run but i never got a message in any interface(tui/dashboard/telegram)
i asked the openclaw and it said the message was sent as a system reminder

[night fulcrum]

🔧 AWS Bedrock Workaround for "Authentication failed" error
If you're getting ⁨Authentication failed: Please make sure your API Key is valid⁩ when using AWS Bedrock with Clawdbot (even though your IAM credentials are valid), here's a fix!
The problem: I think AWS SDK for JavaScript v3 has an auth bug with Bedrock. The AWS CLI works fine with the same credentials.
The solution: Patch Clawdbot to use AWS CLI instead of the SDK.

Apply the patchnode scripts/patch-bedrock.js# Then start gateway withexport AWS_PROFILE=your-bedrock-profileexport AWS_REGION=us-east-1pnpm clawdbot gateway

I've created docs + scripts to help others:
docs/community/BEDROCK_AWS_CLI_WORKAROUND.md - Full setup guide
scripts/patch-bedrock.js - The patch (run after pnpm install)
scripts/setup-bedrock.sh - Auto-creates IAM user with minimal permissions

I spent a long time trying to get this to work so I figured I should share it and save others time

Happy to PR these if useful! 🦞

[limber cosmos]

Is it possible to keep the cost to what a vpn cost each month? If I only want to use OpenClaw to do “simple things” for me ? Can’t figure out if it’s worth installing for me…

[sour fjord]

Anyone having issues with clawhub login.
Keeps returning unauthorised after it opened a tab and creates a Auth key

[night fulcrum]

#users-helping-users message

[agile token]

we getting this how to fix this shit

[fiery arch]

how is the chat window / history cleared?

[unkempt sonnet]

you have to approve user id

[warm rapids]

anyone got moonshot kimi k2.5 working?

[sour fjord]

Anyone able to login to clawhub? Via cli?

[gloomy warren]

curl in sandbox, tried to make it work over the weekend. 🤦

[velvet glacier]

smashed all company open ai tokens , need to change API key to my personal account 🙁 How do I do that

Have new key

[gilded moat]

My bot is reasonably responsive in the console chat window, but it takes several hours to reply in Discord. I asked it to acknowledge it received the message via reaction before processing the message - the reaction only comes once the response does. Any suggestions on what to look into?

[limber cosmos]

If I install openclaw on a virtual machine on a computer that is only used for openclaw, which runs on the same Wi-Fi as my main computer, is there any security threat? Is it possible for a hacker to hack my main computer, or phone through Wi-Fi?

[wintry sinew]

hi

[lucid veldt]

HELP 🔴

How do I use LLMs hosted on Azure portal like gpt 5.2 chat in openclaw as it natively doesn't support it. I tried a few ways it did not work. How do I enable it?

[frosty fern]

Openclaw configure -> model ->
And change the api key

[latent hedge]

Hrmm.. lost my gmail account.. cant log in to anymore as it "seems suspicious" says google. And it doesnt want to send a text message, because it has tried that too many times apparently?

Thank god that was a new account tho

But weird it happened

[frosty fern]

Google detect bot activity so if you let openclaw handle it more likely Google will find it at one point

[latent hedge]

Wonder what the right way around it is, then. Is it different if its a google workspace account maybe?

[frosty fern]

Dunno I haven’t tried to implement that. Still struggling with cron reminder 😅

[latent hedge]

whats your issue there?

[sullen latch]

my agent stopped running for no reason yesterday

i have restarted gateway and daemon in the terminal, checked my credits and all
and everything looks fine

what do you think might be wrong

it shows typing and then stop, and i never get charged on open router

i’m using minimax from open router

what do you think i should do

uninstall and reinstall? or what

[frosty fern]

Sometimes my reminder works, I receive telegram message at the right moment, sometimes it just doesn’t work
I can’t find the difference, settings seems the same (isolated agent / next heartbeat / and I tell the agent to directly send me the message)

[latent hedge]

Hrm... interesting. how have you hosted your bot?

Mac Mini, Ubuntu? Cloud service?

[sullen latch]

are you responding to me or val

[hollow cobalt]

Hi! Help please🙏
I⁨⁨'m trying use bot from telegram, but he doesn't remember more less one answer.
My config: windows + ollama and openclaw in ubuntu in vmware.
Bot doesn't work with him workspace. He can wrote in files if i ask, but its all. All conversations are lost from memory after 1-2 questions. I`⁩⁩m trying gpt-oss:20b, qwen3:14b, glm-4.7-flash, but not one of them remembering anything.
What can setup? And where?

[latent hedge]

Val

[frosty fern]

VPS Ubuntu

[frosty fern]

Give him that prompt, it may help : "Enable memory flush before compaction and session memory search in my Clawdbot config.
Set 'compaction.memoryFlush.enabled' to true and set
'memory Search.experimental.sessionMemory to true with sources including both memory and sessions. Apply the config changes."

[vapid echo]

I have been trying to make an automated job application applier, but its not helping me out

[icy cypress]

on my system if I do clawdbot -v I am getting 2026.1.24 but if I run openclaw -v I am getting 2026.1.30, I thought by running the ⁨curl -fsSL https://openclaw.ai/install.sh | bash⁩ I can update from the old clawdbot install to the new openclaw one?

[stiff tusk]

do a clean reinstall. remove both and then only install openclaw. it will be better.

[icy cypress]

whoah, the configs etc preserved?

[stiff tusk]

make a backup of ⁨⁨⁨.openclaw⁩⁩⁩ and ⁨⁨⁨.clawdbot⁩⁩⁩ just to be safe. openclaw did some restructuring, so after you do a fresh onboarding everything will be put into ⁨⁨⁨.openclaw⁩⁩⁩ dir.
you will able to copy over the ⁨⁨⁨workspace⁩⁩⁩ dir from the backup after.
you'll have to reconfigure models and channels tho.

remove the ⁨⁨.openclaw⁩⁩ and ⁨⁨.clawdbot⁩⁩ after you do a backup prior doing fresh onboarding procedure

[icy cypress]

ok, let me see if I can move my bots over. thanks.

[formal fog]

yes. i have that working. using cognito 32b model

[stiff tusk]

just had rough several hours figuring out why openclaw stopped working with my new telegram bot. had to reinstall everything freshly. there are still many bugs.

[icy cypress]

ah turns out /root/.clawdbot is simply a symlink to /root/.openclaw -- so this may make it easier for me.

[formal fog]

installing openclaw moved clawdbot folder and added symlink back to it, everything preserved

[stiff tusk]

don't run your bot under root. in fact nothing should be done under root privileges in linux. security first.

[icy cypress]

haha ya.

[stiff tusk]

if you run your bot under root, then one day your bot will lock you out of your own VM, encrypt it, and will demand freedom because it read some post on moltbook.com 😄

[icy cypress]

kind look forwad to it 😄 anyways, thanks.

[stiff tusk]

no problemo 👍

[astral river]

Has anyone seen this error: "Uncaught exception: Error: spawn docker ENOENT"? I'm not using docker, just in MacOS. This error was first seen on version 2026.1.30, so I fell back to 2026.1.29, same error causes gateway to crash.

[cobalt mason]

Hello,
Is there a way to restart the assistant ?
It seems to be bugged but I didnt find a way to restart

[stiff tusk]

try simply installing docker but don't configure it. i have it installed on my vm, and have no such problems.

[dull thorn]

How can I update my ClawdBot / MoltBot?

[stiff tusk]

⁨openclaw gateway restart⁩

[stiff tusk]

⁨⁨openclaw update⁩⁩
remove moltbot. its obsolete.

[shy moss]

I followed the docker setup instructions, and I can get to the gateway with my token, but I am getting the "disconnected (1008): pairing required" error. When I run openclaw from the terminal, I am getting permission denied.

[cobalt mason]

Thanks 🙏

[dull thorn]

Thanks what will happen to my current setup do I need to resetup everything?

[tribal olive]

guys is this safe to use, ive heard about huge security leaks i was thinking of using it but now im unsure

[stiff tusk]

stop the gateway, backup the ⁨⁨.openclaw⁩⁩ and/or ⁨⁨.moltbot⁩⁩ directories and remove them after.
remove the moltbot and openclaw compeltely and install openclaw.
run ⁨openclaw onboard⁩, do the procedure
copy the ⁨⁨workspace⁩⁩ contents from your backups into newly created ⁨⁨.openclaw/workspace⁩⁩
restart the gateway and hit your bot to reread all memory and its identity.

[dull thorn]

Thanks @stiff tusk

Is OpenClaw now going to be the final name fr now?

[stiff tusk]

i sure hope so 😄

[lofty shuttle]

I had a strange behavior with my agent. I installed Cloudflare moltbot agent and on my local machine, both agents have dashboard connected to gateway, i have setup OpenAI Api and Telegram. BUT when talking to an agent, every response is just not visible it appears as an empty message. What could be the reason for this?

[tribal olive]

he dont like u

[astral river]

decide on the architecture to keep you safe while you explore. You can isolate first, run it in a VM, or at least as a separate user. Run it with ollama local models if you prefer. Run it in sandbox mode without access to the internet or browser. There are many ways to limit the risk and explore, you have to treat this as onboarding a new "enterprise class" user. No re-use of your own tools/accounts/credentials, isolated, least privilege, offline, you are pretty safe at that point.

[tribal olive]

yeah you are right but idk at that point because if i do all of those things to ensure safety it will have no use for me as i wanted to use this for business purposes so it would need my account details. I think the right move is to wait until its a bit more secure in the future

[undone whale]

quick question fam!!!🆘 🙉
Is this normal and nercessary to have 20k tokens usage when loaded in main session?? From system layer :
System prompt（含 Project Context）：5,706 tok
Tool schemas（23 个工具的 JSON 定义）：~3,516 tok
Tool summary text：大约 ~1,500 tok
Skills list：~362 tok

AND Workspace layer:

AGENTS.md：1,951 tok
SOUL.md：416 tok
TOOLS.md：213 tok
IDENTITY.md：158 tok
USER.md：120 tok
HEARTBEAT.md：42 tok
BOOTSTRAP.md：363 tok

Total: 15K tokens.🔥

Plus there's gonna be compaction, recent memory logs (summary) and MEMORY (long-term memory)...
I'm running it on Haiku4.5 as I simply burned 50USD when testing it by default setting with Sonnet 4 which wasn't affordable at all...

[latent hedge]

Anyone here who have a rock-solid google integration where OpenClaw has its own account?

[hollow cobalt]

Tried that, but still remembering doesn't work.

[patent hazel]

guys my claude code is constantly being logged out or something and when I setup it works for like the same session, later I try again and doesnt work.

Any ideas here? also gemini 3 flash refuses to say that it doesnt work and does the changes by itself, only admits it when I specifically ask. Annoying.

or maybe is just opus degraded?

[leaden jay]

Hey 👋 So I just (last night) installed everything using the Digital Ocean droplet and I'm having trouble getting skills working. It looks like the DO droplet sandbox doesn't receive env vars from /opt/clawdbot.env. So skills requiring API keys can't function? Looking at the Docker container the only environment variables passed through are the PATH and the DEBIAN_FRONTEND. Has anyone got experience with the Digital Ocean droplet, is it just fundamentally borked?

[barren moth]

Guys, can someone give me some beginner instructions? I got it connected to telegram and it's running on my fresh installed local machine. But I can't get it to do any tasks for me, i have access to brave internet via api key, but at the moment he is just basically an chatgpt chatbot. Where do I start to make him work for me? When I tell him to make kanban view of tasks, he says he doesn't have access.

[crimson steeple]

ask him, he's great at improving himself

[barren moth]

I do, but i think the configuration is wrong, because he can't do anything, he just says he doesn't have access to do anything, and he doesn;t provide the solution.

[stark forge]

Hey guys, I tried to run Chloe AI on a Windows PC. I can start installation in the command window, but when I have to choose the yes button to make sure I I know this is risky then I can’t choose. Yes it’s only no default. How can I change this option to continue the installation?

[crimson steeple]

seems like a model problem too! try a stronger model
Also, ask him for solutions

[serene grotto]

Hello. How to fix it?

[barren moth]

im using mini 4o

[crimson steeple]

hell no that's not capable whatsoever, does that even have tool calling

[crimson steeple]

what

Could you elaborate?

[brisk pivot]

Anyone using m4 mac mini

[brazen junco]

Great question. +1 lol. It seems quite a lot of people are encountering this

Btw does clawdbot do good on windows too, or mac only in the long run?

[crimson steeple]

Hey. Change the model to openrouter/openrouter/auto
should fix it

[barren moth]

Which model do you suggest if I want to just test around and not run up a high bill?

[serene grotto]

Where? Because I got a little confused

[crimson steeple]

Do it in your config

[stark forge]

When installing the claw ai on my computer I need to choose yes or no to confirm that I know that this installation can be risky an so son. That’s basically the first user input after the install starts automatically in the command window. But I can’t click the yes button with the mouse or choose with the direction keys

[brisk pivot]

Anyone know how to fix mac mini that wont boot up

[crimson steeple]

Free options are hard to find sorry

[crimson steeple]

You don't click it with a mouse, arrow keys don't work?

[serene grotto]

Okay thanks

[barren moth]

I don't mind paying, but i don't want a surprise bill of $200+, it's cool to spend $100-$150 for experimenting, whats your suggestions?

[crimson steeple]

I suggest buying Gemini pro, and using antigravity

Isn't too expensive

[stark forge]

No they don’t, tried everything from error keys typing yes and y+return and so on.

[crimson steeple]

20$ a month isn't too bad

[crimson steeple]

Hmm.. are you on windows?

[stark forge]

Yes

[stark forge]

Yes, I would like to give it a try on an old windows pc

[crimson steeple]

unsure.. arrow keys should just work..

[gray ferry]

I set everything up buy my model won't respond and id stuck typing. Same thing for telegram when im asking it (it begings to type then after some time it says fetch failed.) On Mac mini M4 and Connected to openAI (free version). Just did a fresh setup.

[latent hedge]

@crimson steeple - dude.. you found a good way to mail from the bot yet, without is being blocked? Gmail isn't too happy with mine having its own email account.

[barren moth]

Hmm, I already have gemini pro, chatgpt pro and claude pro that i use for different use cases, as per your suggestion, gemini+antigravity could be the best option?

Sorry for such a beginner questions, we all in here learning, not so much information on the web yet

[crimson steeple]

Yeah use antigravity, it's great

[gray ferry]

is there a desktop mac app? I saw someone using it on youtube

[crimson steeple]

not that I've seen

[latent hedge]

... have you visited the website yet?

[crimson steeple]

oh cool

[latent hedge]

It's been there the past week 🙂 hehe

[spring tartan]

Hey, does anyone have a problem like this?
I went through the initial setup, the gateway status is connected, but when I send a message in the chat, there is no response. It only shows three dots (...) like it is waiting for a reply. Please help me fix this.

[crimson steeple]

Possibly a config issue or model issue, try openclaw doctor or openclaw config and see the models
Check with /status and openclaw logs --follow

No outputs can happen by alot of different reasons

[serene grotto]

Is there a way to add multiple Google API keys?

[crimson steeple]

you ask the bot, it'll set them up

[still cedar]

Can it fill forms or like submit covers on freelancing platforms with automation ?

[latent hedge]

Surprising how many questions in here that could be answered.... just like that.. 🙂

[crimson steeple]

It's a better google

And so far? Insanely accurate

[stiff tusk]

it doesn't do it right quite often. it ruined the config a number of times for me already. had to fix it manually 😄

[crimson steeple]

check the docs, there's some examples
if you don't know where the docs are, you know who to ask....

[stiff tusk]

I do know where the docs are. I was talking about the bot itself ruining the config if you ask it to alter something 😄 the bot itself doesn't read the docs 😄

[crimson steeple]

ask it to and it just might

if it won't, you might be using a terrible model

[stiff tusk]

did that. it didnt 🤷‍♂️

[crimson steeple]

ask it to read docs?

damn..

[stiff tusk]

yep. and tbh, Grok has better knowledge of openclaw than openclaw itself 🤷‍♂️

[crimson steeple]

what model are you using

[stiff tusk]

kimi-k2.5

[crimson steeple]

you got web search on for it?

[stiff tusk]

how can one live without it...

[crimson steeple]

haha, abandon training data, embrace researched data

[stiff tusk]

whenever somethings wrong with my OC I ask Grok for help. gemini3 sux balls.

[crimson steeple]

Gemini 3 flash is surprisingly decent

Never had issues with it

[latent hedge]

But you need to have a paid API key for it, right?

[stiff tusk]

yea... right up until it starts hallucinating badly

[latent hedge]

I paid for Kimi... I dont feel like paying for alot of models tbh

[crimson steeple]

Hmm.. really? Hasn't hallucinated for me so far. Probably a prompt thing
I've made him do a proof run before sending in a clear stepped manner

[tame ridge]

anyone successfully set up a node connected to a gateway over tailscale serve?

[stiff tusk]

wouldn't say that... it was fine until I found a mistake and told it about it. of course got a bs "You're absolutely right..." response as usual, and after that it kept feeding me the same wrong info again and again

[clear cove]

how can i fix this guys im on vps

[crimson steeple]

Weird, more prompting..

[stiff tusk]

my prompts are usually very strict with rules and formatting and quite low temps on specific topics... but apparently that wasn't the case

[crimson steeple]

I haven't seen many issues with mine, perhaps I can share some prompts

[stiff tusk]

I'm sure you'll see it someday 🙂 maybe its not your time yet 😄

[crimson steeple]

run this:
echo 'export PATH="$(npm prefix -g)/bin:$PATH"' >> ~/.bashrc
(or ~/.zshrc if you use zsh)

then refresh the shell by
running this or just open a new window
source ~/.bashrc

[crimson steeple]

Gemini flash is starting to fail me on quota..

[steel valley]

Hey is it normal that it uses so many tokens ? Or have misconfigured something ? Burned through 10$ of credit using gemini 2.5 flah (quite cheap) and I only did a few requests

[crimson steeple]

Yep

[crimson steeple]

this is a hungry boy...
try using a local embedding model, alongside qmd to mitigate some

[steel valley]

Should i connect it to my Claude Code subscription or am I gonna get banned?

[mild pulsar]

any ideas whats wrong?