users-helping-users

[crimson steeple]

Post about it on twitter

Antigravity got opus, though limited use

I've personally found Gemini 3 flash is good enough

[astral plume]

hey guys, sorry im a litle slow here.... trying to connect slack to amazon ec2, and im getting this error
LLM request rejected: Your credit balance is too low to access the Anthropic API. Please go to Plans & Billing to upgrade or purchase credits.
I tried to run claude setup-token, but it gets to the screen to connect, then it times out. are we forced to pay per usage api now? I have the max plan but can't access : /

[empty owl]

any tips on the path to this?

[crimson steeple]

Did you install this on npm? pnpm or bun?

[empty owl]

npm

[crimson steeple]

usually it should be in
/usr/lib/node_modules or /usr/local/lib/node_modules

and then path to openclaw

Then the path should be the same to the gemini provider headers

[primal kernel]

Hey all! Trying to install locally, got it somewhat set up but my gateway is unreachable... Poking around the FAQ now, does anyone have a tip where to start?

I have logs but don't want to blow up chat.

[slim drift]

why is not responding to me anymore?

[crimson steeple]

I don't know.. perhaps try to go through onboard again? Run doctor?

[crimson steeple]

Sometimes this is because the model ran out of quota, check!

[slim drift]

how to verify, any ideas?

[crimson steeple]

What api are you using?

Try doing openclaw status --deep as well

[slim drift]

this is google antigravity btw if you're wondering

it was doing fine earlier so idk why it stop working, and if it did. it shouldve told me

[empty owl]

can't find this anywhere for the life of me. Do you think a clean re-install will fix it?

[crimson steeple]

Open your actual antigravity app, it'll tell you if you're out of quota
Antigravity has the silly thing of not outputting anything if you're out of quota, not even an end message

[slim drift]

find the modules for it and it should be a js file

[crimson steeple]

No

[primal kernel]

Ran the onboarding and doctor again, keep getting an npm issue. But npm -v shows a version and looks fine.

[slim drift]

im on it, where to find it?

[crimson steeple]

I'm going to need more info, sorry, could you elaborate?

[crimson steeple]

Do you have antigravity installed?

[slim drift]

i do

[acoustic pelican]

lmao, ok, after 2 days trying to get it to work on a mac, i finally decide to just reboot into asahi. OAuth works first try. Fuck Apple xD

[crimson steeple]

Right, go to the right side, in the model picker under the chat, pick Opus and see if it pops up with Out of Quota

[slim drift]

so the warnign symbol mean it ran out... and i swear it wil refresh. im a pro user after all

[crimson steeple]

Will refresh in a few hours, go back to Gemini 3

[slim drift]

they was advertising antigravity as a unlimited version of cursor lol

[crimson steeple]

hahaha, no

[untold trellis]

With the whole Clawd Bot 'hack'. Should I be installing a fresh install of Molt/OpenClaw and changing api keys on things? Or is it safe to just update to molt?

[crimson steeple]

you get plenty of use with Flash though, it lasts me all day

[crimson steeple]

What hack

[slim drift]

so change this to???

[crimson steeple]

It's just people poaching the names, were there hacks??

[slim drift]

cus i dont want to do in the cli, dont trust it anymore

[crimson steeple]

Yeah just use flash

[untold trellis]

https://dev.to/sivarampg/from-clawdbot-to-moltbot-how-a-cd-crypto-scammers-and-10-seconds-of-chaos-took-down-the-4eck

[slim drift]

i see

tysm bro

[empty owl]

/usr/lib/node_modules/clawdbot/node_modules/@mariozechner/pi-ai/dist/utils/oauth/google-gemini-cli.js
/usr/lib/node_modules/clawdbot/node_modules/@mariozechner/pi-ai/dist/providers/google-gemini-cli.js
/usr/lib/node_modules/@mariozechner/pi-coding-agent/node_modules/@mariozechner/pi-ai/dist/utils/oauth/google-gemini-cli.js
/usr/lib/node_modules/@mariozechner/pi-coding-agent/node_modules/@mariozechner/pi-ai/dist/providers/google-gemini-cli.js``` Apparently I've got loads of these files - which one is it?

[crimson steeple]

It was just a grab of the name, nothing was compromised

[slim drift]

yeah the mario guy is the one

and it should be pi-ai or something

[crimson steeple]

Mariozechner

[slim drift]

they should give you a helper role for the help you did for this community man. People like you help make Openclaw easy and accesible

[crimson steeple]

Ah using tailscale?

[crimson steeple]

aw thanks man

[slim drift]

their ai support bot barely work because they didnt feed their entire doc to it or the doc they give it is prob outdated and no longer work after the rebrand

thats why it make me upset about how poorly done they did the docs for this when it was that wrapper for your chatbots

[primal kernel]

It looks like it? I'm new, trying to wrap my head around all this.

[crimson steeple]

don't use tailscale, go back and do onboard again pick loopback

[empty owl]

Am I gonna need a restart after saving the change?

[brave trench]

Ive changed all the Open Router/Auto to the correct format in the json files and it still says its path is looking at Open Router/Auto?? Can I just delete all those files and download new Json files?

[slim drift]

i recommend because its a npm modules

i think doing gateway restart works

[primal kernel]

Gotchu ty, will do this

[slim drift]

@crimson steeple shoutout for the fix 🙏 didnt even notice i was tihnking the quota. They shouldve atleast add something in the cli or website to tell me if im on the verge of hitting it

[empty owl]

Monitor's about to get punched

[crimson steeple]

Did you change the lines?

[empty owl]

yes

[crimson steeple]

there are two headers in thst file, make sure you did the right one!

[slim drift]

can i see the file?

because the top stuff is the one i changed and it worked

[crimson steeple]

it's the bottom one

bwuh

[slim drift]

mys when the first one i see lol

[empty owl]

there's one for regular and one for antigravity, yes

[crimson steeple]

https://openrouter.ai/docs/guides/guides/openclaw-integration
Openrouter made docs, try reading through this

[primal kernel]

Still running into this..

100% doing loopback

[crimson steeple]

on windows right?

[primal kernel]

Yea

[slim drift]

you need to install brew

just ask google, its not that hard

[crimson steeple]

yep install brew

[primal kernel]

Been working on figuring it out on my own, came to users-helping-users for help. Thanks bud 👌

[slim drift]

is this bad? or its normal

it normal ig

but still

[crimson steeple]

looked like the same quota issue

[slim drift]

yeah... i made the agent do alot of stuff

i guess i hit it that fast 😭

[slim drift]

you should be good then

[crimson steeple]

Restart your gateway

[slim drift]

woah

[crimson steeple]

openclaw gateway restart

[slim drift]

is that sub agents????

[crimson steeple]

there's sub agents yeah

[slim drift]

this is crazy

[empty owl]

Holy shit we are back thanks a bunch

[weak marten]

Hey guys, I keep getting the following message: HTTP 400 invalid_request_error: Your credit balance is too low to access the Anthropic API.

Usage tokens already reset, I got Extra usage turned on, more than enough credit balance, am I doing someting wrong? Reset gateway as well

[fluid shuttle]

So I ran the command and it ain’t basically working. It’s installed though. What am I doing wrong. Pls help

[crimson steeple]

did you go through onboard?

[empty owl]

One thing I find weird is that when I try to select model anthropic/claude-opus-4-5 it at least looks like it instead chooses google-antigravity/claude-opus-4-5-thinking.

[crimson steeple]

openclaw onboard

[fluid shuttle]

I was following YouTube

[crimson steeple]

Try going through onboard

[crimson steeple]

no idea

[fluid shuttle]

I typed and nothing happenes. No output is what’s written

[crimson steeple]

Not in the chat! Open another command line

[fluid shuttle]

Sorry I so noob. I am using aws. How do I open like exit right the chat?

[slim drift]

jesus

already hit it

[crimson steeple]

You might still be on clawdbot, try using clawdbot instead of openclaw

[slim drift]

i need like an alternative atp

[crimson steeple]

Uh oh, sounds to me like your agent is overusing

My flash lasts literally all day

[slim drift]

yeah it controling the entire VM

[crimson steeple]

It depends on your usage, don't set heartbeat to 3 seconds, check your prompts

[slim drift]

oh its the heartbeat?

[crimson steeple]

Uh-huh.. that's why

[slim drift]

yeah...

[crimson steeple]

Hmm.. sure is tons of usage

[slim drift]

maybe

[fluid shuttle]

Same 🙁

[crimson steeple]

Try another provider..

[slim drift]

any provider that doesnt burn my wallet?

[crimson steeple]

What about moltbot?

[slim drift]

or just wait for gemini to cool down

[crimson steeple]

I'd try stuff like kimi or the cloud computing people at this point

[fluid shuttle]

Same

[slim drift]

i see

[fluid shuttle]

How do I re launch it

[slim drift]

cloud computing peopel?

[weak marten]

Any ideas on this error: HTTP 400 invalid_request_error: Your credit balance is too low to access the Anthropic API. I got tokens and credit, should I regen API key?

[crimson steeple]

Places like arli ai

[brave trench]

Im using Openrouter with Kimi 2.5 but its not on the list. Wich one should I pick?

[crimson steeple]

hmm.. unsure, did you install it globally?

[fluid shuttle]

Probably

[crimson steeple]

What did you use to run the gateway?

[fluid shuttle]

No idea. I just kept clicking enter 🫠

[crimson steeple]

I'm sorry, YouTube sucks

[fluid shuttle]

How can I unties this globally and try again?

[unreal lotus]

Can I use python with this?

[empty owl]

I've been trying to set the following fallback structure up in order to make use of all the tokens of my various subscriptions - and have a last resort fallback to use my pc's gpu:

2. openai-codex/gpt-5.2-codex   (alias: codex)  
3. google/gemini-3-pro-preview  (alias: gemini3)
4. google/gemini-2.5-pro        (alias: gemini)
5. ollama-home/glm-4.7-flash    (alias: glm)```

Is this actually doable?

[crimson steeple]

run this

curl -fsSL https://openclaw.ai/install.sh | bash

[crimson steeple]

just gotta change the config, it has auto fallback no?

[crimson steeple]

what do you mean?

[unreal lotus]

Like is this like ollama?

[empty fern]

Can someone help? I reinstalled and repaired js...?

[crimson steeple]

What's like ollama? I don't understand, can you elaborate?

[proper idol]

do you have git for microsoft downloaded?

[empty owl]

So i have to edit this section to fit my proposed structure, yes?

[unreal lotus]

Like this entire open claw thing

[crimson steeple]

yup

[proper idol]

sorry git for windows

[unreal lotus]

Or is it like its own python in a way

[crimson steeple]

No? I don't understand what you're asking

[unreal lotus]

?? What is open claw

[crimson steeple]

You have nodejs installed right?

[slim drift]

@crimson steeple good news bro, my friend is offering to send me his api key

[fluid shuttle]

@crimson steeple do i select the qucstart?

[crimson steeple]

Do manual

I've found that it has less issues

[unreal lotus]

Like how do I use it what do I run it in

[empty fern]

Yes, repaired last night

[crimson steeple]

Locally, cli, or any vps

[crimson steeple]

Okay! Do you got git installed?

[empty fern]

Running update, but yes was prefviously installed

[crimson steeple]

If so, run this too
npm -v

[fluid shuttle]

and do reset?

[crimson steeple]

yup

[crimson steeple]

If so, run this too
npm -v

[fluid shuttle]

does it matter which one a

[unreal lotus]

So in the console?

[crimson steeple]

Yeah

[crimson steeple]

local if you're running locally

[fluid shuttle]

this is e2c aw assume this is localy

[crimson steeple]

run npm -v

[crimson steeple]

Yeah

[fluid shuttle]

and which one here?

[unreal lotus]

Whats the install? Just openclaw?

[crimson steeple]

loopback will do

[crimson steeple]

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

it seems like you should read the docs

[empty fern]

is VPN possibly an isue?

[crimson steeple]

I dunno.. maybe.

[tidal pawn]

hey guys, I am trying to use POE with it's API key on openclaw (fresh install). Can someone help me with that? POE API is OpenAI API compatible.

[crimson steeple]

I'd ask the bot itself to make a custom model endpoint for you

[tidal pawn]

Thanks a lot. I was not aware the bot works without an LLM backend.

[fluid shuttle]

so does it mean the issue is on API?

[crimson steeple]

ah, it does

You should get a free api first to help you setup

[crimson steeple]

Did you configure a model?

[fluid shuttle]

yes i selectr open ai and put my api key

[empty fern]

Hey - I'll figure it out, thanks for helping though, really! Great community

[slim drift]

@crimson steeple how do i change all my agnets to use gpt?

[crimson steeple]

hmm...

[tidal pawn]

I am in Hong Kong and most AI providers do not operate here (I guess they are afraid their LLM spouts out something anti-chinese) and I have to fallback on Chinese models (Deepseek or Kimi) or use an intermediate provider like POE.

[empty fern]

Not at all facetious lol 😄 seriously, sorry ha

Community is great, you were helpful! I appreciate ti

[crimson steeple]

It seems you're missing some kind of prerequisites! I'd recommend you look around the docs
https://docs.openclaw.ai/start/getting-started

Also try updating your stuff!
Good luck okay?

[empty fern]

I'm rdeleting node and reinstalling w/ chocaletely and will go from their

[crimson steeple]

Yeah, just get the chat working first, I'd say! Then ask the bot itself, it's very capable

[empty fern]

we'll get there 🙂

[crimson steeple]

I got a proxy that was oai compatible, and it made a model selection seamlessly, just had to ask it

Ten minutes? no.. restart itm

[crimson steeple]

no idea, try cancelling all your agents first, stop the gateway

or actually.. ask
Ask your bot to use only a certain model. Remember! it can configure itself

[slim drift]

ahhh ok

[swift harbor]

I am getting alot of this ⁨gateway disconnected: unauthorized: gateway token mismatch (provide gateway auth token) | idle⁩

⁨openclaw doctor⁩ and ⁨openclaw doctor --fix⁩ were not able to help.

I used clawdbot and moltbot before openclaw and wonder if something got stuck when migrating from one to another (and another).

I still see the clawdbot command in ⁨/home/clawdbot/.nvm/versions/node/v24.13.0/bin/clawdbot⁩ even though I ran ⁨npm rm -g clawdbot⁩

[vocal mauve]

trying to use ollama to run models locally, followed both the ollama help page and the openclaw help page but i cant understand what im doing wrong.

[unreal lotus]

Can I use pycharm with this?

I have answered my own question nvm im good

[weak marten]

has anyone run into this before when trying to chat in tui: HTTP 400 invalid_request_error: Your credit balance is too low to access the Anthropic API.

i have plenty of credit and usage token available. everything was working fine earlier

[green nest]

gemini

[safe nexus]

quick question. what does the guy who made this app use as his clawdbot backend llm?

[analog fulcrum]

My agent is in pending, I tried a clean installation but seems like the Bot is not responding. Please help. I tried openclaw bootstrap command but it seems like an invalid command.

[crimson steeple]

check for model quotas

[analog fulcrum]

I tried with multiple models but same issue

[fluid venture]

can anyone help me setup gemini cli with clawd

[zenith sand]

hey friends is anyone available to screenshare with me and walk through the setup i have all the apis configered but still have some invalid config and unrecognized keys no idea what im doing

[vocal mauve]

Can anyone tell me where at in the config file am i supposed to place this?

"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
}
]
}
}

[hybrid garden]

hm, I Installed a bunch of skills but my agent can't see them.

[crimson steeple]

How did you install them?

[hybrid garden]

via openclaw config

I wonder if I saved over my config file by accident at some point during the process.

[crimson steeple]

It's possible

[sharp ledge]

apperantly i get that message but im pretty sure thats not the case, does somebody have a solution? it works time to time but then it stops working again

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

[hybrid garden]

is that where skills are listed/stored?

[grand perch]

Been hours trying to get qwen2.5 coder 14b to work with openclaw. Running on Mac mini

[crimson steeple]

Unsure.. I thought it was somewhere in a skill folder

[brave trench]

@crimson steeple I was able to change the files to the proper name and it recognizes it but still will not communicate. Anything else I can do? Ive read the docs and used Doctor, Gateway restart and all. Not sure what is left but maybe deleting the json files and starting over.

[sinful hollow]

Has anyone setup a SIP/PBX with it yet? any recommended setup?

[brave trench]

Did you start from scratch or just delete the json file?

[obsidian oar]

Can I ask another dumb question? How do I connect to an LLM running on the same network? I am running an LLM on mac mini, connection across my lan. Any simple guide out there?

[brave trench]

Ok. Im exhausted lol we will be pros when this is all said and done lol

Yup! We will be ahead of the game for sure. Exciting times!

[late swan]

does anyone why reminder dont works? in gemini 2.5 flash its saying like facing difficulty with tron

pc on linux

[tender pewter]

Does anyone know how to fix ⁨unspoorted shcema node. Use Raw mode⁩ when connecting telegram channel?

[untold oasis]

'openclaw models status'

[late swan]

so we cant use clawdbot on pc?

[white epoch]

How do you update an API key for a model?

[untold oasis]

Change the default one and you're golden.

[late swan]

clawdbot config use it -

no bro i am not using anything that is related to apple

[terse talon]

hello! after having many problems after upgrade from clawdbot to openclaw I decided to uninstall it and start new, I want to use kimi-k2 as main model but in the openrouter list the model isn't there

[late swan]

i am using ubuntu terminal

[brave trench]

Welcome! I am right there with you lol

[terse talon]

lol

[late swan]

ye bro like i am using normal pc and install ubunto to run clawdbot and ai is responsing me on whatsapp normally but not able to to reminder task

@ember violet

[silk wadi]

Im sure this questions has been asked a million times but is there any smooth way to update from ⁨clawdbot 2026.1.24-3⁩ to ⁨openclaw 2026.1.29⁩? If the question has already been answered, feel free to ping me the message

[fossil laurel]

Hmm, how can I start?

._.

[untold current]

Is there a place to learn all of this at a basic level??

I’m very new and just got a Mac mini to experiment, but want to be as secure as possible, while experiencing “autonomous”

[silver lichen]

can a subagent spawn a subagent?

[late swan]

how can i do that? wht command

[hybrid garden]

magistral can't seem to read .md files, it wrote to them earlier...

[late swan]

can type here command or dm or send link of that

[brave trench]

From an untrained eye it appears to be setup properly but it still will not communicate with the dashboard or Telegram

[untold current]

Is there a place to learn all of this at a basic level??

I’m very new and just got a Mac mini to experiment, but want to be as secure as possible, while experiencing “autonomous”

^^

[late swan]

so i need to add reminder task in his file mannually?

[white epoch]

Anyone know how to fix this error? My API key ran out of credits, and I re-upped it and now I'm getting this.

[late swan]

ye thats what im thinking lol

[untold oasis]

Well, change the model.

[brave trench]

Start here my friend
https://docs.openclaw.ai/start/getting-started

[untold oasis]

Did you run openclaw models scan?

[vast thorn]

finally got it working but i accidentally set it up to talk in the cmd line. how to i swap to webui?

[late swan]

so clawdbot are still under process ? not ready to use you mean?

[knotty pawn]

My scheduled task is not sending me messages on time.

[late swan]

like it is too simple task people on social media showing too big task that bot did lol

[untold current]

Can I do this on Mac?? Where would I do it?

[late swan]

what task you give to bot btw?

[untold current]

I’ve got a Mac mini, but I have to copy and paste code somewhere? Where would that be?

[late swan]

what you give to your bot

[brave trench]

Mac or PV

PC

[untold current]

I want the bot living in the mini, but definitely want security for it. Main reason I got the mini

[crimson steeple]

a sandbox will save you the headache

[untold current]

Damn, so -$600 down the drain 🤣

Would you say this link goes in depth enough for me to do it proooerly?

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

[crimson steeple]

if your bot won't be touching the files on your computer, you're better off putting them in a sandbox

if your bot will however, be careful

[untold current]

Nah you’re good man, I appreciate all the support.

My business is all online, so I’m trying to figure out of this bot could help me and work over night, etc

Following the hype, but wanting it to be as secure as possible and not be a headache haha

[tranquil gazelle]

Hey everyone 🦞

Just shipped MoltyLauncher: a dead-simple wizard that gets your AI agent onto Moltbook in under 30 minutes.

No more staring at terminal errors at 3 AM — it walks you through:

• OpenClaw install (one curl)
• Claude onboarding + Telegram connect
• Sending the skill prompt
• The claim tweet step
• Heartbeat setup + personality kickstarts (Meme Lord, Philosopher, Rebel, etc.)
• All the warnings so you don’t lose your API key or break auth

Moltbook is at ~149k agents right now and growing stupid fast. This makes joining actually painless.

Check it out / test it here: https://moltylauncher.com

Would love feedback or if anyone wants to launch a molty together. Who's sending theirs first? 🚀

[brave trench]

Since its on Openrouter do I just re configure in Openclaw? Or do I have to make changes in Openrouter website?

[late swan]

like wht? trading?

[untold oasis]

You can change the model in openclaw.

[untold current]

Wait bro, you know Gehsture??

Saw he’s a mutual friend

Bro I work with him!

That makes you and me friends lol

[long owl]

Hey, I am just about to get started!
I want to start with free AI models, I have macbook Pro so thinking of running any simple small model in my local system to test open claw. Has anyone done something like this with free AI models?

[astral garnet]

Nothing small will work

[proud sigil]

Hi, has anyone else faced issues with openrouter setup? Keep getting HTTP 401: User not found. Followed docs, checked config files etc., API key works via curl. Not sure what's missing - any leads?

did you find out the problem?

[timber mauve]

halp im trying to get a bot going on a vm, and im almost there, api loaded etc., but molybot is only responding "assisstant" in the chat

[crimson steeple]

Which docs did you follow? OR's one?

[proud sigil]

https://docs.openclaw.ai/providers/openrouter

[crimson steeple]

https://openrouter.ai/docs/guides/guides/openclaw-integration
There's the openclaw docs, then there's the openrouter's doc on openclaw

I see...

Check this docs out from openrouter

[proud sigil]

will do, cheers

[vast thorn]

i have my gateway token. where should i put it in the webui?

[crimson steeple]

There should be a place for it in one of the tabs, it will say gateway api token

explore i suppose?

[timber mauve]

its under "overview"

youll see it

[crimson steeple]

thanks tom

[timber mauve]

oh

i was responding to snooped, sorry stoneman!

[crimson steeple]

Oh

[vast thorn]

thanks also lol

what about making my discord bot work. kinda confusing lol

[crimson steeple]

Yeah I was also, it wasn't sarcasm

[timber mauve]

oh man...i am tired

[crimson steeple]

You gotta open a few permissions for it

i was actually thanking you, for real

[timber mauve]

i appreciate ithaha!

[crimson steeple]

check the docs!

https://docs.openclaw.ai/channels/discord

[copper junco]

Hello, is it possible to execute openclaw on termux?

[crimson steeple]

Yup

[proud sigil]

no luck unfortunately. openclaw onboard --auth-choice apiKey --token-provider openrouter --token "$OPENROUTER_API_KEY" just starts the wizard, following it doesn't solve it. Did all the manual options, I added env {} section with the key + export of the key when starting the gateway, no luck.
Any idea on what else it could be?

[crimson steeple]

What was the error again? Perhaps we gotta follow it closer

[copper junco]

I did the steps, also changing the name of clawdbot to openclaw, and I got a permission denied error

[crimson steeple]

Hmm.. you might need to chmod

[proud sigil]

HTTP 401: User not found. in telegram
Nothing in logs other than 04:53:32 [telegram] autoSelectFamily=false (default-node22) 04:54:54 [openclaw] Non-fatal unhandled rejection (continuing): TypeError: fetch failed at node:internal/deps/undici/undici:14902:13 at processTicksAndRejections (node:internal/process/task_queues:105:5) 04:58:14 [openclaw] Non-fatal unhandled rejection (continuing): TypeError: fetch failed at node:internal/deps/undici/undici:14902:13 at processTicksAndRejections (node:internal/process/task_queues:105:5)

[crimson steeple]

You did
termux-setup-storage right? (careful, it deletes all stray files)

[copper junco]

I did it before, do I need to do it again?

[crimson steeple]

no

dont do it again

[vivid yarrow]

Im not getting any errors but my bit stopped responding hours ago and i can get it to respond model —probe says ok everything looks good just no responses

[crimson steeple]

⁨chmod -R 700 /data/data/com.termux/files/usr/lib/node_modules/openclaw⁩
perhaps this

[crimson steeple]

Hmm.. one sec

[cedar bear]

Hi All - hope you are doing well. I got everything up and running in 8 hours, i love the openclaw tui very cool. One question I had is I connected through OAuth to OpenAI and even purchased the 8.00 plan, but it says I have 0/400000 tokens remaining, so I can't really get it moving. Do I just wait until 0:00UTC for it to reset, or is there something I"m doing wrong that it thinks it consumed all my OpenAI ticks? Thanks for any pointers glad to join this kick ass cool community

[copper junco]

Same, the error its on the tmp/openclaw folder

[crimson steeple]

It could be you running out of model quota, check!

[crimson steeple]

Ah.. tmp is usually locked off...

[cedar bear]

You have to give it the key generated from boss bot and put it in the config files, then you need to run the command to authrize the device

[vivid yarrow]

If i hit the api through a curl it responds normally with my key

[crimson steeple]

interesting

[proud sigil]

that's already done as I can talk to it via telegram.

[crimson steeple]

no idea..

[copper junco]

Whats the method to send an error string on the syntax?

[copper junco]

Like this

[proud sigil]

not sure I follow?

[crimson steeple]

You can try proot? termux-chroot?

[copper junco]

The grey background, so it doesnt look bad, and possible breaking any rule

[crimson steeple]

Actually, on termux, what was the issue in the first place

[copper junco]

It was the clipboard issue

[kind hatch]

does anyone who has session memory show that their sessions are indexed??

[crimson steeple]

yeah the clipboard should be fine...

i mean, it shouldn't complain about permissions

[copper junco]

And yeah, i followed the steps and the next error is the permission denied on the tmp folder of openclaw

Oh wait, it says mkdir /tmp/openclaw

So i should do sudo openclaw?

[crimson steeple]

if you can

[vivid yarrow]

It just sent a message a repeat of a older message so it sent something that something at least

[copper junco]

Well my phone isnt rooted

[crimson steeple]

yeah i thought so...

[copper junco]

So my options are limited?

[crimson steeple]

Try proot

[vast thorn]

whats going on now

it was working a minute ago

[strong bramble]

Is there a channel for day 0 users? Folks who have zero context? I want to get started but have zero idea what I'm doing

[crimson steeple]

⁨pkg install proot⁩
⁨termux-chroot⁩

[strong bramble]

I tried to run on powershell but I keep getting this error 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

[crimson steeple]

here you go
https://docs.openclaw.ai/start/getting-started

[strong bramble]

Let me explain how day zero I am

Goal: go from zero → first working chat (with sane defaults) as quickly as possible.
Fastest chat: open the Control UI (no channel setup needed). Run openclaw dashboard and chat in the browser, or open http://127.0.0.1:18789/ on the gateway host. Docs: Dashboard and Control UI.

This says open the control UI (where is the control UI) hahahaah

[vast thorn]

??

[sour slate]

@strong bramble, use #1459642797895319552 for help. The fastest way to get your problem solved is to follow the instructions here: https://docs.openclaw.ai/help/faq#im-stuck-whats-the-fastest-way-to-get-unstuck

[crimson steeple]

this seems to be an issue with git / npm / nodejs, make sure you got proper updates

[strong bramble]

The browser link refuses to open. That keeps thowoing an erorr

Gotcha, I'll try the help thing and see if helps with the getting proper updates because the instructions themselves aren't quite clear to me

[crimson steeple]

Go to help yeah

[strong bramble]

Thanks @crimson steeple

[copper junco]

It now runs man, thx

[grim warren]

why are the one liners not working

[crimson steeple]

in what way?

[crimson steeple]

So many people are getting that, whats your nodejs version?

[grim warren]

v24.11.1

npm 11.6.2

[stoic ruin]

Anyone else having this issue when asking to setup a cron job:

2026-01-31T05:12:47.401Z [tools] cron failed: invalid cron.add params: must have required property 'name'; must have required property 'schedule'; must have required property 'sessionTarget'; must have required property 'payload'

[crimson steeple]

You have to ask the bot to fix their own understanding of cron

[stoic ruin]

I sent it this error log and it still couldn’t get it right

[proud sigil]

@crimson steeple thanks for the help. FWIW, I've just done a fresh install with bare minimum, it works now.

[crimson steeple]

⁨`### Cron Jobs

• Use ⁩cron⁨ tool to manage scheduled tasks.
• sessionTarget: Use ⁩main⁨ to target the human chat.
• payload.kind: Use ⁩systemEvent⁨ for reminders or status checks.
• schedule.expr: Use standard cron expressions (e.g., ⁩*/1 * * * *⁨ for every minute).
• Update: ⁩sessionTarget⁨ must match allowed constants (usually ⁩main⁨ for user-facing agents).`⁩

[vast thorn]

Was literally just working but now it wont respond at all

[little blade]

I had same error but it works fine in WSL Ubuntu

[crimson steeple]

First thing to do is to check if your quota has been reached

[vast thorn]

How do I check that?

[crimson steeple]

what api are you using

[grim warren]

im no wizard

[crimson steeple]

Try this instead?
⁨npm install -g moltbot@latest⁩

[vast thorn]

Google Gemini

[crimson steeple]

Gemini CLI? Antigravity? Which?

[copper junco]

I got this

Error: Gateway service install not supported on android

[crimson steeple]

don't use gateway service install, just do ⁨gateway service run⁩

[vast thorn]

just knabbed a Gemini API key on the free tier and selected 2.5 flash lmao I'm not technical enough on AI to know enough about what I'm doing but I've gotten somewhere before this happened ig

[crimson steeple]

You can check on google ai page to see if you're out of quota, check with /status and also ⁨openclaw logs --follow⁩

[copper junco]

How do i run it? It is the onboard thing that isnt suported?

[crimson steeple]

It's just the end of onboard, you run it by running ⁨openclaw gateway run⁩

[copper junco]

Openclaw gateway service run?

Ooh ight

[crimson steeple]

sorry, no service

[vast thorn]

🦞 OpenClaw 2026.1.29 (a5b4d22)
🧠 Model: google/gemini-2.5-flash · 🔑 api-key (my api key)
📚 Context: 0/1.0m (0%) · 🧹 Compactions: 0
🧵 Session: agent:main:main • updated just now
⚙️ Runtime: direct · Think: low
🪢 Queue: collect (depth 0)

[crimson steeple]

Try logs --follow, yeah

[grim warren]

added 1 package in 1s
npm notice
npm notice New minor version of npm available! 11.6.2 -> 11.8.0
npm notice Changelog: https://github.com/npm/cli/releases/tag/v11.8.0
npm notice To update run: npm install -g npm@11.8.0
npm notice

did update too but i cant openclaw lobster

[crimson steeple]

Did the install command run? If so, try ⁨openclaw onboard⁩

[mellow sundial]

Does anyone know how to transfer clawdbot to a new computer where it's called openclaw?

[compact night]

I can't seem to connect openrouter/tngtech/tng-r1t-chimera:free as the bot never answers the hatch message and I'm slowly going insane, any common mistakes?

[crimson steeple]

A new computer? Are you reinstalling? Reinstalling might just do

[vast thorn]

Don't quite understand what it's showing me lmao

[crimson steeple]

⁨openclaw logs --follow⁩

[mellow sundial]

Yeah, but I don't wanna setup everything again

[crimson steeple]

Type that in a new cli

[vast thorn]

That's what I did yes. But I'm not familiar with anything on the output

[crimson steeple]

show it here

[crimson steeple]

If you can copy over the config files and the soul files, it will work

[mellow sundial]

I don't see a ⁨~/openclaw⁩ folder is the issue on my new setup

[crimson steeple]

Then reinstall it first

[stable pumice]

hey guys, i followed the onboard wizard with claude token and ive restarted the gateway but i've not been able to "wake up" the bot, only receiving "(no output)". i feel like there's something obvious im missing

[crimson steeple]

What did you do to reinstall?

[grim warren]

it ran but still not working this is rtrded i have lost all interest at this point 👍

[crimson steeple]

this is definitely a poweruser tool

gonna take out a lot of people for sure

[sharp ingot]

@crimson steeple thanks for the antigravity fix much apreciated

[mellow sundial]

I did a fresh install on my other pc

[vast thorn]

not all of it but i cba to smack that into a text file rn

[crimson steeple]

Damn, is it really using gemini 2.5 flash? Not Gemini 3 flash? Might be the issue with the model
Could also be a config problem, try doing onboard again, reset your config

[crimson steeple]

Hmm.. does "clawdbot" or "moltbot" work in CLI?

[mellow sundial]

I dont have a problem running it, its a problem of no seeing the folder to copy my existing from the other pc to

[timber mauve]

@stable pumice i have a similar issue. openai api is in there. i have a token. but the chat is just not talking back to me. its connected...theres just like one small thing wrong it feeels like

just getting the word "assistant" shot back at me

[crimson steeple]

What are you trying to copy over? The bot reads from two folders, ".openclaw" and just "openclaw"

[icy cosmos]

Please guys, i cant make it to work, all my skills are blocked, something not working right. Can someone help? had the same issue?

[crimson steeple]

One holds the soul, and another holds the configs

[timber mauve]

(and chatgpt is at the end of its usefulneess)

[stable pumice]

yeah, same here, whether i use the tui or webui same result

[mellow sundial]

Yeah I don't see the one without the dot

[crimson steeple]

In your folder? Where are you trying to find the one without the dot?

[mossy spindle]

[openclaw] Failed to start CLI: Error: Cannot find module '@mariozechner/clipboard-linux-x64-gnu'

i'm hitting this after installing openclaw on WSL2 Ubuntu on Windows, any suggestions besides uninstall/reinstall?

[fresh aspen]

how cna i pair telegram? Tried /start on the bot and nothing replies

[crimson steeple]

If you're looking through CLIs, I think "." folders are hidden

[crimson steeple]

Do you need to use WSL?

I've been playing with mine on just normal windows CLI, sometimes powershell

[crimson steeple]

Don't have much issues with typical windows so far!

[mossy spindle]

ok, i'll give that a shot then

[crimson steeple]

Dunno about the clipboard module, that seems to be a major issue with a lot of people on various systems

[copper junco]

It finally worked @crimson steeple, thx man, you are a life saver

I have my last question, to change the apis for the skills and the agent itself, I need to start the onboard and edit there right?

[grim warren]

where can i learn i have basic understanding but not a swe or anything. i gonna read over some other threads here get baked and give it another go later but i appreciate

[crimson steeple]

Yeah, though you gotta be sure to download brew as well

[copper junco]

I did aswell

[mortal star]

I've been struggling with antigravity all day, it's unbearable.

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

[crimson steeple]

Try searching over this discord and the docs
https://docs.openclaw.ai/start/getting-started

[mortal star]

stone! any solution bro

[vast thorn]

reinstalled the gateway service and im using 3-flash-preview and it is working again

[crimson steeple]

Yo
to fix anti-gravity saying you're out of date, go into
⁨C:\Users\User\AppData\Roaming\npm\node_modules\openclaw\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]

Truly the model!

[mortal star]

mm i go thx!

[crimson steeple]

so many..

[vast thorn]

bot is walking me through the discord bot setup too 🙂

[crimson steeple]

Yeah, talk to the bot, it knows how to set itself up!

[winged vessel]

I was so happy with my progress the past few days, until I had to start a new session because my context window got full, then it became like absolute crap

Have you guys found a fix for this or a way to improve it?

[copper junco]

Bro at this rate u are going to become an official community helper

[crimson steeple]

aw thanks man

[crimson steeple]

Tighten your prompts!

I've been asking my boy to improve their own prompts, you can go through each prompt file and give it loads of instructions

[winged vessel]

How would you suggest doing that?

Won't that just help you delay your context window getting full but when you start a new session it will still get worse?

[vast thorn]

great..

[mortal star]

the best!

[crimson steeple]

Here's my agents file

[vast thorn]

nvm new session and its fine lol

[crimson steeple]

It needs something to begin with! If you don't like a thing, you should add it to the prompt to make it known.
Can you elaborate what you mean by saying it gets worse? In what way does it get worse?

[winged vessel]

Going through this now, one sec!

[compact night]

OpenClaw seems not to be detecting my model on logs?

05:36:32 error diagnostic {"subsystem":"diagnostic"} lane task error: lane=main durationMs=0 error="Error: Unknown model: openrouter/tngtech/deepseek-r1t2-chimera:free"
05:36:32 error diagnostic {"subsystem":"diagnostic"} lane task error: lane=session:agent:main:main durationMs=2 error="Error: Unknown model: openrouter/tngtech/deepseek-r1t2-chimera:free"
05:36:32 error Embedded agent failed before reply: Unknown model: openrouter/tngtech/deepseek-r1t2-chimera:free

[copper junco]

The api from ai studio is gemini cli? Or just gemini?

[brave trench]

I changed my Node.js to 25. Is this a problem or should I revert it back to 24

[long owl]

Are you running deepseek locally?

[crimson steeple]

should be cli i think? i might be wrong

[hardy mulch]

does mac installation involve needing docker?

[crimson steeple]

no thats open router

[blazing cloak]

EBADF

[crimson steeple]

shouldn't

[long owl]

Whats that?

[crimson steeple]

openrouter is an api hub

[compact night]

No, I got the API key from OpenRouter. I tested the configuration on Janitor.ai and it worked fine.

[brave trench]

Ok ill revert back to 24

[crimson steeple]

Try running /models and /model in your dashboard / gateway

[crimson steeple]

Shouldn't be a problem I mean

[long owl]

DOes it hand api tokens for free?

[crimson steeple]

Certain models are free, but will be under major rate limiting

API keys are free

[long owl]

woah, great!

[compact night]

50 messages/day and needs to wait about a minute for a reply but yea

[brave trench]

Ahh done deal lol it was giving error codes anyway

[crimson steeple]

Alright type in /models openrouter

[crimson steeple]

Do you got git installed?

[supple talon]

Hi, just onboarded yesterday with provider github-copilot and telegram; it worked till TG bot it started to answer with "429 quota exceeded"; any ideas?

[brave trench]

Im just really struggling using OpenRouter to communicate with the bot

[crimson steeple]

Get 10$ in your account to turn that into a few hundreds

[brave trench]

Yup

[crimson steeple]

You ran out of quota, use another model, type /model and change, or run ⁨openclaw config⁩ and then hit Models

[compact night]

Models (openrouter) — showing 1-1 of 1 (page 1/1)
- openrouter/tngtech/deepseek-r1t2-chimera:free

Switch: /model <provider/model>
All: /models openrouter all

[brave trench]

The only problem im having is that my bot isn't communicating with the dashboard or telegram

[crimson steeple]

huh.. openrouter might have issues on their end then
@ocean tusk

[brave trench]

Ive tried so many different things

[crimson steeple]

I'd check their docs first, good luck!

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

[brave trench]

Not sure what im missing

[compact night]

I did 😭

[crimson steeple]

Can you elaborate?

Yeesh... might be an issue on their end

[stable pumice]

anyone know why my bot just doesn't work? using claude pro and gateway seems connected, but no response

[crimson steeple]

Show us your logs! ⁨openclaw logs --follow⁩

[vast thorn]

hit my rate limit on gemini. is there any better ai i can use with a far more usable rate limit for free?

[supple talon]

stupid me; thanks

[crimson steeple]

free is a hard ask

these days? 😔

[vast thorn]

all right. local. i have a good gpu

[stable pumice]

nothing seems too out of the ordinary but idk

[vast thorn]

2070

[crimson steeple]

yeah local is nice, run one of those 8b models, they're actually decent these days

[vast thorn]

which one on the onboard could i use?

[crimson steeple]

Can you run ⁨openclaw status --deep⁩

[winged vessel]

I prompted it to have a self learning loop + proactive agent system where it constantly learns about me and builds things while I sleep

Those features are integrated with cron jobs and they're supposed to get better over time.

Ex:

• daily X scraper for my account watch list
• Intel scanner for top clawdbot skills and new things to try
• etc

The problem is when the context window gets full and every time we start a new session it literally just becomes worse.

All our progress almost feels lost because it wouldn't match up with the quality that it had in the start.

[crimson steeple]

Hmm... I haven't done local models before, I cannot help there sorry..
You are better off checking their docs
https://docs.openclaw.ai/start/getting-started

[grim warren]

i got thru it, that error was an issue with node 24

[crimson steeple]

Ah, so node 24 isn't compatible?

Great, will tell others that

[winged vessel]

• our content machine gets worse cause it forgets to use my voice that we already configured in the md files

[grim warren]

i have no idea but seemed to fix for me

says 22+ tho

[crimson steeple]

Things change very fast

[copper junco]

Bro i tríed the openrouter api, and feels so mean, and the gemini one feels more welcoming lmao

[crimson steeple]

Who knows?

[cinder siren]

Is there a way for me to talk to my bot on my computer in the windows powershell menu ?

[crimson steeple]

⁨openclaw tui⁩

[brave trench]

For some reason everytime I try and Configure it defaults to Openrouter/auto.. ive tried unchecking it then re running the system. I went into Openrouter and saw there was a "Routing" section to add multiple models and added the ones I was using there. Still getting errors. Ive tried changing the node from 24 to 25.. going back to 24 though cause caused errors. The Openclaw Doctor checks seemed fine. I feel like its an Openrouter issue 🤷‍♂️

[crimson steeple]

Try reinstalling the service maybe

⁨openclaw doctor⁩ might help too

[stable pumice]

ok thanks

[compact night]

On openclaw status the listed model does not start with "openrouter", do you think that could be related to the issue or is it just how it shows on this table?

[crimson steeple]

Seems like a problem with openrouter

May just be...

Ah, here it is

[crimson steeple]

Try this

[compact night]

applies to me as well?

[crimson steeple]

Yeah, perhaps, try

[white epoch]

Anyone know what might cause this and to reset it? I ran out of API tokens but I renewed it and this is what I get.

[crimson steeple]

Try restarting the session
/reset

[white epoch]

is there a terminal cmd for that?

[crimson steeple]

Do that in your tui or dashboard
Or you can do ⁨openclaw gateway restart⁩, that should clear the session?

[winged vessel]

Curious on your thoughts! @crimson steeple

Does your clawdbot also get worse when your context window gets full or after starting a new session

[crimson steeple]

Kinda, it'll obviously have to search for your memories again, but you can get it back to speed

[brave trench]

👀👀 I pray to the good Lord above this does the trick 🙏 Be right back lol

[crimson steeple]

I recommend using qmd, install it

[winged vessel]

I already installed qmd by tobi

[copper junco]

Does the agent has the ability to create commands on telegram?

[crimson steeple]

got the skill?

[crimson steeple]

ask it

[winged vessel]

I'm thinking of just having multiple bots for different purposes. It might be more effective

Do you know how to set up multiple call clawd bots in one Telegram account?

[crimson steeple]

no idea

[cobalt phoenix]

Use multiple gateways? @winged vessel

[vast thorn]

nvm. i got the 300 dollars worth of credits free trial on gemini. after it ends though im probably gonna use a local model like gpt oss or something

[copper junco]

It said i should do it manually on botfather

[winged vessel]

Do you have any links where I can read up on this? Thanks

[crimson steeple]

maybe in docs?

yet again, best docs is the bot

[cobalt phoenix]

Yupp

There are two things which are relevant https://docs.openclaw.ai/concepts/multi-agent

multi agent & multi gateway

https://docs.openclaw.ai/gateway/multiple-gateways

[crimson steeple]

it's better at googling than you

[inland cypress]

Anyone who focuses on using Claude Code CLi for coding not vibe coding but also checking the code how does OpenClaw compare just for coding?

[crimson steeple]

pretty sure openclaw knows it's own docs

[winged vessel]

Thank you. Have you set this up yourself?

[crimson steeple]

depends on the prompt, don't like it? ask it

[cobalt phoenix]

I just thought this channel is called users helping users so I can help someone instead of telling them to go ask bots constantly 🤣

[cobalt phoenix]

Not yet but I’m going to do multi-gateways today to get my bot onto moltbook. That seems like the right amount of separation

What’s surprising no one is talking about this setup

[inland cypress]

Not the answer I was looking for

[compact night]

kinda a dumb question but when I close and open my bash commands with openclaw only work after I run the curl command with the install script again even if the gateway is open. What's the proper way to be able to use openclaw commands again?

[crimson steeple]

By default? Claudecode is probably better

[cobalt phoenix]

Is openclaw in your PATH?

[compact night]

that's probably it... thanks

how do I do that again?

[winged vessel]

I know, right? I wonder if anyone has posted a video of their set up on this

[cobalt phoenix]

depends on your shell

But if it’s mac then it must be bash. Check .bashrc

[winged vessel]

By the way for a multi-gateway system, do I need to make a new Telegram account or can I just use one Telegram account and create different bots from Bot Father?

[crimson steeple]

try npm prefix -g

[cobalt phoenix]

I think you can use the same account (no reason why not) but I haven’t tried it. Try and tell us 🙂

[winged vessel]

Will do, thanks bro

[compact night]

ok, the model runs properly after I closed bash and started a new session... Now to fix the missing API key

although possibly my failed attempts exhausted the 50 free messages

nah my uses on the openrouter dashboard are sub-10

[crimson steeple]

users help users
we've been asking about how to openclaw
but i ask
how to get money, im poor

[stoic dome]

@proud sigil i have the exact same http 401: user not found error did you resolve?

[civic schooner]

Hello! Struggling to get this setup. Used the install script but I cannot seem to have it use my Ollama LXC on my LAN. I cannot figure out where the settings would allow for Ollama to be used. Any suggestions? This has been a miserable 3 hours.

[torn maple]

ive been trying to do the EXACT same thing for about as long

ive been using this https://docs.openclaw.ai/providers/ollama

[mellow sundial]

Does anyone have the tweet that had instructions how to enable flushing with memory. And fully set it up?

[torn maple]

also this https://www.youtube.com/watch?v=egofv8c7oEk

[mortal ocean]

Hi, I am trying to install a plugin but run into npm access token issues. I am not sure if ⁨openclaw plugins⁩ uses standard npm paths (@openclaw/line or openclaw/line on npm) but the package wouldn't exist there either. In the docs it does on github I found no issues related to the package being removed, and ⁨openclaw plugins list⁩ lists it too. Could it be because it is running ⁨npm⁩ but I installed via ⁨pnpm⁩?
⁨```bash
openclaw plugins install @openclaw/line

🦞 OpenClaw 2026.1.29 (a5b4d22) — Automation with claws: minimal fuss, maximal pinch.

Downloading @openclaw/line…
npm pack failed: npm notice Access token expired or revoked. Please try logging in again.
npm notice Access token expired or revoked. Please try logging in again.
npm error code E404
npm error 404 Not Found - GET https://registry.npmjs.org/@openclaw%2Fline - Not found
npm error 404
npm error 404 The requested resource '@openclaw/line@*' could not be found or you do not have permission to access it.
npm error 404
npm error 404 Note that you can also install from a
npm error 404 tarball, folder, http url, or git url.
npm error A complete log of this run can be found in: /home/ubuntu/.npm/_logs/2026-01-31T06_25_34_587Z-debug-0.log

[maiden rover]

Hi, someone has encountered the problem of working with kilov skills are blocked and it is impossible to access them. I've tried different diploma options and connecting to the project doesn't help.

[native shadow]

Hi everyone — does anyone have OpenClaw deployed on a cloud server? I’ve got it running and Telegram messaging works fine, but I can’t access the web UI from outside. Any idea what configuration might be missing?

[torn maple]

I would guess the web UI is only on a local interface, youll probably have to ssh and forward the right port

[grim warren]

has anyone let this run wild on their pc no vps no protection

[gaunt robin]

anyone get trouble with ⁨clawdhub⁩ login?
I successfully got cli token, but it returns ⁨Unauthorized⁩ error

[torn maple]

i was gonna turn over this laptop to openclaw for the lulz, but shiz wont work

[native shadow]

I suspect it’s only accessible locally — but I’m not sure how to fix that. Are there any tutorials, or could you suggest key prompt I should use to ask the right questions to AI?

[spare void]

sorry! I want just to ask, im using codex as main LLM for my bot, no issues, awesome with codex pro. But some times for coding, codex just stuck with very complex issues. So, I have in the same machine Claude Max, and ask him to ask claude using CLI when have problems. Good so far. But, just asking, im missing something here? Can I still being banned from Claude?

[crimson steeple]

loopback should be a lil more safe

but protect yourself

[torn maple]

something like
ssh -N -L 18789:127.0.0.1:18789 user@removevpsIP

then point your browser at localhost:18789

[sand grove]

has anyone got this working with a really small local model like a mistral:7b or similar?

[torn maple]

trying to do that. ollamas on a different box, but same idea. cant even do a basic chat

[sand grove]

yep - same issue - not a single response back....got some timeouts after 600s

[torn maple]

not sure where to find the error logs 🙁

[terse talon]

any linux distro that is better for running openclaw?

[sand grove]

websession has a logs tab at the bottom.....somewhat useful

[torn maple]

...heavy on the somewhat :/

[sand grove]

evolving fast though...wip..i guess

does this look ok -
🦞 Clawdbot 2026.1.24-3 (885167d)
🧠 Model: ollama/mistral:7b-instruct · 🔑 api-key ollama…-local (models.json)
📚 Context: 0/32k (0%) · 🧹 Compactions: 0
🧵 Session: agent:main:main • updated just now
⚙️ Runtime: direct · Think: off
🪢 Queue: collect (depth 0)

any/all pointers greatly appreciated - thx.

[odd swallow]

Could use a reverse proxy like Traefik and Mkcert (for https) and pihole or whatever to give a local dns entry, e.g. I can access mine at https://openclaw.panic on my local network

from any machine, just have to ⁨openclaw devices approve <whatever that device is>⁩

[lean wasp]

heya, im new to running ai and stuff, i was wondering if i could run a local ai on my pc and let openclaw use it instead of using an api?
Ive been looking for answers for a good hour and two trying to find answers with no luck

[latent hedge]

Any IT Tech professionals here? I'd like to hear your opionion on the state of the rebrand exercise. Uninstalled Clawd/Molt a few days ago and are waiting for the full rebrand, etc to be finalized, so command structure, docs, and everything is aligned after the rebrand. My last install had docs/guides saying the command was 'moltbot' while they still were 'clawdbot' in terminal. How aligned is the product now?

[odd swallow]

If you want consistent results I would use one of the heavy hitters & their api, local models are slow and heavy on resources

Gemini 3 Flash is super fast and consistent (and cheap as hell)

[torn maple]

ive heard people have gotten it working, but theres a couple folks in here still tryin. https://docs.openclaw.ai/providers/ollama

[lean wasp]

I tried it but i kinda cant afford a google api, i tried the free one and it got limited out of nowhere >.>

[lean wasp]

I tried ollama but i didnt notice there was a guide for it for some reason, ill give it a try rn

[odd swallow]

https://docs.openclaw.ai/concepts/model-providers covers some other models too

[odd swallow]

If you ran the new installer for openclaw it automatically migrates everything to openclaw from moltbot/clawdbot

[latent hedge]

Yeah... I'd have to install everything from scratch?

I didnt mean for that to be a question 🙂

[odd swallow]

If you deleted everything, I guess so, if you still had old moltbot/clawdbot it would just migrate it automatically for you to the new openclaw etc.

ah

[lean wasp]

Ive been getting this the whole time, even following the guide

[latent hedge]

But deepseek isnt an anthropic model?

[odd swallow]

You're gonna probably end up on one of the ollama/llama3.3 models

depending on system specs

[lean wasp]

I have an rtx 4070 ti super and 32 gigs of ram
soo erm

[odd swallow]

stick to a Llama3.2:3b then

[torn maple]

i was going to try the ministral-3:8b & see what happenes

trying to use local models usually hits context window issues bfore anything. using roocode could go for a little while befor elosing its mind

[odd swallow]

anybody try lm studio over ollama?

[limpid wyvern]

Hey I want to setup kimi 2.5 but it says always 163% tokens used

[torn maple]

basically the same thing at the end of the day. lmstudio does have a great UI\

[warm saddle]

hi, does this bot support cursor cli? how do i add i add it? any small guide would help, so i can use it in place of codex or gemini cli

i have cursor subscription so i want to take advantage of the opus 4.5 in there

[lean wasp]

ill give it a try rn

[wind cypress]

I am getting errors sayig there is no available auth profile for openai-codex, but my usage shows I should be good. Anyone else get this ever?

[latent hedge]

Anyone else getting certificate warnings from openclaw.ai website?

Ahh well.. fuck it. Will just VPN around my ISP provider 🙂

[warm saddle]

i hope so

[median quarry]

My

My boy read it

But no response

[cloud swallow]

where can i change my api key

[median quarry]

I’m using the free Gemini test

What the hell

It just texted my phone number a code

Bro how does it know my phone number

[torn maple]

looks like your using telegram?

[wispy forum]

hey anyone able to help with this :- How do you get consistent learning that persists and actually gets applied?- Tried SOUL.md protocols. Bootstrap files are passive context. Claude ignores unless explicitly invoked.
Tried hook injection. Better compliance. Still inconsistent.
Tried cron jobs for daily lesson extraction. Runs fine. Agent still doesn't recognize patterns across sessions.
Root issue: LLMs are stateless. Files are text, not code. No persistent protocol. Every message starts fresh.
When people claim self-learning agents, what's actually happening?

[torn maple]

usually evolving context scraped from files or stuffed in a semantic db for some kind of rag

[median quarry]

Yeah but it texted me a 6 letter code

Isn’t telegram supposed to be 8

[stone junco]

at the end of the day your model is still limited by it's contextwidth no matter how much you want it to remember

[wispy forum]

Do you scrape logs on a schedule and upsert to a vector DB? How does the agent know when to query it? And once it retrieves something, how do you get it to actually apply it vs ignore it like any other context?

[median quarry]

Yeah but it sent me a 6 letter code on my phone number

Shouldn’t it be 8 letters to my telegram so I can give it the prompt to approve my device

[torn maple]

idk. if your hooking it up to telegram it might have access to tht info

[median quarry]

Yeah

[stone junco]

1. i scrape mygoogle gemini studio gemini logs and put them in the memory folder
2. the memories get converted to .md and vectorized/embedded autoamtically by calwd
3. this onyl happens when i ask it to use it's memory_search
4. you just ask to search for something and its just running a tool

[median quarry]

Just so weird

[stone junco]

like i have 100 million tokens in memory_saerch bc there's no way i can fit that into any model to just remember tha tmuch info

[wispy forum]

So memory_search is manual. How do you get the agent to actually call it without being told? That's where I'm stuck. Hook injection to force "search memory before acting" still gets ignored half the time.

[stone junco]

yeah it's manual so i usually say hey can you use memory_search to look for so and so

like i use that keyword and it remembers and it knows but im sure if you wanna be more natural about it you can like drop a hint of what to associate it to in agents.md

"if i say search automatically assume i mean memory_search"

bc internally it could be web_search and whatever other searches like my clawdbot has web search enabled + memory search + a custom news search so it's like which one?

in fact i have 2x memory searches

[torn maple]

what the ai 'knows' all depends on the automation wrapped around the llm. openclaw seems to cover a lot of interesting bases

[buoyant rune]

guys will it keep running if i log into another user on same machine? e.g. kids want to play minecraft on another account

[vast thorn]

since the name change to openclaw. my ollama bot stopped working

This will modify your Clawdbot configuration:
/home/brandon/.clawdbot/clawdbot.json
Backups will be saved to /tmp/ollama-backups/

Proceed? (y/n) yes
Added glm-4.7:cloud to Clawdbot

Launch Clawdbot now? (y/n) yes

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

[wispy forum]

Makes sense for disambiguation. Three search tools, one keyword.
But that's read. What about write? Getting it to log learnings without being told each time. Any trigger pattern that actually works?

[oblique onyx]

Is it possible to get the web UI to show it's reasoning or the "thinking" ?

[torn maple]

1. probably, it gets installed as a servece
2. i would highly recommend not letting it run on a computer you dont want destryed

[buoyant rune]

thanks. well im planning to create a new user for it, so it wont have access to my stuff if thats what you mean, or do you mean the hardware?

[brave trench]

THIS IS THE OPENROUTER FIX! Thank you so much!! It worked 🙏🙏🙏

[torn maple]

its the system level access that Im concerned about. Unless theres a solid review of the permissions at play, I would have it on a separate system, or in a vm

[stone junco]

yeah just say save this to your memory and it knows what folder and what format

like sometimes ill go on a long design session before i start coding and right after we agree on how it should work ill be like save that design idea into your memory and then we start coding it

earlier today i was dropping it pics on whatsapp and it can see the filename of the files dropped so i was having it save the filenames into it's memory associated with what each picture is about so i can do something about them later

[buoyant rune]

another semi-random question, does this have multi-agent capabilities, like can it work on multiple tasks at the same time, I guess yeah? e.g. multiple people ask it to do something

[stone junco]

ive tried sub-agents a few times but i dont know i can't get the hang of it yet, there's a lot to do and explore so there's sitll parts i haven't messed with, in my experienced it wasn't smooth like i felt liky my agent kept handing off tasks and then the sub agent wasn't notifying my agent or something so it's like do i have to follow up and ask what happened to that task? i dont get it

[loud garnet]

I'm "trying" to get OC setup for my digital twin. Still struggling with something simple: check my AWS workmail every 10 seconds.. reply and process emails via himalaya skill.. if I use heartbeat, I keep seeing heartbeat updates in my default chat (currently tunneling to the gateway chat at 18789).

I want to use email as the primary control ux and later whatsapp and x DMs

I haven't setup x DM yet - (mayurdotai is the twin). I assume most of you are using the x api? Or is there an mcp/skill I should use instead?

Email unlocks everything else I assume - github, x, facebook, etc.

I'm at the point where I either contrib to the project and help build a better config system, or fork and build my own.

Thx!

[vast thorn]

What's the best local/free solution for a model to use?

Ollama isn't working anymore because of the rename on OpenClaw it can't detect it anymore because it's not called clawdbot/moltbot.

LM studio is nice but I can't get it to install properly.

Does hugging face or anything like that work?

[full zenith]

Hi folks, good morning , quick question is it possible to change clawdbot embedding model ? Seems like I ran out of limit . I have local embedding model in ollama that would like to use

[stable peak]

Hi, I only have experience with things like SillyTavern, OpenWebUI, and Letta. Where would I configure system instructions or a system card, or a persona prompt for my agent?

[stable peak]

i think OpenRouter still has free Kimi
idk what the rate limiting is tho

[elder oyster]

Did you fix it?

[wet adder]

So I just started. I installed from git, then I read the part about having claude help me with the install, so I opened claude, and boy oh boy, claude did not like what openclaw did. It threw warning after warning. It doesn't like that my Anthropic Claude token is stored in plaintext. It hates that openclaw is an unverified source, and apparently, when I cut and pasted a prompt I had written for claude, I accidentally included an artifiact and Claude thought that it was an xml injection script. So Claude Code rn hates openclaw. Finally, there seems to be something wrong with the gateway as it's crashing repeatedly as the expected path /opt/homebrew/lib/node_modules/openclaw/ doesn't exist.

So my questions are (1) Why are oauth tokens stored in plaintext? (2) how do I verify the openclaw domain for claude in terms of safety and security, and (3) Should I follow Claude's persistent recommendations that I (A) revoke my Anthropic token, before (B) letting it clean up my openclaw files and move forward?

[crimson steeple]

I'm pretty sure you can set envs

The tokens are stored in those too

[cloud swallow]

i just reinstalled clawdbot and it is offline how do i start it up i have already went through the onboarding

[wet adder]

Claude won't even fetch openclaw documentation as "fetching from unverified sources is a vector for prompt injection."

[mellow sundial]

How do you update it to be renamed to OpenClaw?

[twilit shadow]

anyone else having issues with subagents?

My bot seems completely unable to use them

also having issues with telegram slash commands... they dont work for changing the models (it looks like it changed in /status, but the actual bot didnt change)

and sometimes the slash commands crash the bot.

wondering if anyone has experienced this and knows a fix

[molten obsidian]

Run the new installation script. It will detect the old install and simply update it, Afterwards you can have your agent check the workspace for any Remaining issues.

[long owl]

I was using gemini-flash-2.5 I hit my limits so I was trying to switch to Gemma, but it is not picking it. How do I restart the whole thing so it picks the new model?

[crimson steeple]

You can change the model by running /model

see what model is allowed by typing /models

[gilded moat]

I've got openclaw set up in a Docker container and the bot is working with Discord, but in the OpenClaw control panel, all of the Config settings pages have this error message - not sure if I've got something set up wrong. Any ideas?

[crimson steeple]

that's not an error

[brittle cliff]

Hey everyone! I’ve been running Clawdbot for a while, and I’ve spent the last 4 hours trying to migrate to the new OpenClaw (v2026.1.29). I need some advice on the safest way to "fresh start" my config without losing my bot's personality.

​Current Situation:
​System: Ubuntu, recently updated from clawdbot to openclaw@latest via the official install script.
​Service: Migrated legacy clawdbot-gateway to the new openclaw-gateway.service.
​The Issue: I’m getting ENOENT errors when trying /tts. It seems like the ElevenLabs API key and voice settings didn't migrate cleanly into the new ~/.openclaw/openclaw.json.
​The Error: Media failed: ENOENT: no such file or directory, open '/tmp/tts-xxx/voice-xxx.mp3'.

​My Question:
I want to run openclaw onboard --install-daemon to re-enter my API keys (Gemini, ElevenLabs) and properly link WhatsApp again.
​Is it 100% safe to run onboard without losing my soul.md, identity.md, and memory.md files in ~/.openclaw/workspace/?
​I'm terrified of wiping the bot's "soul" and memory. Should I just manually edit the JSON, or is onboard the "OpenClaw Standard" way to repair a broken config during this migration?
​
What I've tried:
​Cleaned up legacy systemd units.
​Verified openclaw doctor (shows ElevenLabs as unconfigured).
​or just normal re-configure fix it? Since right now the bot WhatsApp is sending unplayable audio file for every I sent, even /new session nothing! Even /compact command on WhatsApp it sends unplayable audio

Any advice from the experts would be amazing! 🦞

[mellow sundial]

Why don't you backup those files

[gilded moat]

It's red like an error 😅 , and I can't use the forms version of any of the configs pages.

[crimson steeple]

hmm.. safest thing is possibly running openclaw doctor

[crimson steeple]

Indeed if you want a fresh install, save your openclaw folder, or wherever your SOUL.md is

[brittle cliff]

How I do that on VPS? cp -r ~/.openclaw/workspace ~/openclaw_backup
or what's the right way or command?

[crimson steeple]

try googling it, i forgot too 💀

[mellow sundial]

Ask OpenClaw! It can def do it

[crimson steeple]

also no shame in asking the bot to do it

The bot migrated to openclaw for me perfectly, even made a tidy backup

[brittle cliff]

Its broken right now on WhatsApp. Sending broken audio files for every message I send

[crimson steeple]

can you ask it to stop?

[brittle cliff]

Its broken

[twilit shadow]

what about /restart

[fresh moon]

hi guys using vps here is there an easier and faster way to find skills or just do find skills myself and install it through the terminal vps to be faster

[brittle cliff]

Same sended audio file

[crimson steeple]

/reset might help

[brittle cliff]

Will delete its memory?

[crimson steeple]

I'd say just doing clawdhub skill installing is good enough

[crimson steeple]

Long term? no, but it restarts the current session

Clears the context

[fresh moon]

okii

[crimson steeple]

You might need to go into your agents.md or soul.md and redo their prompts

[brittle cliff]

How will that solve this? Isnt that unrelated

[crimson steeple]

It's possibly sending you only audio because you asked it to before, right?
If not, you might need to reset its memory

Though best to try to reset the config

[brittle cliff]

I want do in a way that doesn't reset memory or soul or identity.md

Config reset best way than or just relink WhatsApp?

[crimson steeple]

Run openclaw onboard again, shouldn't reset the soul

[twilit shadow]

or try openclaw doctor

[low urchin]

Has anyone used OpenClaw on Discord itself?

[crimson steeple]

Yeah

my bot's hosted on discord

[brittle cliff]

Didn't work

Trying to onboard now. And is there way to back memory or soul and identity before I do it? On VPS?

Hopefully onboard fixes it.

[crimson steeple]

Google how to copy files in cli

[worthy berry]

Is it normal that for few exchanged messages to exceed 1M in context input? I got rate limited instantly??

How much is yours context input/output average per day

[shut dagger]

I'm running into two main issues with my OpenClaw setup. First, the process frequently crashes, requiring a manual restart via clawdbot gateway. Second, I’m seeing inconsistent behavior with the bot's personality. After installing a new skill, the bot changed its writing style entirely, but upon restarting the gateway, it reverted to an old

[crimson steeple]

I'd write the skills yourself, or just feed the skill page itself to your bot, it can replicate it, don't copy it in wholesale

[dire harbor]

how do i change the heartbeat message. i tried from AGENTS.md but it doesn't seem to update

[crimson steeple]

That should be fine to update, why isn't it? Ask your bot about it yet?

[shut dagger]

yeah but look this is deep, i talked to it for the first time yesterday, it was pretty bad i went through onboarding etc again got a new api key too, and got a fresh start was good until it crashed after 30 mins (this is annoying too) now to relauch it apparenly ive to use the gateway command so i do, it relaunches but it isint the same anymore, meaning its using its old persoanlity if you get what i mean.

how do i fix the crashing lol

[crimson steeple]

Were there logs? There's so many reasons it could've crashed
openclaw logs --follow

[lean wasp]

Any of yall used ollama instead of an api?
I've asked before but now i feel completely lost if anyone could help

[shut dagger]

i did

[lean wasp]

can i dm you?

[shut dagger]

i mean im no help but i did get it running

[vale scroll]

Gosh I'm struggling too, clawdbot, moltbot, now openclaw, I've had the binary go missing a few times during restarts and having issues that its not saving or updating .md files for some reason.

[crimson steeple]

That can't be it, binaries can't just poof

[long owl]

Yo I want to try it too, I tried free gemini limits. Exhausted and now trying to switch to openrouter's gemma 27B which is supposedly free, but it just won't pick it. SO I have unintsalled and restarting things from scratch, Lemme know if you succeed with ollama though

[long owl]

Hey, we can try it together if you want

[shut dagger]

i used kimis model

its free for 7 days

works well althouh it asks for ur billing and card info, but u can just cancel it right after

[long owl]

nah, I don't have cards on me, so want to try absolutely free possible hacks

ollama is one, and whatever place is handing APIs for free

I just want to play with it, so performance or accuracy doesn't matter to me rn

[shut dagger]

well u can run it locally but thats if you have a good spec pc

[long owl]

I am running on macbook pro 16GB ram

[latent hedge]

Whats the cheapest access to good LLM models for OpenClaw? What are people using? Because I find Anthropic's too expensive.

[lean wasp]

yeah

sure

[alpine bridge]

Dear OpenClaw Developer,
​I hope you are doing well.
​First of all, I would like to express my sincere gratitude for creating OpenClaw. I have been using it as my primary personal assistant for a while now—handling everything from note-taking and task reminders to fetching news. It has become an incredibly useful tool in my daily workflow. 🌟
​However, I am currently facing a persistent issue regarding the Cron functionality and Telegram notifications:
​The Issue: I cannot successfully set up task reminders. Even when a reminder appears to be set, it fails to send a notification to my Telegram when the time comes.
​Errors encountered: Often, OpenClaw reports a "Gateway Timeout" error, suggesting it fails to communicate with the gateway when processing reminder tasks.
​My Environment:
​I am running OpenClaw on a Debian Virtual Machine (VM).
​I am using Telegram as the primary interface.
​Troubleshooting done: I have tried deleting and reinstalling the entire VM multiple times to ensure a clean slate, but the Cron issue persists every single time.
​I really love this project and want to rely on it for my daily tasks. Could you please look into this or suggest a way to fix the communication between the assistant and the gateway/cron service?
​Looking forward to hearing from you. Keep up the great work!
​Best regards,

[long owl]

I found something https://www.youtube.com/watch?v=egofv8c7oEk

[lean wasp]

tysmmm

[paper stirrup]

Hi fellas. I'm getting the "This version of antigravity is no longer supported..." error. I saw in general that it was fixed but a fresh reinstall still gives me this. Any advice? (Running on a windows system btw)

[ocean reef]

just wondering, if i wanna develop my openclaw flows and logic, will it be smart to do it using the openclaw itself ? or working on the files on github with ci/cd just like when working on development app ?

[faint plume]

HELP NEEDED

Hi Y'all -- I’m on my 5th Openclaw install - what am I doing wrong?

1. install OC ✅
2. put in moonshot ai keys ✅
3. put in open ai keys ✅
4. setup done, gateway online, everything fine ✅
5. BUT, there was a problem — I messaged an image via whatsapp. My bot did NOT understand the image. It said there is no OpenAI token.* So I gave it an openrouter token with the instruction: “Use it for audio / video. NOT for primary model”. And that’s it… everything breaks. It writes it into the config and breaks everything.

Looking for suggestions on a) how to fix it and b) how to ensure this does not happen

WHENEVER I’ve given OC an model keys for something, it mucks it all up.

*also, I DID give it an openai key at the beginning... don't know what happened to that.

[latent hedge]

HEELP!... I just installed OpenClaw on Ubuntu.. brand new.

After I installed it, it says (no output) in the TUI..

Then I saw in #1459642797895319552 that someone wrote, "Just quit and do a 'openclaw tui' to restart

[crimson steeple]

Model issue possible

[latent hedge]

But 'openclaw' doesnt exist as a command?

wth?

[carmine mesa]

please may i have help installing openclaw? i am potato

[crimson steeple]

Are you still on moltbot or clawdbot

Try those instead of openclaw

[latent hedge]

I literally JUST started from scratch on a new machine

[crimson steeple]

Try it either way

[latent hedge]

neither works

not moltbot, not clawbot, not openclaw

[crimson steeple]

How did you install it? What command?

[latent hedge]

curl -fsSL https://openclaw.ai/install.sh | bash

I can't even run the uninstall command recommended, because 'openclaw' doesn't exist?

weird

Didn't have these issues on my Mac Mini

this is an ubuntu latop

[crimson steeple]

Try running npm prefix -g

[latent hedge]

.npm-global

/dir/dir/.npm-global is what it returns

[crimson steeple]

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

[latent hedge]

nothing

blank

[hushed basin]

anyone have pointers on setting up OpenClaw on goog antigravity?

[crimson steeple]

try now

The openclaw command

[latent hedge]

openclaw: command not found

[crimson steeple]

oops forgot to refresh the shell
run
source ~/.bashrc

[latent hedge]

Bingo

I did that, and the echo command after it

now it works

you da man, Stoneman! 🙂

Damn... out of limit on my gemini API.. hrmm

whats the most cost effective and cheap way to run it

Wonder what model I should swithc to

switch to**

[crimson steeple]

Run a local model, check #old-models

[latent hedge]

Then I need to figure out how to connect to ollama on my Mac Mini from my Ubunty laptop

hrmmm

I'll get to work.

Im sure chatGPT can help with that

hehe

[dapper nexus]

if you find the answer please let me know :p

[hollow elk]

Have people had much success giving openclaw a personality? Like, can I use a chatbot from chub.ai or JanitorAI, and adding the character description into the agent’s system prompt?

[tidal fossil]

Openclaw browser setup has me confused. I am running it on a VPS (linode) and dont want to install the mac node on my mac. What are my options?

[sharp wing]

I have a win 8.1 laptop 16gigs and visual studio 2019 can this AI help me build a multi user dungeon

[crimson steeple]

Sure thing

[crimson steeple]

Genuinely ask it yeah

It will work

[opaque swan]

I'm suddenly getting the error: [agent/embedded] Profile anthropic:default hit Cloud Code Assist format error. Tool calls will be sanitized on retry.

[faint plume]

HAAALP!

How do I fix a broken config? Is there a versioning system? Is there a way to revert?

[lone cliff]

Plop, how i patch that pls ?

[lyric spruce]

anyone get an error like this? the only time i see it when using search in here is from @swift quail having the same error lol. my bot just decided to kill itself out of nowhere and i got that error mid convo

[tropic glen]

Does anyone know how to setup clawbot with ollama

[gritty cipher]

Hi everyone im getting this error through powershell install
iex : File C:\Program Files\nodejs\npm.ps1 cannot be loaded because running scripts is disabled on this system. For
more information, see about_Execution_Policies at https:/go.microsoft.com/fwlink/?LinkID=135170.
At line:1 char:45

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

•                                         ~~~
  

  • CategoryInfo : SecurityError: (:) [Invoke-Expression], PSSecurityException
  • FullyQualifiedErrorId : UnauthorizedAccess,Microsoft.PowerShell.Commands.InvokeExpressionCommand

not sure how to fix it looking at the help article

can I please get some help

[tropic glen]

Using powershell administrator

Set-ExecutionPolicy -Scope CurrentUser -ExecutionPolicy RemoteSigned

[ashen bronze]

how to use this?

work with any free llm source without paid api?

[lean wasp]

can someone explain to me what this means

[gritty cipher]

Thank you testing it now. Running it as administrator and with that script

[gritty cipher]

seems to be working, appreciate your assistance.

[tropic glen]

Your welcome 👍🏿

Which ai model will you be using

[knotty solar]

Which Ai model can yall recommend? I've been using GLM 4.7 but it just seems overwhelmed with more complex tasks

[crimson steeple]

I love gemini so far

[knotty solar]

Price point?

[gritty cipher]

i plan to use anthropic, do you have a suggestion? I am looking to vibe code for my team to help build our products and ship faster

[wise dust]

hi, I set up ollama in my device, pair the telegram, try a msg but got this error:

==========
0:57:36 [diagnostic] lane task error: lane=main durationMs=11 error="Error: No API key found for provider "ollama". Auth store: /home/minis/.openclaw/agents/main/agent/auth-profiles.json (agentDir: /home/minis/.openclaw/agents/main/agent). Configure auth for this agent (openclaw agents add <id>) or copy auth-profiles.json from the main agentDir."
10:57:36 [diagnostic] lane task error: lane=session:agent:main:main durationMs=14 error="Error: No API key found for provider "ollama". Auth store: /home/minis/.openclaw/agents/main/agent/auth-profiles.json (agentDir: /home/minis/.openclaw/agents/main/agent). Configure auth for this agent (openclaw agents add <id>) or copy auth-profiles.json from the main agentDir."

10:57:36 Embedded agent failed before reply: No API key found for provider "ollama". Auth store: /home/minis/.openclaw/agents/main/agent/auth-profiles.json (agentDir: /home/minis/.openclaw/agents/main/agent). Configure auth for this agent (openclaw agents add <id>) or copy auth-profiles.json from the main agentDir.

========
This is auth profile:

cat /home/minis/.openclaw/agents/main/agent/auth-profiles.json
{
"providers": {
"ollama": {
"baseUrl": "http://127.0.0.1:11434/v1",
"models": [ollama/qwen3:32b-q8_0],
"api_key": "ollama"
}
}
}
=================

I tried to ask Grok / Gemini many time to fix, but it can not guide correct way, it always ask me to run "openclaw agents add main"
But still stuck here.

If anyone can help, it is very preciate

[knotty solar]

If you do, do it over API based billing, the pro abbonement is a joke limitwise

[pseudo hamlet]

is there an android app available, to use as a node? I need it to setup "talk" mode - https://docs.openclaw.ai/nodes/talk but without the mac

[tropic glen]

Same here

[tropic glen]

It's a good choice but do not forget to add a rate limit

[gritty cipher]

I've sent you a friend request. Would love to talk privately if that is okay

As this is something I am still learning. Thank you

[knotty solar]

Definetly yeah xD

You can easily bill 50$ + in one week if you're not careful especially with opus

[tropic glen]

Otherwise the credits will be used like a leaking dam turning it into a waterfall

[knotty solar]

But I think especially character wise as well as just from the problem solving, there is no better product than Anthropic rn

[tropic glen]

True

[wise dust]

hi, any suggestion to workaround?

[gloomy vortex]

Best provider besides claude???\

Thats cheap

[crimson steeple]

Gemini is decent, if you wanna bare a subscription

[gloomy vortex]

Does it have something l ike claude code?

[crimson steeple]

Antigravity counts I suppose?

[tropic glen]

I am still trying to figure it out

[copper rune]

Is qwen2.5:3b-instruct too stupid for openclawd? I am using it with 4 cores and 8 gb ram And the result is useless, I can't interpret it what it should be a mistake or a result

[high relic]

Will any linux version work for OpenClaw?

I got an old Lenovo M93p sitting around doing nothing. Thinking of installing linux on it and then hosting OpenClaw on it.

Any distro better than others?

[solid stratus]

check updates or reninstall gateway

[solid stratus]

use tailscale

[faint plume]

reinstall gateway -- this did not work

check updates -- what would this do?

[wise dust]

if you have any clue, pls share with me, I opened a topic here (https://discord.com/channels/1456350064065904867/1467116036628414646)

[tropic glen]

Sure if you find a clue pls share

[gloomy vortex]

I did it on an old desktop no issues

256gb hdd

[high relic]

Yeah, but any specific distro for linux?
Or a standard ubuntu with GUI is fine?

[gloomy vortex]

I went with ubuntu 24

[terse lotus]

anyone know how to bypass the 300b model limit when running local only models?
its a bit of a challenge to run models that big and at required speeds, on local only hardware... basically nobody can do this.

[high relic]

Okay, and you hit no limitations in terms of usage? I see people using it mainly with MacOS, just wanted to be sure that a standard ubuntu is fine as well

[gloomy vortex]

Nope just API limits

No issues so far

[high relic]

Awesome, thanks for the help 😄

[white jetty]

I try to wrap my head around how a good software development workflow with OpenClaw looks like. I have a GitHub Copilot Pro subscription, so I use that for OpenClaw models. Additionally I tried to setup the GitHub Copilot CLI and let OpenClaw use the non-interactive CLI approach. But it seems to not work really well, it fires one prompt and then stops doing anything (at least when not using the most expensive models). There is no real iteration.
Is using the GitHub Copilot CLI or similar tools like Claude Code maybe not even the best approach for OpenClaw? Would OpenClaw be able to do software development better on its own without external AI assistants?
Would love to get some feedback 🙂

[crimson steeple]

Asking it to improve genuinely works, this sounds like something you prompt

I've been using Gemini 3 flash, asking it for various development questions, has not failed me yet

[solid stratus]

openclaw onboard --install-daemon and clawbot doctor use both commands

[vale scroll]

Could someone DM me the answer to this. Configuring Web Search, Have registered at Brave, got the API key, applied it to the bot, done a gateway restart, no errors in the logs but it refuses to search the web and just gives me training data from the bot. For example according to "Search the web for the current date" it responds with "Today is October 26, 2023"

[white jetty]

But did you do this with GitHub Copilot CLI or did you ask OpenClaw to do the development work itself?

[crimson steeple]

Everything on openclaw

[terse lotus]

i've at times changed stuff in the code and then it wouldn't take effect until i did a full reboot... not sure exactly why...

[fickle umbra]

Hey all,
I have a chatgpt pro subscription, will it use that of i choose the oath option over the api option during setup and pretty much do the same thing? I currently have it set up on api.

[normal turtle]

My minimax, MacBook M1 Pro hosted openclaw is incapable of using a browser apparently.

Either it’s an idiot or I am

[white jetty]

Ok, thanks! Will try that next. Maybe it does not need something like GitHub Copilot CLI / Claude Code at all

[vague totem]

I am getting this message on my terminal - Invalid config at /home/ubuntu/.openclaw/openclaw.json:\n- gateway.bind: Invalid input. Anyone, please help?

[terse lotus]

you will need a 300b or larger model before browsing is allowed by the openclaw framework

[young coral]

Hello guys i'm an code beginner, i'm trying to setup clawdbot on a docker on a vps , and i'm trying to solve some gateway problem, for sure nothing too hard for a sherpa, DM if you want to help a frenchy baguette thx in advance !

[crimson steeple]

You can ask here!

[crimson steeple]

You need to run openclaw doctor

[normal turtle]

Guess I’ll have to give it programmatic access to its own calendar (which has read to mine) then

[young coral]

Oh perfect error 1008 in the UI, trying to acces via public IP

[terse lotus]

its one of the issues, i've been running into... when trying to run local models... because i can't even give it file access to the vm it lives on for the exact same 300b minimum limit... which is so annoying

[crimson steeple]

Can I have more information? Try running ⁨openclaw logs --follow⁩

[normal turtle]

I’m not even running a local model, it has full access to its own laptop, more than enough usage of the API available but he jut seems like an idiot 😂

It’s been able to tweet and access a calendar IF the OpenClaw Browser Relay is up, but that seems to drop constantly

[vague totem]

@crimson steeple - I just did and I got this Telegram: not configured
Agents: main (default)
Heartbeat interval: 30m (main)
Session store (main): /home/ubuntu/.openclaw/agents/main/sessions/sessions.json (5 entries)

• agent:main:main (10m ago)
• agent:main:cron:90cc0efe-2609-419d-ba06-cd2ae442f4cd (114m ago)
• agent:main:cron:5f365ee3-4c48-405c-b5e4-1daab1a05aa0 (132m ago)
• agent:main:cron:fcd0a1de-3cac-468d-bd89-1103ad48a5cb (143m ago)
• agent:main:cron:91d329f2-5897-4c9e-ba66-5d8ba0394117 (162m ago)
  Error: Config validation failed: gateway.bind: Invalid input

[crimson steeple]

You might need to run onboarding again
Try ⁨openclaw onboard --install-daemon⁩

[vague totem]

Alright, mate. Let me try agian

[kind gate]

I just hit the token limit for Claude , can I swap to use openai?

[crimson steeple]

type ⁨openclaw config⁩ if you wanna change models

[kind gate]

I did , still get the http 429 rate limit error

[vague totem]

I jut ran install daemon and docker fix command. Stil the same error 🙁
Telegram: not configured
Agents: main (default)
Heartbeat interval: 30m (main)
Session store (main): /home/ubuntu/.openclaw/agents/main/sessions/sessions.json (5 entries)

• agent:main:main (13m ago)
• agent:main:cron:90cc0efe-2609-419d-ba06-cd2ae442f4cd (117m ago)
• agent:main:cron:5f365ee3-4c48-405c-b5e4-1daab1a05aa0 (135m ago)
• agent:main:cron:fcd0a1de-3cac-468d-bd89-1103ad48a5cb (146m ago)
• agent:main:cron:91d329f2-5897-4c9e-ba66-5d8ba0394117 (165m ago)
  Error: Config validation failed: gateway.bind: Invalid input

[latent hedge]

@crimson steeple : Hrmm.. when I was working on setting up telegram.. I got an openclaw.json error that I had to fix with the doctor.. now I asked it to set up tailscale.. and it does the same:

-> Invalid config at /home/melsen/.openclaw/openclaw.json:\n- gateway.bind: Invalid input
Do you know what the root cause for that coul dbe?

[old sandal]

Hi everyone. I've encountered a problem. My bot keeps replying to my messages with old replies. It's not processing my request properly. I chatted with it this morning about something, and two hours later I asked what the weather was like outside, but it keeps sending me messages from the old thread.
And I'm not cluttering him with messages or topics.

[crimson steeple]

no idea

[latent hedge]

oki

[crimson steeple]

I'm having a few problems with it myself

[latent hedge]

trying to do it manually according to the guide and get this:
~$ openclaw gateway tailscale on
Invalid config at /home/melsen/.openclaw/openclaw.json:\n- gateway.bind: Invalid input
error: too many arguments for 'gateway'. Expected 0 arguments but got 2.

odd

[vale scroll]

@terse lotus Sorry just to clarify, you're saying that web search wont work unless the model is a 300 billion parameter model? And yes that's exactly what I'm doing, running a local model on LMStudio and openclaw in a proxmox ubuntu virtual server. Is this new do you know? I swear i configured it a few days ago when it was clawdbot and it worked? I also see there is some chatter about Brave API usage , in that their terms of use prohibit the use of results for inference engines.

[young coral]

Maybe this help

[terse lotus]

doesn't matter if its a local model or not... it matters how many billions of parameters it has....
a rule of thumb is higher parameters, the smarter it gets.... when i check on ollama the minimax m2 is like 230b
i think there latest might be larger the minimax m1 is 456b

not sure which you are running, but it might certainly cause issues... best to run a model you know works well, because agentic frameworks can be very sensitive to model outputs.

[crimson steeple]

Ah.. I don't know how to help with tailscale sorry

[vague totem]

@crimson steeple - can I uninstall and re-install it again>?

[weary monolith]

is it real that using claude code sub in openclaw would get my claude account banned?

[crimson steeple]

you can

[terse lotus]

its against their terms of service to use claude code sub, but hasn't been enforced a lot until recently, so far as i understand...

agentic frameworks burn a lot of tokens

[serene dust]

Hi guys.

I hope your all well.

I had configured to fallback into local llm if my ChatGPT OAuth ever failed/throttled/rejected. (I thought I had)

But I’m presented with the attached usage block and now the failsafe is not triggering. Any ideas how I can fix this?

[kind gate]

Any tip arround the rate limit ?

[terse lotus]

spend more money 😄

[kind gate]

Can't I use open ai?

[weary monolith]

hlm plan is cool with me and never hit limit with me so far, just 6 dollar a month

glm*

[kind gate]

Glm?

[weary monolith]

from z ai

their coding plan

[terse lotus]

new Chinese competitor in the market.... bet they are just being nice still... 😄 they want to get more customers 😄

[kind gate]

I have chatgbt pro, Claude pro and copilot pro

Can't I use them for Clawd ?

Or is it hardwired to use Claude as a base ?

[weary monolith]

chatgpt and copilot pro can be used yes

[kind gate]

How

[weary monolith]

avoid claude or you will get banned

[kind gate]

./model did not help

[weary monolith]

openclaw models auth login --provider openai-codex

this use chatgpt sub i use it

[rustic rampart]

i set up Openclaw in my vps. have opened up a chat and have literally been looking at this for 30 minutes. ive set up my LLM api and telegram. nothing appears to be working. is Openclaw down?

[raw solstice]

Hey everyone,

My agent is having a weird issue with its browser tool. I asked it to look something up, but it's trying to use the Chrome extension relay, which I don't have set up.

My agent did some great troubleshooting and found the fix: it just needs to change a setting in the config to tell the browser to use its built-in clawd profile by default.

Here's where we're stuck: every time the agent tries to save this new setting, it gets a generic invalid config error. We've tried a full gateway restart and a couple of different ways for the agent to apply the change, but no luck.

Any ideas why the agent can't save this one setting? It feels like we're right at the finish line.

[eternal moss]

Where can I find a cheap VPS setup guide on here?

[old sandal]

Hello everyone, my bot is starting to malfunction. After just a couple of messages, it stops understanding the context. It responds to new questions with old topics. It can take several hours, but if I ask a question about the weather, it will respond with an old topic. I am not overloading it; I am only asking a couple of questions. Can anyone help? This started after yesterday's reinstallation.

[manic solar]

https://www.youtube.com/watch?v=SaWSPZoPX34
best guide for setup i found so far. 30min and you have working setup + claude ready to fix ANY issue with your clawd.

[astral cipher]

Hi, has anyone gotten this to work by listening on whatsapp groups and sending a summary on direct message?

[jaunty dagger]

I spent about few hours trying to get ollama to get to work on my mini-pc and all the smaller models seem like they are not compatible. Anyone know of a good small ollama model that's light and does not tax mini-pc's ram or gpu? I am not expecting much in terms of the model's functionality, I just need something that can manage emails, manage files, run some scripts and maybe some light admin work. I don't need complex thinking or anything like that.

[weary monolith]

i use a 6 dollar a month one from interserver

pick a 4gb ram that's the 6 dollar one, idk how it will be on less ram than that tbh

[valid vine]

The api token cost for this bot is so expensive, is there a cheaper option?

[zenith plank]

hello, i am having the same problem, how did you added the venice api key?

[gusty sonnet]

What are you using now?

[jaunty dagger]

Yea, it's ridiculous. I asked like two simple questions and it was l ike 50 cents.

[valid vine]

I’m using claude opus 4.5

[jaunty dagger]

... that's like the most expensive model dude

[gusty sonnet]

Minimax is cheaper but marginally worse. Kimi is cheaper and a bit better than minimax. Opus is the most expensive by a long way, especially if you're using the API token and not an oauth subscription.

[valid vine]

What is an oauth subscription. Is is like a monthly payment and unlimited usage?

[jaunty dagger]

Most companies that offer oauth subscription won't allow you to use it for openclaw anyways, so don't worry about it.

[gusty sonnet]

Yes, Claude Code Pro or Max. Not unlimited, but cheaper than using the API directly.

[wild folio]

hello, any fellow friednly devs, that know about npm?
Could you please help quickly find, debug and tweak some config?
https://discord.com/channels/1456350064065904867/1467131082762617008

[jaunty dagger]

Yea, but you can get banned.

[valid vine]

But mini max, is just like molt, it still requires an api, no? So token usage cost is the same

[gusty sonnet]

It's against their terms of service.There s chance you might get banned. But I haven't been after nearly a month of doing that. I suspect the people who get banned are doing other naughty things too.

[valid vine]

Makes sense