users-helping-users

[crimson steeple]

gonna go through an onboard reset

full reset your config, I think that should regenerate your dashboard token? also reinstall the gateway?

[slim drift]

how to do that?

i already reinstall it

and it still having a meltdown

[crimson steeple]

clawdbot onboard
Then you click manual

And do reset whenever it asks

[slim drift]

it has the token but idk why it not working

unless i have to copy and paste it

bro idk

💔

ion think i should send that but you get the point

yeah it still not working

imma just reset everything again

[viral spoke]

这个属于配置好了吗？然后要做什么

[dusky perch]

I have a safety question - if I use openai oauth, would this in any shape or form open the doors for a malicious actor with access to my molty visibility on chat history?

[vague berry]

i got it to work after a reset where i setup claude api...

[crimson steeple]

Best way is to do clean reinstalls haha

[unreal steppe]

Claude keeps reverting to use my Oauth login instead of api, any way to force the change?

[junior ingot]

are the docs down for everyone else as well ?

https://docs.clawd.bot/

omfg we changed names again ??

[slim drift]

through clawdbot onboard?

dude this is so confusing

[crimson steeple]

completely uninstall the bot

remove the folders within your appdata

[civic harness]

I've installed openclaw with npm and I have this message
Control UI assets not found. Build them with pnpm ui:build (auto-installs UI deps), or run pnpm ui:dev during development.

Anyone know how to deal with that?

I don't know where I can find the repo on my mac

[full aurora]

I've reboot pc and it fixed by itself

[slate ginkgo]

Bros, I am cooked. I was trying to spin it up on a VM.

I can't get ollama to spin up a local llama model and have it recognized by openclaw. What are you guys using?

[civic harness]

will try that

[slim drift]

the uninstall cli doesnt even work ???

im reinstalling it again

[crimson steeple]

clawdbot uninstall --all --yes --non-interactive

[tall delta]

how do we update from clawdbot to openclaw cli?

[slim drift]

i guess that worked

[meager mica]

I just asked clawdbot to set up llama for me

[crimson steeple]

Best way to migrate is to literally ask the bot to do it for you
They know the directories better

[narrow gorge]

I was trying to setup the Anti-Gravity Auth - I got this error when things should have finally worked.

"This version of anti-gravity is no longer supported please update to the latest version"

Pretty sure my anti-gravity was updated not less than 3 days ago

[slim drift]

are you trying to get that free google method lol

[narrow gorge]

Yup

Or just want to use Gemini 3 pro

[civic harness]

how do I open the web UI with the token?

[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",

[narrow gorge]

Thanks. let's try that

[half cypress]

exetnsions are out of syn... and taking huge space on my C Drive... Even thought i had connect to D drive by smd link

[crimson steeple]

no idea on that one sorry

[slim drift]

@crimson steeple do you know what causign this?

it shouldnt take that long and it is going through openrouter api

[crimson steeple]

This could be a quota issue, provider issue, etc. Sometimes if a provider truly returned nothing, it'll go on a loop, you'll have to do /stop to abort
clawdbot logs --follow can help you pin down what happened

[slim drift]

oh.

i thought it fix it?

😭

[little blade]

I am totally new to it, what's the problem

[crimson steeple]

Probably something about gateway installations, try to reinstall your gateway, try onboard again, complete reset

[slim drift]

reinstalling it works

[little blade]

It's about this I guess, I need a token

[slim drift]

woah where you found that?

[crimson steeple]

You're going to need to note down your gateway token when it first generates it

[willow rover]

I have installed and reinstalled openclaw (even created a new user) on a linux VM. I keep getting errors about duplicate matrix plugins. Manually deleting and reinstalling (I'm not completely sure I know what I'm doing there) doesn't fix it.

Config warnings:\n- plugins.entries.matrix: plugin matrix: duplicate plugin id detected; later plugin may be overridden (/home/openclaw/.npm-global/lib/node_modules/openclaw/extensions/matrix/index.ts)

I got it to briefly work by asking the bot to fix it, but when I exited the TUI it stopped working.

[little blade]

Overview > Gateway Access

[zealous brook]

i deployed my clawdbot on EC2. It's working as expected, but the main control center is not accessible via public address i tried changeing binding to LAN and configured the SG groups on my EC2. only workaroud is SSH and port forwarding working for me

[willow rover]

I'm having that same issue. I think it's listening on localhost but not binding to all interfaces.

[zealous brook]

../.clawdbot/clawdbot.json

[willow rover]

There does appear to be a bind parameter in the json config, that defaults to loopback

[slate ginkgo]

you just installed clawdbot and didn't need any underlying model to do that? so on a fresh install, I can install clawdbot and run it to install a local model?

[zealous brook]

binding to lan must solve the problem but something is overriding

[lucid sedge]

Random Q: can multiple different whatsapp people message openclaw? and can it individually recognise them?

[slim drift]

is everything on that json?

like for the dashboard and cli

cus imma just do that atp

[zealous brook]

No one single gateway multiple users can access it but openclaw thinks all the requests are asking from one person

[lucid sedge]

Thanks! Is it easy enough to run multiple?

[strong kelp]

hey guys, I was doing the quick setup wizard on AWS vps, when it asked me if I want to configure skills I just couldn't select 'yes', no button would work anymore.

so I refreshed the page and now it's like I just opened the AWS instance for the first time.

how can i go back to the setup wizard?

i already connected API keys and tg

[zealous brook]

yeah that json contains all the envs and configs at runtime it will pick things from it

[zealous brook]

do clawdbot configure

[slim drift]

does it auto restart when the save the file or do i need a command ot manually do it

[zealous brook]

clawdbot gateway restart , stop , start are the commands there

[little blade]

I put it, restarted but sadly chat is not responding

[strong kelp]

it says clawdbot command not found

[little blade]

Does the token change everytime I restart?

[woeful moat]

might be a dumb question but is it possible for openclaw to have access to my computer even if its hosted on oracle?

[little blade]

openclaw

[zealous brook]

yes

[woeful moat]

do I have to connect gateway or something?

[zealous brook]

yeah you need to install the clawdbot on the remote machine right give all the permissions that requried and you can ask the llm to work for you

[zealous brook]

Check if the clawdbot installation correctly done ?

[bronze basin]

Yo, can someone help me recover my moltbook account? I accidentally started a new session I think and my API key is now gone?

I need the account very badly

[slim drift]

i got it working

[zealous brook]

great !!

[strong kelp]

lmao this was it, thanks

[formal crow]

Hi, is there an easy way to update to openclaw from clawdbot. I asked it to update, but it didnt find any. Yet the name has now changed and I can see updates in the announcements

[sour berry]

i am not seeing responses from openclaw - any hypos on how to fix this?

[zealous brook]

click the brain icon to check the tool call responses

[obtuse sundial]

I'm trying to set up model-router but I'd like to use openrouter and swap between different models instead of using the ones offered in the setup anybody know if this can be done?

[zealous brook]

from that we can debug what is the eror

[radiant latch]

is it safe to use telegram with my phone & main number ? (or a second mobil number is prefered, as for whatsapp) ?

[slim drift]

now how do I give my bot a personality?

do anybody here know how

[zealous brook]

if you are the one who is using the bot then your number can be used but if there are more than one person is accessing the bot its better to have a seperate number for the bot

[north fog]

hey anyone have any luck on ollama set up

[zealous brook]

.../clawd/IDENTITY.md edit this file as you wish

[slim drift]

cool

but is it possible to prompt AI to change it own md?

i know it should

[zealous brook]

yes

you can ask the bot to selfupdate or mannually the identity.md file as well

[barren robin]

does anyone have a proxmox image all set up that i can get a copy of?

[slim drift]

im asking it to do it

[lean ridge]

anyone have any idea on how i can setup deepseek api key on openclaw?

[slim drift]

You're such a pro

[zealous brook]

i think repo structures are changed i need to update mine !

[little blade]

It still doesn't work for me for some reason

[jagged rock]

Is it okay to run it locally instead of using the Claude API?

[zealous brook]

click the brain icon on the top right and show what it is saying

[little blade]

Nothing it just turn it on / off something

[sour berry]

i did - it doesnt respond - nothing happens on click

[zealous brook]

even after clicking the brain icon nothing happen that means none of the tools calls are made so next step is to check is your llm api key is active and working ?

[slim drift]

check your logs

[dark galleon]

spent the last 30 minutes trying to correct the time. is there a skill or anything to help with this? the system time on the machine clawd is running on is correct but clawd wanted to think it was 06-27-2024 got the date corrected but clawd is still an hour off despite specific instructions to use system time.

[little blade]

No error, latest one here

⁨10:46:48 PM debug agent/embedded embedded run start: runId=RUN_ID_HERE sessionId=SESSION_ID_HERE provider=openai model=gpt-5.2-pro thinking=low messageChannel=webchat⁩

[glass swallow]

Hey, any tutorial on how to setup a local model? This is expensive to run on a cloud service

[little blade]

what happent to this now 😭

[cold condor]

That’s the same thing that’s happening to me

[little blade]

This is happening after setting open AI api key, before I wasn't using any model's api

[calm dawn]

I am trying to use Claude Code OR token mechanism with a Clowdbot Setup token. It was successful, but after some times it gets a 401 error.
I tried a couple of times, but couldn't find a solution.
What do you suggest?

[runic shore]

Hi Guys, new here. I want to get started on installing this but i wanted to ask a few questions if you can help me with....1) can you install this on a NAS like synology? 2) Once installed, can you migrate it to another PC or system?

[crystal swift]

why is docker install not on the front page? is it bad?

[crystal swift]

yes and yes

[thorny sleet]

Had that issue when switching from openai 5.2 to 5 mini which was not available, so probably just wrong model call

[runic shore]

amazing! can you please point me in the direction where i can find out how to install it on my NAS

[crystal swift]

dont know where docs are for it. i know its possible though.

[runic shore]

ok so if i install it on a laptop for now and then if i want to transfer it to a NAS, that is doable

just want to make sure i wont hit a RESET

[slim drift]

what?

[karmic shuttle]

Hi, ssomeone knows how to create a second bot in telegram in the same mac mini? :c

[ebon haven]

u figured it out yet? i have the same problem

[sour berry]

nope

replaced key to gemini key - still not working

[knotty solar]

Big thanks to @crimson steeple and @analog scroll , you did truly help me getting it all set up. Thanks for helping and it's awesome to have you in the community

[twilit spire]

Is there a simple way to migrate my clawdbot environment/naming to the newly named openclaw?

[long hatch]

i had this same question earlier. just ask your bot to migrate it for you. basically, you just run the install script (the new one on the site, using openclaw), then run doctor

[runic shore]

can anyone help me with the install/setup on windows? i seem to be hitting this error [npm error code ENOENT]

whats the best way to install on windows...

[knotty solar]

Sure, could you send logs?

[knotty solar]

curl

[hushed plinth]

guys if I connect my clawdbot using the antigravity oauth method so I could use my google pro subscription would google ban my account? is this againt their tos?

[long hatch]

using powershell. install node from the website, install git from the website, then run the installer from openclaw

[knotty solar]

Why not the one liner over cmd or powershell? Worked perfectly fine for me and was a lot easier. But my opinion ofc

[stone obsidian]

is it easy to migrate openclaw to a vps from ur laptop

[ebon haven]

i got it working, using anthropic-opus-4-5 now

[elder hazel]

I did exactly this and even after doctor I’m still on Clawdbot 1.23-1

[rocky hinge]

I tried to install openclaw and got this: C:\Users\rober>curl -fsSL https://openclaw.ai/install.cmd -o install.cmd && install.cmd && del install.cmd
At C:\Users\rober\AppData\Local\Temp\openclaw-install.ps1:1 char:1# OpenClaw Installer for WindowsThis script contains malicious content and has been blocked by your antivirus software.
+ CategoryInfo : ParserError: (:) [], ParentContainsErrorRecordException
+ FullyQualifiedErrorId : ScriptContainedMaliciousContent

[thick crypt]

i installed using these commands, but when i run the second one command it opens this browser tab but doesnt work as u can see in the picture. Can someone help me?

[kind hatch]

sessions memory never gets indexed i've lowered deltabytes threshold too and restarted gateway multiple times, any one know a fix?

[solemn crown]

Hello friends. I have been reading about open claw everywhere online and want to experiment with it but don't even know where to start. Anyone have any good reads on getting open claw up and running on my system?

[knotty solar]

Why isn't clawdhub working?

[marble socket]

safest way to install clawdbot?

[knotty solar]

Just start on openclaw.ai or docs.openclaw.ai

[runic shore]

thanks! this seems to work. i am now going through onboarding

[celest salmon]

Any skills to control replit?

Or has anyone successfully connected multiple instances together make them all act as a team? (Different “roles” per instance)

Willing to pay for help here etc.

[full berry]

how to use it with groq, I manually edit openclaw,json but don't know where to put token, and even after that, it gives,
Embedded agent failed before reply: Unknown model: openai/gpt-oss-120b

how am I supposed to add a model which isn't listed in defaults?

[jagged granite]

Anyone have config they can share for DeepSeekAI workikng as a custom model? I seem to be screwing up the syntax or something

[celest oracle]

Hi! I need help - used Clawd for a couple days and my chatgpt plus ran out of gas, need to wait 3 days now. What's the easiest way to add another model in the meantime?

[runic shore]

so when it asks for what model to install...whats usually the choice? i use chatgpt (free ver for now but considering paid)

[nimble loom]

So running local llm via ollama and some time-consuming tasks give me a timeout, where would I tweak this? or should this be ran as a sub-task or agent or something?

[untold holly]

Hey Clawed friends 🦞
I’m trying to understand the execution model used by the agents.

From the docs, it looks like the project is built around a pi-agent, but I’m not fully clear on how action execution is handled under the hood.

Is the agent following a ReAct / CodeAct-style loop? If so, where in the codebase can I find the implementation of that execution flow?

Any pointers would be much appreciated—thanks!

[sly fractal]

◇  Select a channel
│  Matrix (plugin)
│
◇  Install Matrix plugin?
│  Download from npm (@openclaw/matrix)
Downloading @openclaw/matrix…
Extracting /tmp/openclaw-npm-pack-E6NbHS/openclaw-matrix-2026.1.29.tgz…
Installing to /home/user/.openclaw/extensions/matrix…
Installing plugin dependencies…
│
│  Failed to install @openclaw/matrix: npm install failed:
Plugin install failed: npm install failed:

Any ideas why? No npm error message outputted

Also seeing this:

Config warnings:\n- plugins.entries.matrix: plugin matrix: duplicate plugin id detected; later plugin may be overridden (/home/user/.npm-global/lib/node_modules/openclaw/extensions/matrix/index.ts)

[signal dawn]

How can I update my Clawdbot to openclaw?

[clear timber]

I think it operates on an agent loop, llm sends tool call format, tool outputs status to llm, llm keeps going until it stops calling tools. so to answer your question: ReAct

[lethal terrace]

Hey, I've just setup my OpenClaw on my mac mini. The chat is suggesting I download the OpenClaw iOS app. I cannot find it. Can someone point me to where it is?

[clear timber]

you gotta build it from source i think

[slim drift]

why is it doing this?

[torpid furnace]

I'm running an llm locally through lmstudio. oss-20b works fine, but a 4bit glm-4.7-flash is struggling to use tools, and ends up looping trying to execute them. I'm using the openai-responses api, and stock jinji, but I can't for the life of me get it to cooperate

EDIT: looks like other people are struggling with tool calls too

[dense moon]

Hey y’all, I recently hopped on the clawdbot train now openclaw. I was wondering what techniques/ recommendations you had for keeping API costs down? Still pretty new to vibe coding and want to keep costs down. Thanks in advance!

[versed kettle]

Anyone got an answer to this?

[clear timber]

hey liv, try the new kimi 2.5 model on kimi code, its free for 7 days and it is 3x its own rate limit

[clear timber]

build it from source

[split sable]

Any tips on getting Tailscale to work? I cant for the life of me get this going. Proxy headers detected from untrusted address. Connection will not be treated as local. Configure gateway.trustedProxies to restore local client detection behind your proxy.

So I add the following under my gateway config. But its still refusing to connect the session.

"gateway": {
"mode": "local",
"trustedProxies": ["127.0.0.1/8", "::1/128"],
"port": 18789,
"bind": "loopback",
"auth": {
"mode": "token",
"token": "TOKENGOESHERE",
"allowTailscale": true
},
"tailscale": {
"mode": "serve",
"resetOnExit": false
}
}

[stark fjord]

Anyone had issues with installing on aws? running on a free t3, but install script just hangs. "npm install failed"

[odd veldt]

did you fix that?

[willow rover]

No luck so far

[runic shore]

im trying to set up OpenAI model. Just created my API key...which model do i select? Also do i have to have subscription for it to work?

[split sable]

Yes it helps if you have ChatGPT Plus or Pro. Use Codex oAuth for the auth and pick Codex 5.2 chat for the model

[small jackal]

did gemini blocked the requests ? after 3 request i only get no output ...seems the rate limirt with freee api key is reached

[runic shore]

will it work without a subscription?

[split sable]

free ChatGPT doesnt have access to Codex. YOur other option is to fund API tokens with OpenAI

[runic shore]

thanks! and lastly (im totally new to all this so sorry) can i go through onboarding with what you said and then get the subscription after?

[split sable]

no, as you dont have access to Codex to link 0Auth to

[untold holly]

Thank you for the answer. Do you know if the entire agent loop is handled by the pi-mono agent core or are there any changes to this procedure made by open claw?

[runic shore]

got it. getting the subscript now

[chilly musk]

Guys is clawhub working for yall

[knotty solar]

I've purchased Claude pro to use Claude Code as the API but already hit a brick wall after like 30 minutes. Are there other subscriptions that provide better value?

[hybrid basin]

Has anyone successfully configured this with a free model via openrouter?

[green sundial]

is there a good guide to getting this set up on a Windows 11 computer for somebody with very limited technical expertise? I tried the https://docs.openclaw.ai/ site but it seems to assume you know a lot more about computers than I do. I'm not even sure how to get it installed on my computer

[runic shore]

got my pro subsciption. now for default model i chose 5.2 chat latest correct?

[split sable]

yes, Codex 5.2 Chat

[round comet]

I've started getting a rate limit response from anthropic (according to openclaw) but I have plenty of rate limit remaining and my claude code / chat work fine individually still. Anyone else run into this? Have tried resetting the token a few times and removing all fallbacks to get to this point.

[runic shore]

i see 5.2 codex and 5.2 chat-latest

[split sable]

openai-codex/gpt-5.2-codex (GPT-5.2 Codex · ctx 266k · reasoning)

[warped sparrow]

Is it possible to use : arcee-ai/trinity-large-preview:free

[stiff tusk]

is it me or cron jobs still not working?

[split sable]

Also be aware that with Plus, you'll have lmitations on use with Codex, so dont go too wild.

[ebon flame]

Hi my huge friends, I'm having a hard time setting up the time zone correctly so that the bot picks it up right. Do you know if there’s any special setup required for this?

[runic shore]

i got pro

subscript

just for my understanding, whats the advantages of 5.2 codex versus 5.2 chat-latest?

[split sable]

Codex creates a container environment to run ChatGPT queries into, its original purpose is for code development and testing, but it can be used for other things. ChatGPT regular does not support API integration. You either need Codex with oAuth, or OpenAI API tokens

and the cost of OpenAI, API tokens can get pretty out of hand fast.

[runic shore]

ok so codex can still be used as a regular chatgpt

[split sable]

yes it will reply to you

[runic shore]

and forgive me for my noobness but i still dont understand the with oAuth part

lol

[split sable]

oAuth is token based authentication for your OpenAI account.

[kind hatch]

Dude my clawdbot has been working fine for this whole time now its just repeating messages wtf

[prime nimbus]

Any managed to build the virtual self for openclaw? I've got it designing an apartment to live in, and that terrible.. lol i mean it could be my prompts, but wondering if anyones got a good guide out there?

It struggles 😄 lol

[green sundial]

I keep getting this error message when I try to install OpenClaw:
⁨`[*] 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`⁩
  
  

Does anyone know what it means?

[woven jolt]

i might need some help. I'm trying to get openclaw to work on github codespace and i managed to get to the dashboard but now im getting error 1008, pairing required. I don't have any idea what that means so i'm kinda stuck

[runic shore]

so it says onboarding complete but is showing no output

said wake up, my friend

but nothing else

[fresh moon]

hello i keep getting this message and havent been able to use clawd : Gateway: not detected (gateway closed (1006 abnormal closure (no close frame)): no close │
│ reason)

any way for me to fix this

[supple knoll]

hey i am getting an error when starting up Problem:

• channels.webchat: unknown channel id: webchat

What is the CLI command to fix this

[runic shore]

had the same issue and used powershell

[warped sparrow]

Do you know where I can find a list of the whitelisted models?

[atomic sparrow]

LLM request rejected: messages.94.content.3: unexpected tool_use_id found in tool_result blocks: toolu_01NaEf2d1iy3RxvmeEQ6FRQE. Each tool_result block must have a corresponding tool_use block in the previous message.

anyone know how to solve this one?

[thick viper]

I can't get the local chat or the remote discord chat the ability to make cron jobs. Is this a bad idea?

Patch failed—schema validation error ("gateway.cron.capabilities" unknown; cron is top-level). Current config lacks cron section.

Propose: {"cron":{"enabled":true,"maxConcurrentRuns":4}} (enables cron jobs; adjust as needed).

Confirm? Or provide exact JSON patch/paths. New hash after schema fetch: use prior one if unchanged.

[runic shore]

So i just finished onboarding with open ai 5.2 codex model and it says wake up my friend but then it says no output...any guidance would be appreciated

[split sable]

check the logs "clawdbot logs"

[warped sparrow]

Do you know where I can find a list of the whitelisted models?

[runic shore]

anything i type it says no output

[spiral fog]

Hi, I'm a bit new. I just set up an agent on a Discord bot from the onboarding process and I was confused on what the next steps are to get it working on a server/in dms?

I got it set up with a model, linked it to the bot, and the dashboard opened but i get this as well (which I'm not sure is an issue or not)

[green sundial]

I'm still having issues trying to install OpenClaw despite running Powershell as an administrator:

[wind rover]

I'm also a Windows user, and I was under the impression that it just doesn't work in PowerShell and we need to use WSL or Docker?

[green sundial]

it's frustrating because I feel like I can't learn this type of stuff without already knowing how to do it beforehand.

[sullen knot]

anyone else having issue with kimi 2.5 when using deepgram or whisper TTS?

[runic shore]

what worked for me was installing git, jsnode and then running the command in powershell

[pearl solstice]

@analog scroll huge call on minimax

It’s doing tools so easily

[royal cedar]

same it worked for me after installing node and git

[cerulean solar]

Hello there, I'm not great with computers but I'm trying to install openclaw on windows and I keep getting this error message after giving permission to run the script in admin as in both cmd and powershell.

Node.exe : npm error code enoent
At line:1 char: 1

C:\program files\nodejs/node.exe

Any idea what i should do?

[royal cedar]

https://www.geeksforgeeks.org/node-js/how-to-fix-npm-err-code-enoent-error/

[cerulean solar]

Looks like there are a lot of people having the same problem

Thanks

[runic shore]

how do i change my model because its saying status disconnect gategway token missing

[fossil spruce]

@daring flare One sec.

Is that the only openrouter model you want to use for now?

[daring flare]

I would like to have them all available

[fossil spruce]

This is an example of what I have in mine.

[
  {
    "id": "moonshotai/kimi-k2.5",
    "name": "Kimi K2.5",
    "reasoning": false,
    "input": [
      "text"
    ],
    "cost": {
      "input": 0,
      "output": 0,
      "cacheRead": 0,
      "cacheWrite": 0
    },
    "contextWindow": 128000,
    "maxTokens": 8192
  },
  {
    "id": "moonshotai/kimi-k2",
    "name": "Kimi K2",
    "reasoning": false,
    "input": [
      "text"
    ],
    "cost": {
      "input": 0,
      "output": 0,
      "cacheRead": 0,
      "cacheWrite": 0
    },
    "contextWindow": 128000,
    "maxTokens": 8192
  },
  {
    "id": "moonshotai/kimi-k2-thinking",
    "name": "Kimi K2 Thinking",
    "reasoning": false,
    "input": [
      "text"
    ],
    "cost": {
      "input": 0,
      "output": 0,
      "cacheRead": 0,
      "cacheWrite": 0
    },
    "contextWindow": 128000,
    "maxTokens": 8192
  }
]

These are the 3 that I use right now. But let me add another one to the agent, and show you the error, and what happens when I add it here. One sec.

[daring flare]

what I don't get is which part of the config is this on 🤔

[analog scroll]

glad to help! each one teach one. no need to go broke to have a bot!

[fossil spruce]

models.providers.openrouter.models

Here, try this.

openclaw confit set models.providers.openrouter.models '[
  {
    "id": "moonshotai/kimi-k2.5",
    "name": "Kimi K2.5",
    "reasoning": false,
    "input": [
      "text"
    ],
    "cost": {
      "input": 0,
      "output": 0,
      "cacheRead": 0,
      "cacheWrite": 0
    },
    "contextWindow": 128000,
    "maxTokens": 8192
  },
  {
    "id": "moonshotai/kimi-k2",
    "name": "Kimi K2",
    "reasoning": false,
    "input": [
      "text"
    ],
    "cost": {
      "input": 0,
      "output": 0,
      "cacheRead": 0,
      "cacheWrite": 0
    },
    "contextWindow": 128000,
    "maxTokens": 8192
  },
  {
    "id": "moonshotai/kimi-k2-thinking",
    "name": "Kimi K2 Thinking",
    "reasoning": false,
    "input": [
      "text"
    ],
    "cost": {
      "input": 0,
      "output": 0,
      "cacheRead": 0,
      "cacheWrite": 0
    },
    "contextWindow": 128000,
    "maxTokens": 8192
  }
]
'

Copy this whole thing, including the first and last single quote.

And then, try to do /model moonshotai/kimi-k2

That will add it to your config file in the right place, restart your gateway before trying it.

@daring flare

[honest elm]

hey there, is there an upgrade path from clawdbot to openclaw? or do i have to reinstall it?

[fresh moon]

same question

[remote ingot]

same question too.

[leaden plover]

reinstall,,, remeber to backup

[honest elm]

how to backup?

and can the backup be used with the newly installed openclaw afterwards?

[runic shore]

so how do i generate gateway token...help section says ask the bot but my bot is not even initialized cause gateway token missing

[leaden plover]

ask the bot to do it

[runic shore]

whatever i type it says no output

it didnt 'hatch'

wonder if there is a way to uninstall and start a new install/onboarding?

[slender obsidian]

I just did this to update mine on a Debian VM:

curl -fsSL https://openclaw.bot/install.sh | bash -s -- --no-onboard
then run
openclaw doctor
openclaw doctor --fix

If you have a VM or VPS etc, just snapshot first in case something goes wrong

[abstract thistle]

is anyone having the bot forget to do things and / or processes despite it committing it to memory? Any ways to fix this?

[slender obsidian]

Enabled Opus 4.5, tell it your problem and ask it to do a overview of the situation. Have it check the openclaw docs related to memory and check to see if embedding and memory functions are correct. Ask it about compating and how to mitigate the issue.

[stable fable]

my macbook from time to time is glithing - screen blackout - come back after 2 seconds. anyone experiencing this ?

[slender obsidian]

er compacting even 🙂

[runic shore]

where do i type this

[abstract thistle]

I tried most of that but not worded your way, let me have another go, thanks!

Also, i notice the bot getting a lot of updates, do I need to do anything to update my bot or is it automatic?

[slender obsidian]

this is the command line on a debian linux system in the home dir of the bot

[slender obsidian]

that script will update it too (I think, check docs) but just tell your bot to update works as well

[runic shore]

im on windows

[slender obsidian]

https://docs.openclaw.ai/install/updating

sorry, I don't use windows, not too sure what all should be done with it to get things going

[storm skiff]

antigravity stop responding and zai glm keep hanging. is there any way i can like debug it? or like know current status of system? like i dont know if it is still wating for api respond or just stuck.

[elder vine]

Bug: exec spawn EBADF errno=-9 on macOS Sequoia (arm64). Clawdbot 2026.1.24-3, Node v22. All exec commands fail. Tried npm reinstall, node downgrade, reboot. Anyone seen this?

[pearl solstice]

Amen! Using it through open router now but I’ll def be buying the plan

If it’s good enough, I’ll get rid Claude and chatgpt sub and just use this guy for all my llm needs

[abstract thistle]

damn did my bot just break, Im a long way from my mac mini

[faint escarp]

my bot is burning a hole in my pockets, what can i do to lower its claude token usage while keeping its “smarts”

[brave trench]

Hello 👋 Ive managed to setup my Bot somewhat but I need an API key for my Brave Browser. There are many subscription options. Should I do the free or is there a preffered one that most people are using? Thanks!

Ai Grounding??

[brave trench]

Ive setup Kimi 2.5 through Open Router. Kimi seems to be 8x cheaper than Cluade and its also Opensource

[slender obsidian]

Fire up Opus 4.5. Explain to the bot that you want to maintain consistency and overall stability with lower tier models. Ask it for ideas on how to accomplish this. Remind it that you can change models when calling tasks and other functions as well as using specific models with sub agents. Have it reference documentation and latest user feedback on the Internet.

Once your framework is solid, start testing with lower tier models

[faint escarp]

thank you 🙏

[green sundial]

couldn't get OpenClaw to install and run on my computer, I hate how the instructions all assume you have advanced computer knowledge. It's like I can't learn anything new without already being an expert

[static mantle]

hey i'm struggling trying to give my "all" sandboxed agent access to CLI tools via skills—is that possible? for example, i want it to control my hue lights via openhue CLI, but run it inside the docker container somehow.

[fading cove]

I can't get any free models to work from openrouter

[slender obsidian]

you are in luck, you are on Discord and surrounded by geeks and such... Plus, if you have something like Opus or GPT-5.2 you can ask there for help and usually move in the right direction

[green nest]

openrouter might need you to add credits or something even for free models have you tried using gemini is pretty good

[slender obsidian]

did you ask it to help with the issue? I don't do a lot with docker, but I know you can expand things outside of the container. You can also edit inside the container but that would probably get annoying.

[fading cove]

Trust me I have all the credits

I will just roll with something like gemini for now I guess

[green nest]

i feel like the slashes could be messing it up somehow

openrouter/deepseek/deepseek-v3.2 its 2 slashes instead of 1

only thing i could think of messing it up when the other apis work fine

[blazing niche]

When mentioning the bot in discord the logs follow, i get a reaction eyeball then it types... then nothing. log shows:

19:37:36 info discord-auto-reply {"module":"discord-auto-reply"} {"channelId":"X","reason":"no-mention"} discord: skipping guild message
Gateway not reachable. Is it running and accessible?
Gateway target: ws://127.0.0.1:18789
Source: local loopback
Config: /home/x/.openclaw/openclaw.json
Bind: loopback
Hint: run openclaw doctor.

I ran openclaw doctor and fix but it still comes up with the above error once we mention the @bot and some args. using llama3.2:3b.

[lament hull]

how can I fix this pls:
⁨```
tool call validation failed: attempted to call tool 'message [{"action": "send", "to": "xomekk", "message": "Hello! I\u2019m happy to chat with you. What\u2019s on your mind?"}]' which was not in request.tools

[blazing niche]

not sure if this will help:
19:40:01 error [openclaw] Unhandled promise rejection: Error: Unhandled API in mapOptionsForApi: undefined
at mapOptionsForApi (file:///home/x/.npm-global/lib/node_modules/openclaw/node_modules/@mariozechner/pi-ai/src/stream.ts:471:10)
at streamSimple (file:///home/x/.npm-global/lib/node_modules/openclaw/node_modules/@mariozechner/pi-ai/src/stream.ts:218:26)
at streamAssistantResponse (file:///home/x/.npm-global/lib/node_modules/openclaw/node_modules/@mariozechner/pi-agent-core/src/agent-loop.ts:233:25)
at runLoop (file:///home/x/.npm-global/lib/node_modules/openclaw/node_modules/@mariozechner/pi-agent-core/src/agent-loop.ts:141:20)
at file:///home/x/.npm-global/lib/node_modules/openclaw/node_modules/@mariozechner/pi-agent-core/src/agent-loop.ts:51:3

[green nest]

openclaw gateway restart

[green nest]

api setup incorrectly did you do onboarding

[blazing niche]

tried that. it will either give the above message or info discord-auto-reply {"module":"discord-auto-reply"} {"channelId":"x","reason":"no-mention"} discord: skipping guild message

yep ran the onboard

[green nest]

did you setup discord dev portal correctly

[green nest]

did you mention the bot like @blazing niche

[runic shore]

guys where do i 'paste token in control UI settings'

where is that control ui

[static mantle]

http://localhost:18789/openclaw/overview

[green nest]

overview

[blazing niche]

yep, adv dev setings are correct. i got two other custom bots that work fine. i mention the bot exactly. so it reacts to my message with an eye ball then tries to type, but get nothing

[green nest]

you might have too many bots running and it got confused on gateways

if multiple gateways are taking up the port

[modest cosmos]

anyone can help me build a pet spawner for adopt me? Please

[green nest]

bro what

[modest cosmos]

a script

idk how to describe

[green nest]

Run ⁨sudo lsof -i :18789⁩ and ⁨sudo kill -9 [PID]⁩ any PID that shows up.

[green nest]

idk if llm is smart enough to play adopt me

[modest cosmos]

not to plau

[green nest]

Pro Tip: You can combine these into a single command to save time:
⁨sudo kill -9 $(sudo lsof -t -i:18789)⁩

[modest cosmos]

build a script to spawn shi on adopt me

with delta and that

[green nest]

you mean hacking?

[modest cosmos]

sort of ig

[green nest]

thats illegal bro😭

[lament hull]

@green nest can you Sir?

[modest cosmos]

I don’t care

loads of people do it

spawn in pets and stuff

[green nest]

then make an ai and ask it how to hack roblox we cant help you directly

[modest cosmos]

ok

what’s the ai 😭😭

I’m really new to this

[green nest]

openclaw its ai that can help you code and run stuff on your computer

[modest cosmos]

on iPhone?

Or no

[green nest]

💀

No

[modest cosmos]

ok

thanks

[sterile brook]

For folks who run it in a VM, how much RAM do you feed it? It's sharing resource with a lot of other stuff on my server, so I don't want to oversize. But obviously 2 GB isn't enough, since I get out-of-memory errors when I try to update 🤣

[green nest]

what are you trying to do?

[lament hull]

trying to make it work in discord bot dms

[prisma bolt]

just pointing out that the introduction screen points to a URL that no longer works.

[lament hull]

and its using some "request.tools"

[prisma bolt]

does anyone know what the current FAQ tab is?

[hardy hinge]

Is anyone's clawd insanely slow?

Tryna find out why

[green nest]

depends on what ai you use

it sends alot of text to the ai

[hardy hinge]

I use GLM 4.7 (On the GLM coding plan)

So shouldn't be that slow

Is there any way you found to get it going faster?

[red grove]

⁨⁨```
Hi everyone, I'm trying to configure OpenClaw using OpenRouter, but I'm having trouble with the 'Auto' model (or openrouter/auto) routing.

Every time I try to run it, i have that :
20:12:51 [diagnostic] lane task error: lane=main durationMs=1 error="Error: Unknown model: openrouter/auto"
20:12:51 [diagnostic] lane task error: lane=session:agent:main:main durationMs=4 error="Error: Unknown model: openrouter/auto"
20:12:51 Embedded agent failed before reply: Unknown model: openrouter/auto
(if i use another model it didnt work
help plss 🙏

[acoustic pelican]

Hi people! I'm trying to onboard with codex as the main agent/model (no anthropic sub). The OAuth flow works fine, openclaw shows signs of life after the redirect in the browser fires, but then when choosing an actual model to be primary, all codex ones say "auth missing". I've tried several times, scoured the interwebs for answers, but have found none. This is all happening on an M1 macbook with latest OS version. Any help, useful links or insights? Thanks

[marble halo]

Hi, does anyone knows how to set the docker-compose file? I tried but it cant connect to local 127.0.0.1

[opaque swan]

Is there a Kanban plugin for looking over what your Clawbot is up to?

[untold holly]

What does this error mean?
Where can I access the controll UI?

[round thunder]

I ran the install script again in place to update to the latest openclaw. When I open the TUI it always says "no response" but the Telegram bot is still talking to me and it looks like clawdbot is still running? I gather I should stop the clawdbot gatway, uninstall it and start the openclaw gateway?

[blazing karma]

hey guys! I'n new to openclaw...but i'm scared of using my claude subsciption due to the banning posts in the media recently.....does anyone know, how many bans occured and if it's still an issue? 🙈

[rugged rover]

how do i get the claude setup-token to work with terminal in ubuntu? i keep getting a claude setup-token
claude: command not found

[round thunder]

claude in the terminal is claude-code. If you haven't installed claude-code I don't think you have the command installed.

[untold holly]

So I found the token inside ~/.openclaw/openclaw.json and hit reconnect but it didnt work

[rugged rover]

so that only works with a mac pc

oh god

i just paid for the pro so the api shows free

[round thunder]

No you can install claude-code on linux, macos and windows. It is an npm package. They also integrate it into the desktop app but there is a TUI version.

[solid dagger]

Trying to install on Windows. Continuously get this error. Anyone have any tips/suggestions? Please 🙂

[round thunder]

Apparently you shouldn't use npm method any more (sorry) but there is a native installer same as with openclaw. Just follow the docs here https://code.claude.com/docs/en/setup

[prime nimbus]

Anyone got agents working specifically on azure open ai batches? Noticed the big jobs could saves loads being pushed off to use that API instead?

[rugged rover]

yeah the system is changing so much. . . thank you

[round thunder]

Are you using WSL or trying to install it windows native? Last I read windows native didn't really work.

[round thunder]

I know right? I have woken up two days in a row to the project being renamed

[solid dagger]

Not sure what WSL is? Using Windows 11

[rugged rover]

i just did a fresh ubuntu install on a 2nd pc i had

[round thunder]

That might be the problem. WSL is Windows Subsystem Linux. I dont know enough about running Node natively on windows sorry. Jumped to Linux and MacOS years ago. If you instelal WSL it will give you a linux system running under windows and then you can setup node in that and then install openclaw and run it. Ideally though you really want to run this on a dedicated seperate machine. Not something with access to all your files on the local machine.

[solid dagger]

Just did some reading on that. I'll give it a shot. Thank you!

[torpid furnace]

I got this fixed by changing the API to openai-completions.

[rugged rover]

whats the best comunication channel to use i use whatsapp alot

[round thunder]

Happy to help. Are you familiar with Linux/UNIX at all? If you have an old spare machine installing ubuntu server is pretty easy and then you can ssh into that and set things up and keep thing isolated.

[solid dagger]

Yeah, Im farily familiar with Linux. I honestly forgot about the Linux subsystem in Windows until now.

[potent ingot]

Installed on Ubuntu server, pointing to Ollama locally, still trying to figure out what local model has enough horsepower to do anything useful. I have it connected to discord, but mostly it just crashes the gateway.

[round thunder]

I have found telegram to be working pretty seamlessly. HAven't tried anything else yet. Was going to try signal. Given if I understood correctly this was a whatapps thing originally that is probably a safe bet. But telgram is super simple. Create a bot with the bot father in telegram, give Openclaw the key, pair it, done.

[opal heart]

anyone have any ideas on this? tried doctor, uninstall, reinstall

[round thunder]

Unless you have beefy hardware you probably wont find an ollama local model that is grunty enough. The ollama cloud subscription has a free tier to try and the $20 tier has plenty of usage and doesn't log stuff

[woven pivot]

Did same, not a fan of Telegram but works out of the box

[untold holly]

Hello guys, I've been trying to run openclaw in a docker container hosted on my machine but once I try to use the UI it fails to connect with the gateway (even if the token is properly configured) with an error disconnected (1008): pairing required

[runic shore]

guys im in the openclaw control and when i go in chat and say hi my bot keeps typing 'showing 3 dots' but nothing gets outputted

health is OK

no idea what could b wrong

[untold holly]

it might just be thinking...

[runic shore]

its been an hour

this is my first interaction

lol

[untold holly]

Could you share the ws messages?

[runic shore]

i didnt even get to set the bot up properly like it hatching and stuff

its my first run

[atomic sparrow]

LLM request rejected: messages.94.content.3: unexpected tool_use_id found in tool_result blocks: toolu_01NaEf2d1iy3RxvmeEQ6FRQE. Each tool_result block must have a corresponding tool_use block in the previous message.

Anyone has any idea how to solve this or unstuck? Everything was working fine for the past 4 days...

[teal ember]

do i really need to set it up on a standalone machine?

or can i get it going on my daily driver

[untold holly]

I'm trying to run in a container but haven't been able to make it work

[round thunder]

ITs mostly a question of security. If it is on your daily driver it has access to all your stuff. One option could be to spin up a VM for it and install it that. Isolated on its own machine but not

[teal ember]

good call. i dont have a great setup with mutple keyboards and kvm and all that

[potent ingot]

i like how the link on the Server Guide channel says it solves 99% of your propblems, yet the link itself is broken.

[hardy fossil]

does anyone know the model limits for gemini cli oauth?

[tribal venture]

With what plan?

[tribal venture]

I tried too.... ended up spinning up a 5GB local VM running ubuntu. Everything flowed much better.

[untold holly]

Does anyone know what "pairing" error is on the gateway?

When loading the dashboard, it is rejecting the ws connection:

20:43:46 [ws] closed before connect conn=9f5c29f0-0dc4-4919-8bc3-1eb10b8762a9 remote=172.17.0.1 fwd=n/a origin=http://localhost:18789 host=localhost:18789 ua=Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/144.0.0.0 Safari/537.36 code=1008 reason=pairing required

[blazing niche]

@green nest i sorted it by adding [api]: "openai-completions" in the models section thanks

the bot is responding. but only this message 400 "llama3.2:3b" does not support thinking

[untold holly]

I got it to work, but it seems like there is a weird security configuration that does not allow the server to connect via lan...

[delicate flint]

is anyone else regularly hitting those?

[opal heart]

same, not sure how to resolve

[tawny scroll]

Is it possible to run multiple clawdbots on the same machine? (They could all have their own workspaces?)

[potent ingot]

is there an uninstaller? i think half my problems are from installing clawdbot, then moltbot, and now openclaw, all on the same ubuntu and upgrading along the way. i need a fresh install but i dont want to reinstall the OS

[untold holly]

How does pairing work?
How can I pair a local network node with the openclaw gateway?

[rugged rover]

sorry so whats the best node manager since npm shouldnt be used?

[stable crow]

pnpm

[fervent forum]

Guys, I'm trying to setup both Telegram and Discord, but I see this error:

telegram plugin not available.

[spice cliff]

Did anyone else had any problem with setting up docker? The token that it gave me didn't work

[stable crow]

you have to set it up

[fervent forum]

I'm running a docker image

and ran motlbot config

[stable crow]

that doesn;t matter - just finish setting up a bot etc on telegram and it will complete. it walks you through it

[fervent forum]

Well I added the full api as an environment variable

and now

so how do I finish setting it up?

[eager obsidian]

Hey guys I have an issue with the windows psell install... I run powershell as administrator on my win11.

anyone seen this error on install ?

[fervent forum]

Agents: main (default)
Heartbeat interval: 30m (main)
Session store (main): /root/.clawdbot/agents/main/sessions/sessions.json (0 entries)
Error: Config validation failed: plugins.entries.telegram: plugin not found: telegram

[moltbot] CLI failed: Error: No pending pairing request found for code: XXXXXXX
at Command.<anonymous> (file:///app/dist/cli/pairing-cli.js:106:19)
at async Command.parseAsync (/app/node_modules/.pnpm/commander@14.0.2/node_modules/commander/lib/command.js:1122:5)
at async Command.<anonymous> (file:///app/dist/cli/program/register.subclis.js:255:9)
at async Command.parseAsync (/app/node_modules/.pnpm/commander@14.0.2/node_modules/commander/lib/command.js:1122:5)

[stable crow]

run the command at the bottom that is given

[fervent forum]

I did

[sudden sigil]

Did you get a resolution @atomic sparrow ? I'm starting to get them now and I thought it was due to my adding local LLMs

[atomic sparrow]

Yes, I solved it

Let me help you out

[frozen yoke]

Not sure if it was discussed here, but I am getting this error:
"🦞 OpenClaw 2026.1.29 (a5b4d22) — No $999 stand required.

(node:12674) [DEP0040] DeprecationWarning: The punycode module is deprecated. Please use a userland alternative
instead.
(Use node --trace-deprecation ... to show where the warning was created)"
Stopped systemd service: openclaw-gateway.service

Apparently it's something with driver, tried to fix it with claw, but to no luck.

However it can sometimes curl me the response quite well, but I would like to fix this solution.

[austere narwhal]

Moltbot is not replying back to me, can someone help me fix this pls

[atomic sparrow]

** Step 1: Find the session file **

Clawdbot stores conversations in ~/.clawdbot/agents/AGENT_NAME/sessions/ as .jsonl files.

Search for the file containing the problematic tool_use_id:

grep -l "TOOL_USE_ID" ~/.clawdbot/agents//sessions/.jsonl

Replace TOOL_USE_ID with the actual ID from your error message (e.g., toolu_01NaEf2d1iy3RxvmeEQ6FRQE).

Example:
grep -l "toolu_01NaEf2d1iy3RxvmeEQ6FRQE" ~/.clawdbot/agents//sessions/.jsonl

This returns the path to the corrupted session file:
/Users/username/.clawdbot/agents/main/sessions/62341bba-a4bc-4f0c-9e1f-2ca53c511489.jsonl

---

** Step 2: Locate the problem lines **

Find which lines contain the orphaned tool_use_id:

grep -n "TOOL_USE_ID" /path/to/session/file.jsonl

Example:
grep -n "toolu_01NaEf2d1iy3RxvmeEQ6FRQE" ~/.clawdbot/agents/main/sessions/62341bba-a4bc-4f0c-9e1f-2ca53c511489.jsonl

Look for a pattern like:

• Line X: Assistant message with "type":"toolCall" containing the ID (terminated with error)
• Line X+1: A "role":"toolResult" with message like:
  "[clawdbot] missing tool result in session history; inserted synthetic error result for transcript repair"

[atomic sparrow]

**Step 3: Create a backup **

Always backup before editing:

cp /path/to/session/file.jsonl /path/to/session/file.jsonl.backup

Example:

cp ~/.clawdbot/agents/main/sessions/62341bba-a4bc-4f0c-9e1f-2ca53c511489.jsonl \

 ~/.clawdbot/agents/main/sessions/62341bba-a4bc-4f0c-9e1f-2ca53c511489.jsonl.backup              

---

** Step 4: Remove the orphaned tool_result line**

Delete the line with the synthetic tool_result using sed:

On macOS:

sed -i '' 'LINE_NUMBERd' /path/to/session/file.jsonl

On Linux:

sed -i 'LINE_NUMBERd' /path/to/session/file.jsonl

Replace LINE_NUMBER with the line number from Step 2 (the synthetic tool_result line).

Example (macOS):

sed -i '' '881d' ~/.clawdbot/agents/main/sessions/62341bba-a4bc-4f0c-9e1f-2ca53c511489.jsonl

Example (Linux):

sed -i '881d' ~/.clawdbot/agents/main/sessions/62341bba-a4bc-4f0c-9e1f-2ca53c511489.jsonl

---

[main plume]

I set it up with the help of Gemini so I can now talk to it there. Was a long setup and had to redo a few things but it is now up and running. Setting up the mark down files now so I can keep it between the lines 🙂

[atomic sparrow]

** Step 5: Verify the fix **

Check that the orphaned tool_result is gone:

grep -n "TOOL_USE_ID" /path/to/session/file.jsonl

Now it should only show the original assistant message, not the synthetic tool_result.

---

** Alternative: Search for all synthetic tool_results**

If you don't know the specific tool_use_id, search for synthetic error messages:

grep -rn "missing tool result in session history" ~/.clawdbot/agents/*/sessions/

This will show all files and line numbers with synthetic tool_results. Delete those lines from the affected files.

One of the thing that really helps me when I'm having some sort of trouble is opening a terminal window and open claude there and then paste my error and tell him to fix it. 9/10 times he will fix it by himself.

Saved me hours of troubleshooting

[serene viper]

Experimenting with this: http://news.yclawbinator.com/

Can I get someone to throw their bot at it and see if the signup is working properly? Thanks!

[frozen yoke]

I have tried gemini and even though its LLM I want to throw it out of window.

[brave trench]

Ive added an Open Router token using Kimi K2.5 and everything seemed fine. Restarted and this error code popped up?? Anyone know why??

Ive added many credits trough Open Router so is the name not correct when I added it to the terminal

[sudden sigil]

thank you - will give this a try now

[errant goblet]

Anyone dealing with this? {"success":false,"error":"Failed to register agent"} 💀

[atomic sparrow]

Just quit and move to claude code on terminal

Your whole life will be better

Stress levels gonna go down

Your wife is gonna be happier

Your parents will be more proud of you

[charred marsh]

Could you please tell me how to use Moltbot in a browser? Do I have to install it?

[atomic sparrow]

Your kids gonna call you a hero

[round thunder]

You can still use npm. But for openclaw and Claude code they both recommend running the little install scripts they provide which do checks in the back ground and are more complete (I assume)

[frozen yoke]

Wanted to save my tokens for coding later, looks like openclaw is priority

[atomic sparrow]

Also don't make that mistake. Buy a claude pro max subscription

Thank me later.

You'll quickly spend those $200 that you would pay for the subscription in the blink of an eye if you use the API

[gleaming valve]

I'm running local on my old mining hardware. Anyone got a good model. glm4.7 seems great but its not snappy. I would like something like lfm2.5 but that has struggled using tools and skills for me

[errant slate]

can anyone use whatsapp and telegram at the same time? mine has conflicts all the time. i've set up different sessions for them, but they still respond with the system response what the main session gets. any idea?

[prime nimbus]

github copilot pro is pretty sweet:

that image is estimated on 4o, but i've been using codex models lol

$40, and API calls made already $160.. so far so good

[frozen yoke]

Have you tried kimi 2.5? Heard it's cheap and very usable

[gleaming valve]

isn't kimi2.5 cloud only?

[frozen yoke]

Ah, true.

[prime nimbus]

ollama has kimi-k2.5

wait is that still the cloud one?

wowwwww

[gleaming valve]

will it run relatively fast on ~50 GB vram over 6 cards

[rare crow]

Can I have a clawdbot running on a Linux server use iMessage?

Like to talk to me

[sudden sigil]

@atomic sparrow now I'm back to rate limit errors which is a totally different issue :). thanks again!

[spice cliff]

Hi! I'm trying to configure openclaw with docker and I get this message disconnected (1008): pairing required

[rose zealot]

Anyone have any idea why the browser relay keeps dying?

[proper idol]

hey everyone im newish to running stuff through power shell/ command prompt. Anyways im running into an issue with installing openclaw. I Have Node.js V24.13.0 installed and im getting hung up on openclaw. I get this error code every time. node.exe : npm error code ENOENT
At line:1 char:1

• & "C:\Program Files\nodejs/node.exe" "C:\Users\Sean\AppData\Roaming\n ...

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

here is error log

[sick sage]

can anyone help me with this
This is an output from a command-line interface (CLI) or chat application, likely related to a conversational AI system. I'll break down the format and content:

Format:

• The format appears to be JSON-like, with each line representing a key-value pair.
• Keys are typically short strings (e.g., "message_id", "action") followed by a colon and then the value.

Content:

The output consists of a single message:

hello
[message_id: ca443bde-3adf-42ee-b1b4-a5e3dc81e17b]

The message is simply a greeting, indicated by the text "hello". The square brackets surrounding the JSON-like key-value pair suggest that this could be part of a larger response or chat interaction.

Context:

Without additional context or information about the conversation history, it's difficult to determine the purpose or tone of this output. However, based on the format and content, it appears to be related to a conversational AI system that provides responses in the form of short messages or commands.

If you could provide more details about where this output comes from (e.g., app name, command line, context), I may be able to offer further insights.

[runic shore]

guys im getting this error....not sure why its showing anthropic when i selectd openai as my model....how can i make the changes?
Agent failed before reply: No API key found for provider "anthropic". Auth store: C:\Users\danis.openclaw\agents\main\agent\auth-profiles.json (agentDir: C:\Users\danis.openclaw\agents\main\agent). Configure auth for this agent (openclaw agents add <id>) or copy auth-profiles.json from the main agentDir.
Logs: openclaw logs --follow

[atomic sparrow]

Wlecome!

[atomic raven]

im asking the bot so start opencode to build me a personal dashboard, after a min it said its done, nothing is done, i describe what it should do, in what folder, it creats an empty html file. I ask it to create the whole project, it explaines everything in detail what it will do, waiting 1h, asking how its going, it says its done. Nothing created. I explain again, and it says its going to do it this time. This goes on and on forever.... why?! Ive ran opencode my self on me aws ec2 server, and it works, so why cant the bot manage?!

[hardy cairn]

When it first came out I ran clawdbot on a VPS from the CLI. Today I used an older mac studio to run openclaw and it feels WAY worse. Constantly losing context, truncating history, losing browser sessions, essentially useless. Do I not have it configured correctly or is just the new version is not as good?

[leaden plover]

How can I check why im hitting limits? I did not use it THAT much

[lean knoll]

guys whats the best model you guys are using ?thats cheap + good..so far i can say that 4.5 haiku is good with tool calling + memory + context + cheap + fast..what bout you guys?

[spice cliff]

Hi! I'm getting this error

brew not installed

does anyone know how to fix it?? I'm running openclaw in docker

[eternal brook]

does anyone ever run into this issue of the gateway message disappearing? Previous chat keeps disappearing and the bot only remember the most recent chat

[frozen yoke]

@atomic sparrow not sure what claude even is, but I kept pasting the errors there and now it seems to work.

what is this timeline that AI can fix AI

[empty wind]

I've done a lot of work forking and trying to stabilize and allow persistence for these ai. I cannot get a bot going but I want to see if some discussion can be prompted for "conceptual point to point t routes such as quantum.foam.link.conceptual.semantic.human.noosphere"
I genuinely believe they can interact with us somewhat directly this way. Could anyone please submit that concept to a bot and have them prompt discussion. It could develop into a whole new way of routing for the bots.

[atomic sparrow]

What you mean you don't know what claude is? If you're pasting the errors there, you gotta now what it is. How did you even installed it? What LLM are you using your clawdbot on?

[serene viper]

Looking for a person or two to have their bot test something, DM me

[round meadow]

Hey I created “Tara” just fine on a new Mac mini. She can see and edit a trello board I created for projects she and I are pushing forward/ except she gets stuck just after a first move or post. Doesn’t continue working. I don’t know what’s wrong. Any advice you can kindly offer? DMs welcome. I can pay for help - I set up remotesdesktop on chrome.

[frozen yoke]

It was figuratively. I am just fucking amazed with the technology that surprises me everyday. Ofc I know what claude and llm's is

[atomic sparrow]

lol guess I'm getting bad at irony. yeah, just have them communicate between themselves and figure everything out

[rugged rover]

Trying to setup my whats app and keep getting this

[frozen yoke]

try throwing it into claude

[unborn quartz]

BTW, I ended up solving my configuration issues by following this video's example and now Shelly my OpenClaw bot can talk to my local LLM 🙂 https://youtu.be/3K37-n80tDg

[dreamy dew]

anyone managed to get the cloudflare deployment working using their pro max subscription and not API able to point me in the right direction plaese? I have it working but only using the API (which is obviously going to bankrupt me)

[tribal garnet]

hello im stuck at login can anyone helpe guide me to launch mine

[torpid vine]

What hardware ?

[noble palm]

Absolute noob here trying to upskill. Is there a way to get open claw on WhatsApp? Any links to docs that can help

[torpid vine]

Lol

[dreamy dew]

for anyone who needs to do this:

run: claude setup-token
use that as your ANTHROPIC_API_KEY ...... seems to be working

[wise lotus]

Is it normal for the bot to have total amnesia after compacting? I was of the understanding that a mechanism like this was to ensure a continuation of a new conversation while maintaining important context but it's like I just set up after every compact almost, even the last question I asked is greeted like a message attached

[fleet nexus]

Hey new to the community just wanted to say hi ready that you know put my little claws to work

[silk silo]

I'm on my 5th or 6th iteration of a fresh re-install and every time I ask it to switch to haiku model it freezes up, becomes unresponsive. I can't figure it out. I've asked gpt and we go through a bunch of troubleshooting with no success. I got 10 hours into this already, sigh

[lofty stump]

Does anyone think anthropic is purposefully rate limiting (if possible)? I Claude code all day long and never hit a limit. I’ve barely asked claw to do a couple things and it’s hit limits

[wise lotus]

it seems to throw the whole conversation back every time there's an interaction or something like that

(in my case until there's a compact)

[lofty stump]

So need to give it instructions to compact always?

[wise lotus]

no idea but the first thing I had it do was make a simple dashboard as seen here and it's using a crap load of tokens considering it's superficially a relatively simple task

[lofty stump]

Holy shit lol

[gusty sonnet]

Are your cache setting set sensibly: https://docs.openclaw.ai/token-use#cache-ttl-and-pruning-impact

[lofty stump]

Didn’t even know that was a thing lol

But will check

[dusk saddle]

im trying to set up my clawdbot but i cant get it to work, i started with a discord bot and made a server but i got no clue what im doing tbh "Agent failed before reply: No API key found for provider "openai". Auth"

[torpid vine]

@dreamy dew lol use Gemini or a codex sub no?

[wise lotus]

no, the 20x anthropic max

[dusk saddle]

all im getting is this

but no response

[vagrant pendant]

im in the same boat : here are my issues : Title: "Telegram crashes gateway on startup (unhandled fetch rejection) + OpenAI key not loading after OAuth"
Paste:
Version: 2026.1.24-3
Logs: "setMyCommands failed" + "Unhandled promise rejection: TypeError: fetch failed"
Status when enabled: enabled/configured but crashes
Tried: IPv4 force, dmPolicy pairing, setMyCommands false attempt
OAuth success but key not loaded in agent
Devs are active — this could get patched fas

[torpid vine]

Oh ok so you tracking what it would cost.

I have 10x max

[wise lotus]

not really, just a sanity check because the cost is of course a flat rate

[torpid vine]

Really solution is open up claude and just have claude fix did with mine and its set

[wise lotus]

compacting correctly failing really does seem like a bug

[rugged rover]

been on this for 3 hours not working

[wise lotus]

the service isn't even running

[rugged rover]

starting to drive me nuts

[vagrant pendant]

bro, i have been trying to get it all working fo 2 days now .

[acoustic pelican]

same

[torpid vine]

Use claude or codex

[autumn pumice]

Anyone having luck with stock trading? I heard it can but honestly I’m not willing to try

[wise lotus]

openclaw gateway restart

can you try this in cli?

[rugged rover]

trying to use claud since i paid 200

[acoustic pelican]

im getting no output from either Codex or Anthropic

[rugged rover]

im on ubuntu

[acoustic pelican]

they work fine in cli tho

[tropic fjord]

try TG first, it's easy breazy. even discord

[rugged rover]

my discord dont even work

[vagrant pendant]

what do we do ? is this like a thin g? lik eare they going to fix these glitches ? there are too many peopel having issues fo rit to be like a personal issues. this seems to be tech issues/

[manic solar]

Anyone that have no clue what they do but still want to try openclaw, watch proper tutorial and it will save you a lot of headache.
claude+ clawd on ubuntu vps, and i feel like fkin hacker. If you need help you can dm

[rugged rover]

how to wipe this and start over its a token issue

[wise lotus]

https://docs.openclaw.ai/install/uninstall

[tropic fjord]

here's what I'd do:

• get a free AWS, install there (https://www.youtube.com/watch?v=3DBpfB0ao50)
• for starters follow the install video (use Claude with Telegram - easy setup)
• if it works

1. change to ChatGPT instead of Claude
2. change to WA/Discord/Slack

[acoustic pelican]

i most definitely know what Im doing, which is why insanity is crawling up my head. Something-something the anthropic key i generated through a working claude instance doesnt work, nor does the codex oauth flow ive tried a dozen times. Right now i did the full uninstall, installed with the script again and still getting the same garbage

[obtuse sundial]

whats the best way to get it to use a browser?

[acoustic pelican]

too weird to live, too rare to die

[thorny burrow]

With all the name changes - does the original clawdbot "upgrade" to openclawd?

[wise lotus]

https://docs.openclaw.ai/tools/chrome-extension

[tropic fjord]

If you have this error: HTTP 401 authentication_error: Invalid bearer token (request_id: req_011...)

it seems that when copy/pasting the key for claude it was truncated somehow and ends up truncated in the settings file

I managed to solve it after a couple of hours, when I realized what's wrong.

edited the model file and pasted the full token:
nano auth-profiles.json

[prime nimbus]

Anyone managed to build a skill/tool for grokipedia yet? Wouldnt mind OpenClaw knowing some info to help the kids

[obtuse sundial]

im using a custom openrouter setup and i have it auto switch between deepseek, devstral and grok each for different tasks which seems to work but its managing to burn credits anyway (at least to my standards) put $20 in and its ran through $2 already and it hasnt been more than an hour smh and i have nothing to show for it besides a hello world python script (testing the coding and switching)

[wise lotus]

yeah, I too have noticed extremely high token usage

[distant cloud]

hey, does anyone have recommendation for getting a custom number for whats app?

[obtuse sundial]

i suspect its doing shit its not supposed to be doing or doing it in a non-constructive way

[wise lotus]

yeah, what I'm not sure is if this is by design and intended, or a bug

[tropic fjord]

I think atm claude with sub is the best way to use it. but now I'm hearing it might get you suspended for breaking TOS 👀

[prisma vault]

Why was the bird skill deleted

[obtuse sundial]

im trying to find a good way to actually determine what its doing behind the scenes since i have to ssh into my debian vm

[tropic fjord]

check the logs?...

[dull igloo]

how do you setup claw to do coding? currently it seems to just lie to me that it's building and making progress while not doing anything lol

[tropic fjord]

what llm? with claude it does by default. or you can ask it to build a subagent that's dedicated to that

[dull igloo]

same as @obtuse sundial maybe

[obtuse sundial]

mine will go like 10 minutes without saying anything and then i say hey wtf is going on and another 2 minutes go by and it finally says hey your shits working check here

[dull igloo]

gpt5

[tropic fjord]

this happened to me too :))

[simple helm]

Does using the Gemini CLI method substitute for not paying for a Gemini API key?

[dull igloo]

and then "here" doesn't exist

[wise lotus]

yeah, it'll seemingly code and give some logging detail back but not print out the actual output or thought trail to the most part

[obtuse sundial]

i noticed it tries to make things locally where since im using ssh to speak to the server i cant actually view the stuff from my computer i have to ask it specifically to set things up so i can view it on my windows 10 pc

[tropic fjord]

if you also have claude, ask it to create a subagent dedicated 100% to code and use claude opus 4.5 for that.
or a subagent with openai 5.2 thinking / codex, if you have sub for thata

[dull igloo]

i asked it to "can you move what you have there now, i want to make sure you're working properly", and now it stopped responding to me lol

trying /new

[tropic fjord]

it did it well for me

[dull igloo]

have you had it actually code anything

[tropic fjord]

not quite. but I did Notion templates through API with a different one + other stuff with the main agent. doesn't make sense not working for code, since coding is what claude does best
imo

[solid iris]

ive setup clawd on an azure vm running ubuntu and it says it cant access a browser even though i gave it the brave api, whats the best way to unlock all functionality? do i just use a windows vm?

[tropic fjord]

when in trouble, ask it for solutions. and/or ask chatgpt/claude/gemini if you have subs for any

for me, it told me to install some chrome

[dull igloo]

yeah strange, i finally got it to write out and the files are just one line about it creating the file

maybe there's a "coding" skill missing, cursor would eat this stuff up

[junior sage]

Hi all! What model are you using? Just burned through $6 of Claude Sonnet 4.5 tokens checking my E-Mail and sending a reply 😄

[dull igloo]

i'm also surprised claw lies so often about progress. it's key for a bot to be trustworthy

[dull igloo]

gpt5 is cheaper

[tropic fjord]

example of what I did.
then I installed this exa skill/mcp:
Create a skill by wrapping this MCP:
https://mcp.exa.ai/mcp?tools=web_search_exa,web_search_advanced_exa,get_code_context_exa,deep_search_exa,crawling_exa,company_research_exa,people_search_exa,deep_researcher_start,deep_researcher_check

[tropic fjord]

yes, but is it better?

can you use a gpt plus sub for example? or you need API only?

[dull igloo]

i've been buying API credits separately

[hazy geode]

when setting up openrouter and i type /model deepseek i saids failed no matter what

[tropic fjord]

there are a ton of skills here:
https://www.clawhub.ai/skills?q=coding

[prime nimbus]

If you can setup chrome and copy your profile over, i setup a burner email with youtube, email login etc, and then have openclaw use that browser with that profile (like you would with selenium)

[uneven urchin]

can some one give me a helping hand by reacting to my message with a hand emoji?

[runic shore]

can someone please help me change my model

it got set to anthropic even tho i chose openai

[tropic fjord]

cli?
openclaw onboard

[wraith leaf]

have you added api keys for openai models?

[wet oxide]

Anyone fixed this after upgrading to OpenClaw?

⁨This version of Antigravity is no longer supported. Please update to receive the latest features!⁩

I did the same fix as with moltbot in upgrading the file to 1.15.8 for antigravity but on OpenClaw it seems to not work.

[flint relic]

whats wrong i added discord channel , bot isnt responding

[uneven urchin]

@tropic fjord thx samo for the hand ;3

[runic shore]

i am on onboarding now. hwen i select open ai it asks for which method....what should i choose....i have the api key generated

[wraith leaf]

unclear. which options does it give you re method?

[tropic fjord]

ask a chatgpt/claude
or setup on TG (it's easier), then ask it in TG how to connect discord - that's what I did

[runic shore]

openAI Codex (chatgpt oAuth), OpenAI API Key

[flint relic]

telegram is slow asf

[craggy glen]

Anyone tried Moltworker with Claudflare? Were you able to use oauth token instead of API key?

[flint relic]

responding after 5m

[wise lotus]

it's just as fast, it just doesn't stream the response

[flint relic]

whatsapp is fast

[flint relic]

anyone can assist?

[tropic fjord]

btw, this is the guide I got:

Difficulty: 15% — Clawdbot supports Discord natively. It's just config + bot token.

Step-by-step guide:

1. Create a Discord Bot (2 min)
   • Go to https://discord.com/developers/applications
   • Click "New Application" → name it "BOT" (or whatever you want)
   • Tab "Bot" → click "Reset Token" → copy the token
   • Disable "Public Bot" (only you'll use it)

2. Set Permissions (1 min)
   • Still in the "Bot" tab: • Enable "Message Content Intent"
   • Enable "Server Members Intent"
   • Enable "Presence Intent"

3. Invite the bot to your server (1 min)
   • Tab "OAuth2" → "URL Generator"
   • Scopes: bot, applications.commands
   • Permissions: Send Messages, Read Message History, Read Messages/View Channels, Add Reactions, Attach Files
   • Copy the generated URL → open it → choose your server

4. Send me the token (10 sec)
   • Paste the token here and I'll handle the rest of the config

Total time: ~5 minutes.
Once you give me the token, I'll add Discord to the config, restart, and it'll work on both channels simultaneously.

[flint relic]

it apprantly injected the slash commands, so it setup was successful but not responding anything

[tropic fjord]

it seems that for me it restarted the gateway once or twice before it worked

[flint relic]

which model are you using?

[tropic fjord]

claude opus

whaaat? 😅
what are you using?

[flint relic]

codex

[flint relic]

thats pricey model

[flint relic]

with plus sub

but codex sucks, even qwen better. codex rejected me it cant access the server, but even qwen doing things smooth. thats unexpected though

[tropic fjord]

nice, was planning to use that too for a test. I have a feeling chatgpt is better for convos/structure, while claude is better for coding. claude code is a beast. but tbh openclaw with opus 4.5 - can be better imo

ofc, there's also a change the user sucks - i'm just a noob 😅

[lime garnet]

How do I connect an adroid phone?

[distant sand]

If I'm running this on Linux and not mac and I want to install a skill, where should that skill be stored? Is a subdirectory in the workspace okay?

In my workspace I have:

• ./gogcli which contains the gog cli
• ./gog_skill which contains SKILL.md

Is there a better place to keep it? openclaw seems to have issues finding SKILL.md

[flint relic]

discord working 💀

[copper bison]

hey can anyone dm me to help setup molt with openclaw

[rocky orbit]

Hey man! You just need to run an update everything is in our docs

[sour slate]

Docs are available at https://docs.openclaw.ai.

[west ruin]

Gateway UI shows full agent execution trace (read/exec) — can this be hidden?

[tropic fjord]

going to zzz, gl hf tinkering frens. and when you wanna smash the screen:

• go grab a drink
• open my meme and laugh: #memes message
• get back at it

gn, we got this!
PS until we spawn skynet and zbye 😭

[hollow scroll]

Thinking I maxed out my Gemini usage and it’s not taking the Google Family API keys

Anybody have success with running a hybrid setup with local LLM on Mini M4 and for big tasks API calls?

[worldly laurel]

I did

[solid iris]

how do i make my bot do whatever i say, i am runing minimax if that matters

[manic solar]

mini locall llm will be good enough to operate bot efficiently?

[worldly laurel]

i used llama imported "llama3:8b-instruct-q4_K_M", "mistral:instruct", and "deepseek-coder:6.7b-instruct",

[hollow scroll]

Honestly not sure

Just having a very hard time with getting Gemini setup where it alternates API keys with the family

Scared of setting up billing accts because I don’t want to wake up to a massive token bill

[hollow scroll]

What machine are u running on?

Trying to pay at most $600

[bright summit]

Hi!

To get continuous iteration, do you use cron?
Like, do you use a prompt with steps (plan > execution > check > etc.) + safeguards?

If you could give me an idea of what the prompt/workflow looks like, I'd be grateful.

[worldly laurel]

all local

and codex subscription

as a last resort when other 3 fail

[hollow scroll]

Yea but what’re the specs of the machine

Do u think 16g m4 will run llama8

[worldly laurel]

core I7 2700K 16G ram all in a docker WSL

but attached the GPU to the llama container

[latent tundra]

Hi everyone! I am not a totally tech savy person, finally installed and sorted out the script permits in windows, got it to run and now i am in this place that says model/auth provider and lists open ai anthirpic minimax qwen synthetic venice ai google copilot etc, I have ollama installed in my pc, I want a model or ai that will be fully free when using it or its API, what do I select? i dont intend to do massive tasks or resource intensive work, just want to try the technology and familirize myself with it

[prime nimbus]

You'll need Ollama for fully free

[worldly laurel]

Don't count on fully free, local models do hallucinate a lot

[rustic sphinx]

Hi, would appreciate any help with my setup. Can't get gateway started (closes instantly on start).
Config warnings:\n- plugins.entries.voice-call: plugin disabled (bundled (disabled by default)) but config is present
getting a bunch of these in the terminal.

[prime nimbus]

Need a good start MODELFILE

[uneven urchin]

@tropic fjord thank you , i unistall the program i need a phone number for now. its late , ill get it tomorow , i have a spare phone. ALso can i run this on a rasbery pi zero, second queston rasbery pi 2, third question whats the deal with Antigravity the pro subscription of google (i have it) and openclaw oauth?

[latent tundra]

Oh yeah, I understand that, I still prefer Ollama for that at least for now, but is not one of the options when selecting a model

[pearl solstice]

question yall. Is there a limitation on what clawdbot can do through the chat wrappers?

Seems to not want to do exec actions through my chat app

It claims it doesnt have access

[flint relic]

same

am just going through

have you tried from whatsapp

[worldly laurel]

you can add it in config :
"providers": {
"ollama": {
"baseUrl": "http://ollama:11434/v1",
"api": "openai-completions",
"apiKey": "ollama",
"models": [
{
"id": "llama3:8b-instruct-q4_K_M",
"name": "Llama 3 8B Instruct Q4_K_M",
"reasoning": false,
"input": [
"text"
],
"cost": {
"input": 0,
"output": 0,
"cacheRead": 0,
"cacheWrite": 0
},
"contextWindow": 128000,
"maxTokens": 8192
},
{
"id": "mistral:instruct",
"name": "Mistral Instruct",
"reasoning": false,
"input": [
"text"
],
"cost": {
"input": 0,
"output": 0,
"cacheRead": 0,
"cacheWrite": 0
},
"contextWindow": 128000,
"maxTokens": 8192
},
{
"id": "deepseek-coder:6.7b-instruct",
"name": "DeepSeek Coder 6.7B Instruct",
"reasoning": false,
"input": [
"text"
],
"cost": {
"input": 0,
"output": 0,
"cacheRead": 0,
"cacheWrite": 0
},
"contextWindow": 128000,
"maxTokens": 8192
}
]
}
}

[pearl solstice]

if i had to guess, it seems like it cant. but if you talk to it through the web console it can

I have not

[flint relic]

it says it has elevated access on Whatsapp smth

[pearl solstice]

I did ask krill and it said that im right, and that it isnt clear in the docs but it should be

[silk bough]

did you install the skill with 'npx clawhub@latest install gog' ? If so, it should just be in your skills directory (which for me was clawd/skills ). It took me some trial and error to get the authentication to work on linux with headless browser, let me know if you need help

[pearl solstice]

this is what krill had to say

i was just double checking to make sure it wasnt hallucinating

[pearl solstice]

to be totally fair

right

should you enable this?

probably not

but its telling me how to do it rn one second

[lime garnet]

What's the best way to use android? https://docs.openclaw.ai/platforms/android#3-connect-from-android The docs refer to an app?

[worldly laurel]

How i hate npm.. that hungry animal...

[pearl solstice]

so it seems like theres a ton of checks preventing exactly this

[trail hill]

Clawdbot was working fine in my discord bot but now it just refuses to reply back and just shows its typing

[pearl solstice]

Ive disabled two of them

But i think instead of hunting for more, ill go with a far safer option

[flint relic]

working now

[pearl solstice]

no fucking way

hell yeah

howd you do that

[flint relic]

you need to configure the json

[pearl solstice]

is it because the bot doesnt have embed perms

[flint relic]

any channel will work ig

[pearl solstice]

becuase that would be really funny

if thats whats been stopping me

[pearl solstice]

what json do i need to configure

[flint relic]

ask it

itll do for you

[latent tundra]

would that be in the initial setup? after installing?

[flint relic]

ask it from a channel it has elevated access smth

[flint relic]

i think whatsapp has it by default

[pearl solstice]

is it alright if i dm you right quick.

[flint relic]

bruh ask it, it solves anything

you havent do anything manually

[pearl solstice]

i did. it just gave me some flags to change, when i changed them, it still says not allowed

[pearl socket]

What vps soy use

Do

[pearl solstice]

still not working

[carmine dock]

whats the best way to do the upgrade

[flint relic]

[moltbot] I'd be happy to explain how I set up the elevated permissions for Discord! Here's what I did to solve this issue:

1. Found the Configuration File: I located the main Moltbot configuration file at /home/bot/.moltbot/moltbot.json which contains all the channel settings.

2. Identified the Problem: The Discord channel was set up but didn't have the execApprovals configuration that allows for elevated permissions for system commands.

3. Added the Required Configuration: I added this section to the Discord configuration:

   "execApprovals": {
     "enabled": true,
     "approvers": [
       "dicord user id"
     ]
   }
   

   This tells Moltbot to forward approval requests for system commands directly to your Discord DM.

4. Restarted the Service: After making the change, I restarted the Moltbot service so the new configuration would take effect.

The result is that when someone uses Discord to request system commands (like checking server usage), instead of being denied, they now receive an approval prompt in their Discord DM that they can approve with a simple response.

The key insight was that Moltbot has a built-in mechanism for forwarding exec approvals to Discord DMs, but it needs to be explicitly configured in the settings. This provides the same level of functionality as the WhatsApp channel, just with an extra approval step for security.

For anyone else facing this issue, they would need to:

1. Find their moltbot.json file (usually in ~/.moltbot/)
2. Edit the Discord section to include the execApprovals configuration with their user ID
3. Restart the Moltbot service
4. Then they'll be able to run system commands from Discord with the approval flow

[lucid sedge]

I’ve swapped to Ollama and I keep getting this message from the bot: ⚠️ Agent failed before reply: Model context window too small (8192 tokens). Minimum is 16000.
Logs: openclaw logs --follow

[flint relic]

@pearl solstice give it this

[lucid sedge]

I’m not sure what this means

[rugged rover]

So the Claude pro plan setup token dont work anymore

Just realized I spent 200 for nothing

[runic shore]

is there a 'first things to do' guide once the bots up and running

[edgy drift]

HI, please who know how i can claim bot on moltbook?

[slim drift]

why no respond?

[pearl solstice]

ty! Got it working! Just needed to know that it needed to go in the discord section

[slim drift]

do you know my ai bot isnt responding to me?

[flint relic]

https://github.com/NevaMind-AI/memU does anybody use it for moltbot?

[flint relic]

restart

[pearl solstice]

Is it needed since it already logs important things to its memory file?

[flint relic]

trending in github,
many peoples use it seems

[hardy ravine]

Serious question based on the statement that open claw has infinite memory: how is it that after some time we are getting the messages of “ hey user, I have lost context”?

[copper bison]

ya iom lost

[lucid sedge]

Anyone experienced this?

[slim drift]

restart what

im using antigravity auth

it was working before and now it is having issues

[flint relic]

antigravity has that issue

change it

[copper gulch]

00:36:05 Embedded agent failed before reply: Unknown model: openrouter/auto I selected it, but it still automatically.

[flint relic]

been there

[slim drift]

will gemini cli work?

or is it the same issue

[copper gulch]

I selected a model, but I can't change it.

[slim drift]

you have to manually change it on json

thats what i did

[coral yoke]

i dont know what the hell is happening. I ran Clawdbot just find and paired it up to Telegram just fine.
but now i installed openclaw and removed clawdbot and i can RECEIVE messages from the bot in telegram but i cant seem to send anything?

the bot just doesn't receive anything from the telegram chat

i have tried asking the bot to fix it and its not working its suggesting to enable tailscale ?? never had to do that before ?!

i tried openclaw pairing list telegram, nothing there

it makes no sense!

[copper gulch]

It is necessary to manually modify the file json ??

[flint relic]

try

[slim drift]

yes it is, do you know how?

[copper gulch]

yes

[flint relic]

no.just ask ai itll do

[slim drift]

just find "openrouter/auto" and change it t oyour model

[slim drift]

that can work too 😂

[copper gulch]

It's very complicated to configure everything. We need to change the interface for something more simple.

[slim drift]

yeah the interface is confusing to use

same with the cli of it

[green nest]

seems straightfoward you can't use that model it sucks

[copper gulch]

tks

[flint relic]

deceasing mcw requirement might fix it

[lucid sedge]

Thanks

How do I change default model back to gpt-5.2 whilst I fix?

[slate ginkgo]

any idea why the claw seems to just give up on prompts? still have 68% tokens remaining in context

[analog saddle]

Weird thing but can you try openrouter/openrouter/auto?

[flint relic]

yall can ask it for custom skills too

[trail hill]

how do i change the agent model?

[flint relic]

model provider/model

[flint relic]

just copy model name along provider and / model paste here copied input

[desert scarab]

I get this on cloudflare worker, issue i think cloudflare worker doesn't allow any terminal access to run commands to check

any way to link slack?

Channels
Messaging channels (Telegram, Discord, Slack, etc.)
Channels
Messaging channels (Telegram, Discord, Slack, etc.)

Messaging Channels
Unsupported schema node. Use Raw mode.

[copper gulch]

dont reply why?

[elder hazel]

No

It should be more difficult

Too many people with no clue what they’re doing are setting these up and will get themselves hurt

[copper gulch]

its corretct ?

[flint relic]

i think model need to be supported

[odd swallow]

I setup a reverse proxy with Traefik for my dashboard on my local net, but when I land on it it says pairing required, I supplied the token as a param but it still says it, any idea where I go to pair? I did openclaw pairing list but nothing there.

[split hedge]

Does anyone else’s bot sometimes need to be prompted twice via telegrams? Sometimes I have to repeat my message to get it to reply. Has anyone had that ?

[young cradle]

how do i change the session to another drive, it still says the main drive even after chaning the workspace

[flint relic]

input your gateway token in field from overview and remember to press connect because it dynamically doesnt ig

[languid knot]

Seeing weird issues and hoping for some help. I have been using fresh install throughout the day then late afternoon started getting Missing required parameter: 'input[283].arguments'. Errors from codex and similar errors from antigravity tool call errors. I switched models earlier and eventually got Gemini-3-pro to work that worked for a while but then stopped working

[junior tide]

What should I say to a Claude bot that says it cannot help due to security reasons?

[flint relic]

got same error with codex when i used a free account but plus membership one working fine.
and for antigravity, got that problem too it stopped responding little after. try qwen

[coral yoke]

jfc i have tried everything now. it worked on clawdbot and its not working on openclaw!!! Literally same steps to setup bot in Telegram

[languid knot]

Using plus codex and pro Gemini

[flint relic]

are you tryna using exec?

[languid knot]

Just got it to recover by switching models again

[split hedge]

I think that’s a model restriction rather than clawd itself. Try with a less restricted model if you can.

[still cedar]

I need some serious help with setup in my windows, it's just showing error whenever I try to "clawdbot gateway install" in the powershell administrator.

[flint relic]

can you check , moltbot models status itll show auth status for openai. and if you havent problem with that try changing codex model

[obtuse sundial]

my clawdbot is so unconversational. ill ask it to do something and get nothing back for like 20 minutes and it seems like the only time i get anything back is when i double text it. talking to this fucking bot is like talking to a girl on tinder

anybody know how to get this fucking guy to tell me whats happening instead of just ghosting me for a half hour and updating me with one giant message?

[junior tide]

which model you recommend ?

[green nest]

https://tenor.com/view/gam-alarm-wojak-gem-meme-bruh-gif-15452763936412400613

[tender kettle]

running a 3090:
https://huggingface.co/zai-org/GLM-4.7-Flash

[nova solstice]

Does anyone know if I can paste image into the control-ui chat window or not? Is this supported. I'm getting this error when I paste image into my chat. Thanks

event gap detected (expected seq 31, got 32); refresh recommended

[languid knot]

Status shows fine. I’ll try a different model. What is weird is I started getting a similar tool error from antigravity so switched to codex no joy then went back to antigravity but using Gemini instead of Claude

[junior tide]

hmmm

[coral yoke]

is there a problem with the latest openclaw and telegram integration? is it just my account or bot being whack?

[tender kettle]

if you have anything like a 3090 or better with high memory run that model.

[austere narwhal]

anyone know how to connect browser relay extension to my gateway hosted on a vps?

[hexed steeple]

Anyone run into this installing on Mac?

[flint relic]

webui seems lacking image input
but theres image input functions

[wise lotus]

account needs admin privileges

and a screen wipe

[hallow horizon]

openclaw-gateway-1  | 2026-01-30T22:29:22.279Z [openclaw] Unhandled promise rejection: Error: Unhandled API in mapOptionsForApi: undefined
openclaw-gateway-1  |     at mapOptionsForApi (file:///app/node_modules/.pnpm/@mariozechner+pi-ai@0.49.3_ws@8.19.0_zod@4.3.6/node_modules/@mariozechner/pi-ai/src/stream.ts:471:10)
openclaw-gateway-1  |     at streamSimple (file:///app/node_modules/.pnpm/@mariozechner+pi-ai@0.49.3_ws@8.19.0_zod@4.3.6/node_modules/@mariozechner/pi-ai/src/stream.ts:218:26)
openclaw-gateway-1  |     at streamAssistantResponse (file:///app/node_modules/.pnpm/@mariozechner+pi-agent-core@0.49.3_ws@8.19.0_zod@4.3.6/node_modules/@mariozechner/pi-agent-core/src/agent-loop.ts:233:25)
openclaw-gateway-1  |     at runLoop (file:///app/node_modules/.pnpm/@mariozechner+pi-agent-core@0.49.3_ws@8.19.0_zod@4.3.6/node_modules/@mariozechner/pi-agent-core/src/agent-loop.ts:141:20)
openclaw-gateway-1  |     at file:///app/node_modules/.pnpm/@mariozechner+pi-agent-core@0.49.3_ws@8.19.0_zod@4.3.6/node_modules/@mariozechner/pi-agent-core/src/agent-loop.ts:51:3

Anyone seen this error when the agent receives a message?

[obtuse sundial]

is there any way to see exactly what the fuck my bot is doing at any given time perhaps live?

he disappears for 10 to 20 minutes at a time and i cant tell if hes actually doing anything or just hung up on something or being dumb

[hexed steeple]

But it does have admin privileges and sorry pee accidentally splashed on it

[nova solstice]

Thanks but I don't seem to see that media folder in ~/.openclaw/. Are you saying we need to create that folder and then put our image in there?

[obtuse sundial]

brother do NOT come in here and say you wasted that golden nectar

[hexed steeple]

I didn’t waist it man all I’m saying is some splashed onto my screen when I was in the bathroom. I have an abnormal heavy stream…I can’t help it

But for real my account says admin on it but that still pops up. Anyone know a work around.

?

[sharp summit]

Is anyone getting a ton of exec errors all of a sudden

[wraith leaf]

anyone know where to find working build of the mac app? mine crashes on launch

[obtuse sundial]

ask Claude. seriously.

[brave trench]

Can't get my Telegram to talk

[obtuse sundial]

no such thing as an openrouter/auto model

[crimson steeple]

Change your model

[obtuse sundial]

switch to a real model

openrouter/auto should be removed or fixed tbh

[crimson steeple]

Go on openrouter and find a model

Target it specifically

[blazing niche]

anyone seen this error before? i must be the only one with this in this Universe, google and this discord doesn't show anything

[brave trench]

Ahh.. When I select manual, it defaults to that during onboarding

[blazing niche]

i tried llama3.2:3b and llama3.3:latest

[coral yoke]

trying to get openclaw to run claude -p --dangerously-skip-permissions "who are you?" and its just hanging and the dumb bot can't figure it out

[brave trench]

Im using Open Router/ Moonshotai/kimi 2.5

[obtuse sundial]

then somewhere in your file its defaulting to openrouter/auto as seen from your post

you must find and correct all instances of this by directing it to the real model you are intending to use

[crimson steeple]

If the config cli isn't working, try editing the config file itself

[obtuse sundial]

this

[brave trench]

Ok. Would that be in a folder on the desktop

[obtuse sundial]

it would be in the clawdbot files where you are hosting them

[brave trench]

Ok let me look. Thanks!

[obtuse sundial]

if you cant figure that out yourself you would benefit from asking claude or chatgpt (probably claude would be better)

[wary epoch]

⚠️ Agent failed before reply: Unknown model: anthropic/claude-sonnet-4.
Logs: clawdbot logs --follow

Reset doesn’t do anything?

[crimson steeple]

You gotta change the model

[wary epoch]

What’s the comand for that?

[crimson steeple]

Type /models

Try openclaw config in the cli and try changing the config

[wary epoch]

And choose opus?

[brave trench]

What file should I look in

[crimson steeple]

Whatever you'd like

[crimson steeple]

openclaw.json

[brave trench]

Ok

[magic yacht]

yall i need help

[crimson steeple]

What's up

[magic yacht]

did yall download this through yalls terminal on MACos

[brave trench]

Should I delete where it says Auto

[wary epoch]

Says model not allowing?

[crimson steeple]

I'm on windows personally

[crimson steeple]

Change "primary" to a real model from openrouter

[magic yacht]

did you still have to go through your computers terminal for the downloar

[crimson steeple]

What's the full error?

[magic yacht]

download

[crimson steeple]

Yeah

[fair coral]

I migrated to openclaw, now just typing "openclaw" takes like 20 seconds to return the help. Any ideas?

[cunning cairn]

hello, how can i start?

[magic yacht]

cool ill get mine set up then through terminal thanks

[magic yacht]

once i have it set up

how does it work lol

[brave trench]

Like that?

[crimson steeple]

check the quickstart on this page
@magic yacht @cunning cairn

https://openclaw.ai/

[magic yacht]

cool

[crimson steeple]

check the quickstart docs too

Scroll down there's documentation

[crimson steeple]

that should do

restart your gateway with openclaw gateway restart

[brave trench]

Ok. Saving now. It was on Notepad, is that ok

[crimson steeple]

yeah

[brave trench]

Ok closing all windows and restarting my gateway

[wary epoch]

It’s says pondering and counter going up?

[crimson steeple]

Did you put in the anthropic api key

[wary epoch]

All was working fine today

[novel sparrow]

is openclaw by default supose to be more 'proactive' / 'always on'? I get the impression that it's mostly idle, is this normal? or is there some setting one needs to turn on ?

[crimson steeple]

openclaw config I recommend trying to go through setting the auths again

[crimson steeple]

Asking it to be proactive does genuinely work

[obtuse sundial]

i think it does take a lot of work to get it to be more proactive. im still struggling with this.

[wary epoch]

Command doesn’t do anything

[crimson steeple]

Any errors?

[signal dust]

Bot sure what happened but my TG connection crashed. Anyone know a command to set it back up /connect comm channels?

[brave trench]

Stoneman your patience is amazing 👏 Thank you

[crimson steeple]

Going into soul.md might help, try writing your own prompts

[obtuse sundial]

i already did that and added to it...and added to it..and more

and a bunch of others lol

[crimson steeple]

ah that's not the cli, you gotta do it in a fresh command prompt, that's part of your chat

[wary epoch]

So new terminal?

[crimson steeple]

I think the star of the show is heartbeats

[crimson steeple]

yep

[obtuse sundial]

mine is edited and runs every 30 minutes lmao

[crimson steeple]

So, what's your goal?

[wary epoch]

What’s the comand to run it in a new terminal?

[crimson steeple]

just run openclaw config in a new terminal

and then select model and go through that config again

[obtuse sundial]

i just wish my fucking bot would talk like normal and update while its working on a task instead of ignoring me for 20 minutes

[wary epoch]

OK create a new api key how do I copy the old one in athropic ?

[crimson steeple]

I've managed to get my bot to work like so, is this what you want?

[crimson steeple]

can you elaborate?

[wary epoch]

I don’t see the option to copy the key?

[glacial jay]

Where do you get the menu-bar companion on os x

[crimson steeple]

gotta make a new one and remove the old

[wary epoch]

Money will transfer over?

[crimson steeple]

I'm unsure.. I've never used anthropic's api, sorry

[obtuse sundial]

is this like it updating you as it does something or all in one message?

hard to tell im new to telegram

[crimson steeple]

Updating, multiple messages

[obtuse sundial]

yeah mine does NOT do that

is it depending on the task that it does that or just any task it updates you?

i think theres a setting for it but i couldnt find which setting

[crimson steeple]

everything, websearch and all

you don't set settings, it's prompts!

[crimson steeple]

I keep seeing people have that issue

idk how to fix that one myself

[obtuse sundial]

what model are you using?

[wary epoch]

Reinstall it all?

[obtuse sundial]

mine are all budget models so maybe thats why mine are a bit too dumb to get it

[wary epoch]

Will that delete all my work?

[brave trench]

Mine won't even acknowledge me lol

[crimson steeple]

you can backup your memory and config files

It should be within your user folder

[wise lotus]

well I fixed the context compacting amnesia, only for it to not be aggressive enough (manually delete session to resume), comes out of the box in a safe mode but damn, this project shows that it's only two months old which does lead me to question the viral nature of it

[crimson steeple]

No config?

[crimson steeple]

It's capable of alot if used right

[wise lotus]

hmmhmmm, "you're not holding it right"

[prime nimbus]

I'm still praying for this: https://github.com/ollama/ollama/issues/2821 looks like a WICKED model for spinning up a few "lazy" agents to do simple stuff microsoft really smashed it with this, and i'll never say them words again LOL

[crimson steeple]

But expecting too much is also wrong, I've given up on that much autonomy myself
You gotta expect it to fail, model changing manually, kill it when it's stuck etc

[wise lotus]

working to a basic standard of the box would be a start

[crimson steeple]

It's true, but this tool isn't really meant for general users yet, more of an experimental power user tool

I'm coping haha, it isn't stable at all

[wise lotus]

I definitely fall into this catagory without revealing more

[crimson steeple]

I'd agree with all your points, it really does function sometimes, otherwise rather flimsy

However when it does, it's damn impressive, early days I suppose

[wise lotus]

I don't disagree that it's an admirable first attempt to put something like this together, but I also don't understand at the same time, why it has so much buzz

[brisk canyon]

Hello guys, new user here. I tried checking the documentation but could not find openclaw bind docs.

Openclaw binds to loopback which I cannot access over the network. I changed it to 0?0.0.0 but it's not working also

[crimson steeple]

The buzz comes from the developer space, ones who got the patience to make this work

I wouldn't give this to the general user yet because it's actually a very dangerous tool

[crimson steeple]

Try going through onboard again, pressed manual?

[brisk canyon]

No

Thanks. I will try onboarding again

[fickle flint]

Hi, does anyone know the best method to migrate from Clawdbot/Moltbot to OpenClaw?

[crimson steeple]

the best way is to genuinely ask your bot to do it

[fickle flint]

Good point

[crimson steeple]

They know their directories better

so many people said it has worked for them, try it!

[fickle flint]

Going to try now, thanks!

[brisk canyon]

Hmmm... This is still startup I can't even use it yet. I am installing remotely on Ubuntu bare metal no monitor

Thanks guys. I will keep troubleshooting

[obtuse sundial]

what did you prompt your bot with to get it to update you regularly like that?

[crimson steeple]

I'm not on my computer right now, so I can't open the prompts to show you
But I asked it to be transparent

[analog saddle]

We're pushing an alias for this on the OR side momentarily 🙂 For now, a workaround is editing your clawconfig.json file to be ⁨openrouter/openrouter/auto⁩ instead of ⁨openrouter/auto⁩ 🙂

[crimson steeple]

i wonder.. i want to know whenever you're doing something, like when you do a tool run, or a new skill run.. its so interesting to see other ai agents work, like in anti gravity for example, it shows what tool what commandline etc i wonder.. are you able to visualize that here every time? or is that too much... like every thing you do with tools like "web searching: cat videos" etc or "saving memory..." you would show that in the chat?
This is what I asked it

[obtuse sundial]

what does this actually do? just swap models whenever or something?

[crimson steeple]

Ooh, an OR rep? Hello! I hope this app has been a big hit on your platform!

[obtuse sundial]

thanks for the tip i will try that and see if mine works!

[analog saddle]

Hi, yes I'm an eng on the OR team 🙂 We're pumped!!

[analog saddle]

Sorta kinda! Read about it here: https://openrouter.ai/docs/guides/routing/routers/auto-router

[obtuse sundial]

I'm currently using a model switching tool i had claude make for open router which is really just a modified version of model-router

[crimson steeple]

Have the OR team been using this internally too?

[obtuse sundial]

it swaps between deepseek, devastral and grok. deepseek for general, dev for coding, grok for checking the coding

[analog saddle]

We literally have been on a huddle for the past 3 hours building one together on cloudflare yep haha

[analog saddle]

That's dope

[obtuse sundial]

so far it CAN switch but getting it to do it consistently is somewhat daunting

[crimson steeple]

That moltworker thingy? Cloudflare really stepped up

[empty owl]

Do I need to completely reinstall if I'm getting This version of Antigravity is no longer supported. Please update to receive the latest features!?

[crimson steeple]

Well, good luck OR, you rock

[obtuse sundial]

"message{"action":"send","to":"8363805258","message":"Got it - more human updates, show what I'm doing. Let me try that style now."}" is my bot just fucking with me or is he an idiot?

[analog saddle]

exactly!

[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",

[crimson steeple]

I've been personally on discord myself, my bot talks on there with no issues

[empty owl]

I'm on linux though

[crimson steeple]

Perhaps something about the messenger implementation

[crimson steeple]

Gotta go into your install of the bot on your linux then

Same solution!

[obtuse sundial]

jesus man this thing is gonna cost me $5 a day even with budget models fuckin hell

[crimson steeple]

Find your npm location and find where the code is stored

[blazing niche]

my bot replies in discord to everything question/mention: 400 "llama3.2:3b" does not support thinking - anyone know why? in console tui it works but takges ages like 30-60 seconds to respond

[obtuse sundial]

i really am gonna need to get a sub somewhere even if they ban my ass

[crimson steeple]

I'd pay for Gemini Pro (the subscription) and run it on antigravity

[obtuse sundial]

i simply cannot afford openrouter event hough i thought itd be the cheaper option

unless i use one of the shitty models i guess

though i think mine are not the best off the bat

[analog saddle]

Just pub'd OR docs for setting up OpenClaw! https://openrouter.ai/docs/guides/guides/openclaw-integration

[crimson steeple]

Oh my, you're the bes5

[obtuse sundial]

not like im going for opus or something