users-helping-users

[errant lagoon]

whew maybe someone already said this but you will want to change default disk size on a lume vm to over the default 50gb... ideally before your'e halfway through the setup

[silent steppe]

Hey guys! I'm curious how would I use clawd if I want to use clawd for my business with stocks and managing orders and things like that, wondering if you guys can point me to some available SDK / plugins that works best for the above usecase?

[pulsar sinew]

Is Cerebras possible or not?
https://docs.clawd.bot/gateway/configuration does mention it, but it does not work?
agents.defaults.models.providers: Unrecognized key: "cerebras"

[sour slate]

@golden sphinx, use #1459642797895319552 for help. The fastest way to get your problem solved is to follow the instructions here: https://docs.clawd.bot/help/faq#im-stuck-whats-the-fastest-way-to-get-unstuck

[golden sphinx]

How do i remove clawd from my mac?

[random thunder]

Hey anyone seen this error and how to resolve? Can’t move past it:

LLM request rejected: messages.360.content.1: unexpected tool_use_id found in tool_result blocks: toolu_01TXK5QmxTae578da66PFgSJ. Each tool_result block must have a corresponding tool_use block in the previous message.

[formal fox]

The model was configured successfully, and during the first interaction it did provide responses. However, after a few more interactions, it suddenly stopped producing any output at all. Why is this happening? Can anyone help me figure this out?

[formal fiber]

I recommend trying Discord. My only issue with Discord running on a local system is that cron jobs don't work, even when the computer is on.

[left spear]

apologies! but how do i get the clawdbot chrome extension?

[dire mirage]

Can we get banned from Anthropic for using CLI instead of API? Seeing conflicting info online

[formal fiber]

CLaude code max is still working here.

[next elbow]

/status says I’m outta tokens in this window… but clawd keeps replying.

what’s broken? my understanding?

[zealous raptor]

infinite money glitch...?

[random thunder]

Fixed by doing /new

[lone tangle]

"gateway": {
"port": 18789,
"mode": "local",
"bind": "loopback",
"controlUi": {
"allowInsecureAuth": true
},

That was the solution, the "allowInsecureAuth" flag

[astral garnet]

That flag got me also

Each update has many changes you can trip over. I created a clawdbot guru agent with document links to help me tshoot.

[formal fiber]

How can I get the AI to install skills from Clawdhub? is there a CLI tool and/or something else to make it easier? Or an integration?

[empty hemlock]

out of interest is the claude pro plan sufficient ? or does one need to use max ?

[next elbow]

I've just started... and pro runs out very fast. Half my weekly budget was blown on the install, and it went relatively smoothly for me. Can always use pro for a while and see how limiting it is for you.

Most opinions seem to be that you want Opus on Max to do anything substantial.

planning to upgrade to max soon myself

[pallid yacht]

Has anyone found a good way to work with clawdbot with voice? I currently talk to it over discord

[compact stream]

use GLM-2.7 through openrouter

ah it also fuckin hoses it but z-ai have a sub

[sand grove]

Has. Anyone able to configure clawed to run with a locally running deepseek model

[real rivet]

Hi, What are people's thoughts on how much permission/credentials to give to clawdbots? Do people grant access to clawdbot to their main email or main Twitter so that it can help with email categorization and other stuff? How are people thinking about the risks v.s. productivity gain there?

[ornate tinsel]

any method of compressing context - i hit rate limits quickly

[empty hemlock]

Keep getting the following issue with telegram during setup -> this via ec2 running ssh locally on my mac

06:28:12 [telegram] telegram setMyCommands failed: HttpError: Network request for 'setMyCommands' failed!
06:28:12 [clawdbot] Unhandled promise rejection: TypeError: fetch failed
at node:internal/deps/undici/undici:14902:13
at processTicksAndRejections (node:internal/process/task_queues:105:5)
at runNextTicks (node:internal/process/task_queues:69:3)
at listOnTimeout (node:internal/timers:549:9)
at processTimers (node:internal/timers:523:7)

some network issue via ec2 i suspect but cant seem to find the root cause

[rigid vigil]

hi I keep getting no response. Anyone know why this is happening? Thanks in advance!

[empty hemlock]

The issue was that Node.js was trying to connect to Telegram via IPv6 which was timing out on the EC2 instance, and disabling IPv6 with sudo sysctl -w net.ipv6.conf.all.disable_ipv6=1 fixed the connection problem.

[hard moss]

I have the exact same problem with my two setups I have triied.

My issue: Clawd is runninig, health shows OK, I send messages in the Gateway Dashboard chat, empty responses are received.

[rocky orbit]

Did you try #1459642797895319552?

[hard moss]

This has happened the same way on two different Macs. Running Clawdbot 2026.1.24-3

[bright plinth]

Ok lol - I blew through my weekly limit on claude and I have it running on my macbook air m2 with 8gb of ram. Am I hosed? Should I fall into the hype and by a mac studio to run a local LLM tomorrow lol?

[lime lodge]

What will be minimum specs to run claedwbot into t=my system if I am on windows

[rocky orbit]

No links between hardware and models it can run on a raspberry

[rocky orbit]

Not much… it can run on an old computer you have at home

[oak portal]

But if I want it using a specific I Message it needs to be that Mac's I message no ?

[rocky orbit]

For iMessage it needs to be a Mac mais it can be an old one you have. No need for a Mac Mini

[steep vale]

started out in a 1 GB vps on racknerd

[oak portal]

Got it but for example each company would need its own Clawd & Mac if each has its own I message I guess ye ?

[rocky orbit]

Each company?

[oak portal]

Like each of my companies basically

Each company has unique I message #

So need a Clawd & Mac for each since can't use multiple I message on 1 Mac ye ?

[rocky orbit]

Yes or you can setup groups on Telegram

Telegram integration is better than iMessage imo

Then you just need one bot and create a group with topics for your separate companies

[oak portal]

Telegram can go thru I message ?

I'm looking to also manage my text chains on each phone I have aswell

[scarlet shell]

burning credit even when it's idling, anyone have same issue:

[rocky orbit]

Oh ok I see @oak portal I’m no expert but I think you have than to get one separate computer per company and run one bot per company

[rocky orbit]

Yes it’s cron jobs or heartbeat

[oak portal]

Do you think I need multiple chat Gpt, Gemini, claude etc subscriptions aswell ?

[scarlet shell]

thanks @rocky orbit , not cron job set yet, it must be heartbeat then, can it be disabled or delayed?

[rocky orbit]

Yes on the config file

[rocky orbit]

Hmm I guess no but depending on the use you have for each bot it can become really expensive quickly

[oak portal]

Clawd uses tokens of models or just subcriptions ?

[rocky orbit]

Can be both! I guess you’d have more precise answers on #1459642797895319552 as your use case is pretty specific and Krill has access to the codebase

[kind pebble]

Shivanshu@Shivanshus-MacBook-Pro ~ % curl -fsSL https://clawd.bot/install.sh | bash

🦞 Clawdbot Installer
Because Siri wasn't answering at 3AM.

✓ Detected: macos
✓ Homebrew already installed
✓ Node.js v22.14.0 found
✓ Git already installed
→ Installing Clawdbot 2026.1.24-3...
→ npm install failed; cleaning up and retrying...
Shivanshu@Shivanshus-MacBook-Pro ~ %

need help with this

[ornate tinsel]

why does clawd bot keep mentioning a mobile app 😭😭😭😭😭😭😭

[stiff tusk]

try installing via brew brew install --cask clawdbot

[viscid goblet]

anyone keeps getting no ouput & randomly get unresponsive

[rigid vigil]

hi, i heep getting this error, anyone know why or how i can make bearer not required? HTTP 401: authentication_error: Invalid bearer token (request_id: )

[somber scroll]

are you trying to use your Claude sub on a user account that doesn't have keychain access (non admin for example)?

[rigid vigil]

I might be, how can i enable that?

[somber scroll]

**Slack thread replies not triggering agent
**
I've got Slack connected with socket mode and message.channels event subscribed. Main channel mentions work fine, but thread replies that @mention my bot don't trigger the agent.

What I've tried:

• Confirmed message.channels, message.groups, app_mention events are enabled
• Set requireMention: false on the channel
• Gateway restart after config changes
Config:
"#studiomanagement": {
"allow": true,
"requireMention": false
}
Thread replies show up in Slack but don't invoke the agent. Is this a known limitation? Any workaround?

Running 2026.1.24-3

[somber scroll]

you can run it on the admin account of the machine if you've got nothing else of any value running on it and not logged in to anything sensitive on that machine (like an old laptop or macmini you had lying around which isn't in use for anything important), or you can switch to using the Claude Code API but that's usually more expensive.

[brave igloo]

Hi all, a new user here on clawd using Claude sonet 4.5. Anyone got issue with rate limit and context overflow: text too large?

[silent stag]

Hi. I watched some videos and couldn't find a definitive answer. Can you use clawdbot with claude max subscription (i have the 5x one) ?

[lean knoll]

guys, can anyone help me out here
https://discord.com/channels/1456350064065904867/1465260336796930171

[amber mortar]

Hi why does the code that Clawdbot produces always seems to be worse and less "thoughtfull" than Opus 4.5 in Claude Codeß

[austere bluff]

hi, i am looking to run clawdbot, if i buy a 1 vcpu 1gb KVM(burstable) will it be enough?

[stiff tusk]

should be ok, but look for 2GB RAM at least

[austere bluff]

Okay

[stiff tusk]

I'm running mine on 2vCPU/4GB RAM and my VM mostly idling

[austere bluff]

are you using dedicated vm or kvm?

[stiff tusk]

low-priority kvm

[austere bluff]

which provider are you using? I am thinking about getting one in racknerd

[stiff tusk]

i'm running mine @ hetzner

[austere bluff]

is it cheap compared to racknerd?

[stiff tusk]

dont know about racknerd, but I'm paying €6.5 a month and i'm happy

[austere bluff]

i can afford this, i will checkout hetzner as well

[floral sand]

Is there a known issue with plugin cli command registration >=2026-01-23-1?

[stiff tusk]

seems so, yeah
clawdbot voicecall... produces error: unknown command 'voicecall' even tho its documented 🤷‍♂️

[floral sand]

I've been fighting with this since yesterday when voicecall was working but now I've got the same. Wanted to verify if just my install was screwed up somehow. Thanks. Do you know if there's a GitHub issue on it?

[stiff tusk]

nope, didn't check github for the issue yet

[stiff tusk]

it works when the bot uses voice_call tool, so I'm getting a call via Twilio, but the content wasn't what I expected... still can't make it talk 😄

[floral sand]

What I'm seeing is CLI commands look like they could be broken for all plugins... but I don't want to jump to conclusions or file garbage tickets... Is anyone else seeing broken CLI commands with voice-call or any other plugin?

[stiff tusk]

and there are issues will call ending as well

[floral sand]

I'm working on a plugin that needs CLI in order to login and authorize.

[fickle depot]

https://github.com/clawdbot/clawdbot/issues/2151

did anyone have this problem? Wondering if there is workaround

[austere bluff]

hi, if i use vps to run clawdbot and want to migrate to another vps what are the files i should copy?

[tough marsh]

hey ya'll i'm setting up my clawdbot but i'm getting this error on the gateway.
i've installed the cli

[stiff tusk]

.clawdbot and clawd from your homedir

[sage condor]

Hi guys, can we use our Cursor usage in Clawdbot?

I got Cursor Ultra plan and I really want Clawdbot use it's tokens from that package if its possible?

[glacial maple]

Apple Notes on a Linux VPS Machine
I am hosting clawdbot on a VPS linux machine. My main machine is a macbook pro. I want clawdbot to be able to read/write Apple Notes. I understand that I should connect my macbook as a Node to the Gateway to do that, but the drawback to that, I believe, is that the Mac would need to be always on/awake. My use case, though, would be to use clawdbot when on the phone (iOS as well btw).

[stiff tusk]

I'll give you a little hint: you can ask exactly that your own clawdbot, and it will even prepare everything for you automatically :)👍

[austere bluff]

🙂

btw which model you are using?

[stiff tusk]

minimax 2.1

[austere bluff]

i already have minimax 2.1, is it working fine? i know it wont work like opus, but is it good for clawdbot?

[stiff tusk]

no problems so far. i'm pretty satisfied. will try other models of course to compare, but rn MiniMax2.1 is pretty solid... and it's cheap 😄

[austere bluff]

yeah, i have a started package in minimax, i guess its enough for clawdbot

[stone dome]

Hey there i've encountered the following problem: ⚠️ Agent failed before reply: Unknown model: anthropic/claude-sonnet-4.
Logs: clawdbot logs --follow. Can anybody help?

[tacit slate]

how much is minimax?

is it free

[austere bluff]

10 $ per month for starter pack

starter pack is enough for coding related task, so i assume it would be enough for clawdbot as well

[stiff tusk]

i'm using direct API with MM2.1
also very cheap

[tacit slate]

can we use groq?

how can we implement that into clawdboy

[stiff tusk]

my clawdbot says it supports groq

[agile topaz]

how to use direct API? I'm new to this, still trying to configure my qwen api into this

[tacit slate]

huh where

[stiff tusk]

Login to minimax.io platform. add some $ to your balance, generate API key, copy-paste it when configuring CB, set it to use MiniMax2.1. you're done.

[stiff tusk]

clawdbot config set providers.groq '{"baseUrl":"https://api.groq.com/openai/v1","api":"openai","models":[{"id":"llama-3.3-70b-versatile","name":"Llama 3.3"},{"id":"mixtral-8x7b-32768","name":"Mixtral"}]}'

[tacit slate]

raspberry@Khatib-Pi:~/clawd $ clawdbot config set providers.groq
error: missing required argument 'value'
raspberry@Khatib-Pi:~/clawd $

[stiff tusk]

try pasting the whole clawdbot config set providers.groq '{"baseUrl":"https://api.groq.com/openai/v1","api":"openai","models":[{"id":"llama-3.3-70b-versatile","name":"Llama 3.3"},{"id":"mixtral-8x7b-32768","name":"Mixtral"}]}'

I haven't tried this myself yet tho

[atomic raven]

I can’t access the Clawdbot browser. The Gateway runs correctly and shows the browser control server on 127.0.0.1:18791. The Chrome Browser Relay extension is installed but fails to attach (icon turns red). I’m running the Gateway on AWS EC2 and accessing it via SSH port forwarding. Please confirm whether this setup is supported and how the browser should be connected.

[brisk valley]

Anyone figured out how to fix Cron jobs ?

[silent stag]

Does anyone know if you can use claude subscription or it must be api?

[median ivy]

you can use the subscription, even subscriptions for other services

[median ivy]

I think there are multiple issues related to cron, I'm on a Raspberry so I could be seeing some that are linux-only

[brisk valley]

Yeah my issue specifically it looks like the Cron is starting but sending the msg to telegram fails dunno

[median ivy]

(multiple as in: different ways in which parameters required for cron.add are missing, jobs non firing sometimes,ecc....)

do you get the "An unknown error occured" message?

[silent stag]

Thank you 🙂

[shy sigil]

Hi All,

My ClawBot not answering in chat after install, Installed in the Local directory and auth Claude Code via browser. Any tips?

[crystal oyster]

Yeah same here.

When I send message it wait for 1 sec and responsed with actually nothing, there's no error found in Logs

[median ivy]

that is hard to debug.... I was trying to fix it a few days ago when I identified a commit from Krause on normalize.ts that refactored a check for a missing payload.kind moving it somewhere else but there are basically 0 info in my case in the logs... and at some point ti started working again, and then stopped working again.... I understand that the issue could be the model not being able to allway calls cron.add with the correct message structure but it's hard to debug...
The main problem being not enough logs to figure out what is happening or understanding where some error is from (yep, I could submit a PR)

[median ivy]

yeah, the last few days I'm getting some strange "empty responses"

[median ivy]

but btw, adding jobs with the clawdbot web gui works, so there is a workaround

[slender pulsar]

@floral sand did you manage to find a fix?

[floral sand]

@slender pulsar not yet... are you experiencing the same bug? With which plugins?

[slender pulsar]

voicecall lol error: unknown command 'voicecall'

It's enabled in my plugins though

% clawdbot plugins info voice-call

🦞 Clawdbot 2026.1.24-3 (885167d) — Ah, the fruit tree company! 🍎

Voice Call
id: voice-call
Voice-call plugin with Telnyx/Twilio/Plivo providers

Status: loaded
Source: /opt/homebrew/lib/node_modules/clawdbot/extensions/voice-call/index.ts
Origin: bundled
Version: 2026.1.24
Tools: voice_call
Gateway methods: voicecall.initiate, voicecall.continue, voicecall.speak, voicecall.end, voicecall.status, voicecall.start
CLI commands: voicecall
Services: voicecall
% 

I started to think I was the crazy one then I saw you also have the same issue

[floral sand]

Yes this seems to be a bug that appeared after 2026.01.16-2.

It's not clear to me if it applies to all plugins or not but the fact that it applies to a bundled plugin suggests it's a real core bug.

I'm trying to figure out a fix now.

[wind flicker]

Hey guys how can I change the model from claude (hit rate limit) to openai?

[sullen knot]

will z.ai coding plan subscription work?

[floral sand]

Use the /model command.

[wind flicker]

ty

[undone inlet]

how high should i set context length to in lm studio? i keep getting a context length error with gpt-oss-20b

[floral sand]

@slender pulsar @stiff tusk #1465038352876044623 message This is working. Not sure if it's the best fix but it is working.

[glacial maple]

Apple Notes on a Linux VPS Machine
I am hosting clawdbot on a VPS linux machine. My main machine is a macbook pro. I want clawdbot to be able to read/write Apple Notes. I understand that I should connect my macbook as a Node to the Gateway to do that, but the drawback to that, I believe, is that the Mac would need to be always on/awake. My use case, though, would be to use clawdbot when on the phone (iOS as well btw).

[floral sand]

I wonder if there is a way to let your Clawdbot access and write notes through icloud.com - I bet there is.

[glacial maple]

apparently theres not 😢

at least nothing that i could find

[floral sand]

@glacial maple Apple definitely let's you access your notes through icloud.com if you "Allow web access to iCloud data" in your device settings. Clawdbot can access websites. It might not be the cleanest approach but it should be possible.

[median ivy]

and you can configure a fallback module, I asked clawd to do it on its own and it updated its config

[floral sand]

There is a Snap app “iCloud Notes” and broader “iCloud for Linux” wrappers that expose iCloud Notes in a native-like app.

Your service could interact with that client indirectly (e.g., by writing local files or using whatever storage the client uses, if documented), but there’s no stable, supported API surface to call.

[glacial wyvern]

I've just read Alex Volkov's tips on Clawdbot (https://x.com/altryne/status/2015229089574228169) ..
And wanted to ask how to add a Grok 4.1 to CB and what skill to use? 🙏

Alex Volkov (Thursd/AI) (@altryne)

⚡ @clawdbot tips and tricks : stream of consciousness style

X · Jan 25, 2026, 1:03 AM

[wind flicker]

Is memory persistent across primary and fallback models?

[floral sand]

It is.

[wind flicker]

Ty gang

[median ivy]

yes

[meager meadow]

i'm really having loads of trouble getting Clawd to reply to me using venice models through the default onboard

can someone lend me a hand'

[gritty reef]

hey guys quick question, if i use the codex oauth will it use my codex quota or is there a way to use the normal chatgpt 5.2 quota? thanks!

[proud dust]

Hello guys! Pretty new to this

I tried to onboard multiple times but I don't get any answer when I test it I don't know why, even though it's supposed to be logged to my claude account and it says it's connected

Also, it's an old mac mini I had for some time now, but stopped using few years ago (still got data on it)
Is it best to reset it beforehand?

[floral sand]

Have you tried some other providers just to see if it works at all?

Have you tried running clawdbot doctor?

[white cipher]

Hey! Having issues with Control UI on Railway deployment. Getting persistent "disconnected (1008): pairing required" error despite following the Railway setup docs.

Seems like gateway generates its own token that doesn't match the CLAWDBOT_GATEWAY_TOKEN env var. Tried editing /data/clawdbot.json via SSH but it resets on redeploy.

Any workarounds for Control UI on Railway, or should

Filed issue: https://github.com/clawdbot/clawdbot/issues/2205

[lament flare]

Hey y’all! I currently have two Raspberry Pi’s in a k3s cluster serving up some apps. Would it be possible to deploy Clawdbot through k3s as well?

[devout nacelle]

hey all, how to fix this telegram issue? HTTP 401: authentication_error: Invalid bearer token (request_id: req_011CXVzFqFeAT8EVJ9bHRimz)

[median ivy]

I don't see why not

it doesn't create additional containers for itself so it should run

[glossy iris]

For Clawd, do I have to buy Claude API credit or a Claude subscription?

[lament flare]

what better way to get it working then let claude do its thing?

[floral sand]

Both work.

[glossy iris]

Which one is the best? And which one is cheap and has the best performance?

[floral sand]

You want a subscription. You will be bankrupt in the next 48 hours if you use API 😂

But beware Clawdbot if used heavily can exhaust even the $200 plan.

For getting your bearings MiniMax-M2.1 $20 plan works great.

It's not as smart as Opus-4.5 but it's not terrible and it's fantastic for the price!

[glossy iris]

Thank you

[paper pilot]

Reposting here for visibility:

I'm trying to get my agent to consistently use the message tool to send multiple short messages instead of one big block (feels way more natural in chat).

The instruction is in AGENTS.md and SOUL.md, but Claude's base training defaults to responding in one complete message - so it keeps forgetting to use the tool despite clear instructions.

I was thinking: if I could customize the prompt that fires on /new or /reset, I could add a reminder like "Remember to use the message tool for multi-message replies" right at session start.

Currently the reset prompt seems hardcoded ("Say hi briefly and ask what the user wants to do next").

Questions:

1. Is there a config option to customize the reset/greeting prompt that I'm missing?
2. Any other workarounds people have found for reinforcing specific behaviors at session start?
3. Would this be worth a feature request?
   Thanks! 🤠

[devout nacelle]

hey all, how to fix this telegram issue? HTTP 401: authentication_error: Invalid bearer token (request_id: req_011CXVzFqFeAT8EVJ9bHRimz)

[paper pilot]

Make sure ur bot id is correct

[karmic gale]

how’s everyone dealing with compaction? At the point it happens so often I’m finding myself having to start over in conversations nearly every ten minutes… recoverable, of course, but I’m sure there are better ways of handling context

[floral sand]

Speculating here so take this with a grain of salt... Most of these models are designed to stream their reply so short bursts of replies would make little sense. Not 100% sure but getting multiple short replies I think requires multiple requests which means sending context again - this would blow up your quota really fast. I am not sure if any messaging apps can accept the streams or not. It might be possible to create something that catches the streaming reply from the model and sends it out in short messages but not sure it's worth it.

[devout nacelle]

where do i set that?

[floral sand]

What model are you using?

What's the context window? Sounds weird if it's happening every 10 minutes unless the model has a tiny context window

[devout nacelle]

where can i set my telegram bot id?

[paper pilot]

I thought so too:

[paper pilot]

In the onboarding, did you do it?

[devout nacelle]

i did yep

[paper pilot]

And you selected telegram there?

[median ivy]

try being more explicit stating "maximum telegram message size".... had the same issue and it seems to work.... for now....

[devout nacelle]

yep now i set it correctly, i am no longer getting that error

but anything i type in telgeram nothing happens

[floral sand]

Most things that the onboarding gathers get set in ~/.clawdbot/clawdbot.json from what I understand, so if you want to check it you can look there. Just back it up before you screw with it.

[rustic linden]

Hello all,
Can anybody provide a cost estimate (I know "it depends") of what it does cost to use Clawdbot for a day using Opus 4.5 API?

[floral sand]

You can run clawdbot tui from the terminal and ask it to help you fix itself.

[floral sand]

run clawdbot tui and ask it to help you debug the telegram connection.

[floral sand]

A fortune. Don't do it. Get a subscription.

[devout nacelle]

this is what i get

[paper pilot]

I checked:

[karmic gale]

Opus4.5

[rustic linden]

https://tenor.com/view/faints-fresh-prince-fell-down-falling-down-fall-gif-17320032

[umbral storm]

I'm struggling to get a cron job working correctly. I can define the job and have it run, but it seems to just stop half way through. I'm asking it to do some searches on a topic and give a report, but it seems like it just stops half way through the job, acting like it finished but in reality it didn't deliver anything other than "let me search according to the preferences:"

[rustic linden]

Thank you!

[paper pilot]

Claude subscription makes it 20x cheaper

[rustic linden]

Thanks guys

[devout nacelle]

any idea why it says no output?

[floral sand]

Very strange. That has a 200K context window. You can go to ~/.clawdbot/agents/main/sessions/ and look at the session logs and try to reverse engineer what is filling up your sessions so fast.

[floral sand]

Try asking again a couple of times. If it doesn't work try another provider.

[wind flicker]

Hey guys it seems clawdbot keeps crashing/shutting down the process that is running it? Not even sure where to begin looking for the culprit. Anyone have ideas?

This usually happens when I prompt it to do something

[floral sand]

What OS are you running on? And what are your system specs?

[wind flicker]

Windows and running on a beastly gaming rig I don't think specs is the issue

[floral sand]

The only time I've seen it crashing like that is when the node heap is too small.

I haven't used windows in more than a decade... but I believe node heap is the same on all platforms. You want 3072 or 4096 to run smoothly in my experience.

You can check the logs too of course to see if ti says anything about why it's crashing. I dont' know where the logs are on Windows. On linux on my install they are in /tmp

[wind flicker]

Ok thanks for the advise Ill take a look!

[dense girder]

Can I use Opus through antigravity? Anyone manage to make it work?

[median ivy]

I do use gravity... but with gemini, any particular issue with opus?

[stiff tusk]

why is everyone so obsessed with Opus 🤷‍♂️

[fickle escarp]

it's really good probably 😄

[stiff tusk]

I'm sure its amazing, of course, but there are a lot of cheaper alternatives 😄

[fickle escarp]

need to use all the models imho.

[dense girder]

Getting a 401: invalid_client error

[proud dust]

Do people use it with their personal icloud or a new one?

Because I see people saying they make a new account of everything but then what's the use of it

[fickle escarp]

it's to seperate concerns. depends if you want the bot to replace you or be an assistant.

also it has it's own imessage which you can chat with then. a lot of people seem to use that as their main chat client (mainly US i think, europe whatsapp, asia line, etc.)

[glad ibex]

I have a strange issue:
I created a new Claude account to use with Clawdbot, switched my auth to use that (via claude setup-token) and since then I've been getting this error:

⚠️ Agent failed before reply: No available auth profile for anthropic (all in cooldown or unavailable).. Check gateway logs for details.

I checked the logs and it says the token for that account is on cooldown. I tried it again as soon as the cooldown was over and I got the same error and it went into another cool down for 1 hour.

Does anyone know how to fix this?

[proud dust]

Thanks for the answer

Tbh I'd like it to assist me as much as it could lol

Idk if there's security concerns though

[fickle escarp]

yeah personally i don't want it to have access to my whole history. i'd rather treat it like a team member that i am give more and more access to my life over time.
mine has own apple id, gmail, github, whatsapp, hardware, etc. as it helps me to build the structure for a larger team. also security is a big concern and it should.

[fickle escarp]

did you try the search, there are some results for that.

[glad ibex]

Yep, found some questions, but not answers on how to fix it (atleast for Anthropic)

[jade musk]

Marning all. Anyone got a gold standard quick start guide pls? I have a spare Mac Mini M4 for The Right Hon Clawd. Best to set up in docker?

[median ivy]

does gemini 3 pro/flash work? (works fine for me, just checked switching the model, update: it burns token quite fast)

[floral sand]

Gemini 3 Pro had a bug where the "thinking" was leaking through into the chat. Not sure if it got fixed.

[paper pilot]

i am trying to connect to clawd with my macbook. my gateway is running on a vps. i downloaded from source and built and packaged the app. what am i doing wrong here? the ip address is correct, the ssh key file is correct too. why is it trying to connect to a localhost gateway? whats going on here?

[unreal bear]

Make a new one

Don give it access to your personal information

[analog birch]

Good Day, has anyone did a Proxmox installation? I am busy with it now. Just want to know if there are any pointers to look out for in advance

[median ivy]

fixed since yesterday, doesn't do it anymore for me

[paper pilot]

kind of pointless to use clawd if you do what you're suggesting

[unreal bear]

Maybe try this

1. Open terminal, run "claude" to open claude code cli
2. Inside cli, run "/login" to create keychain credentials
3. open a new terminal, then inside the new terminal, run
   cd~/clawd/skills/claude-connect
   ./refresh-token.sh --force

[still condor]

Quick question — is there a way to set a user avatar in the webchat control UI? I see ui.assistant.avatar for the bot, but didn't find an equivalent for the human side. Feature request if not!

[tulip hinge]

Hey people. Have anyone managed to set up clawdbot to use your google AI Pro subscription for Gemini? Does anyone know if it’s possible and how?

[outer storm]

configure, then go to gemini, gemini OAuth

but right now, my clawdbot is stuck on typing infinitely, it isn't responding to me (not giving empty responses, its literally just stuck on typing)

[tulip hinge]

I’ll try that

[median ivy]

gateway restart

[outer storm]

tried that

[royal wyvern]

I set ip local llm qwen3 8b. And it is messing up it’s own thinking and replies. Is it because the model is bad?

[median ivy]

ouch, that session got broken in some way? does /new do something?

[outer storm]

weirdly its now saying clawdbot gateway service isn't installed despite installing it again a couple of minutes ago

[outer storm]

nope it doesnt even respond to that

[astral garnet]

yes, local models give you crazy dumb output

you cannot use this with low context, small models who don't understand tools well.

[outer storm]

im so baffled its just not responding

[novel spruce]

Hi guys. just started with clawdbot today. awesome stuff. quick question. just wanted to activate brave search. when actiavting the API in brave search it seems that it is not allowed. any thoughts?

[astral garnet]

does anyone ever read TOS?

[median ivy]

/abort from the terminal tui? or if you use the website ui that the gateway exposes you can see and delete sessions

[outer storm]

now it just says no active run
connected | idle

[median ivy]

now doing /new you should start with a clean session that doesn't get stuck if the issue was the content of the session (long running task? task that ended in a reasoning loop?)

random "(no output)" happen to me too sometimes...

[outer storm]

Still stuck on typing, I typed 1+1= and its been a minute now and its still not responding

in tui

[median ivy]

and with gemini also some "Cloud Code Assist API returned an empty response"

[outer storm]

i mean mine isn't even giving some sort of output

its just stuck on loading / typing

even after your fix

[median ivy]

remaining tokens? quota is ok?

[outer storm]

0/1m tokens

[median ivy]

no I mean the subscription tokens

[outer storm]

I normally use OAuth but i swapped it a couple minutes ago to test if using the API would work better

[outer storm]

uh im not sure, how would I check?

[median ivy]

like what codexbar displays

I have no idea on how to find it on the gemini website... too messy, I just installed codexbar that show it in a way that a human can understand

[outer storm]

do you just mean how many tokens I have left?

with my api plan?

[median ivy]

in the subscription, yes

yep

[strange crow]

Please help me with this error:
HTTP 401: authentication_error: invalid x-api-key

[outer storm]

I still have all my tokens left it seems, every input has returned an API error

[median ivy]

and if you have gemini pro rember that Gemini Pro and Antigravity have two different quotas you can use

[outer storm]

I could try fully uninstall and reinstall clawdbot

[gritty reef]

hey guys quick question, if i use the codex oauth will it use my codex quota or is there a way to use the normal chatgpt 5.2 quota? thanks!

[outer storm]

is there a way I can save the context and scripts from when it was working?

[median ivy]

the sessions are just json files in clawd/ or .clawdbot/ , one of the two, if you were building script they are in clawd/

[outer storm]

dope ty

ill try reinstall and see how that goes

[median ivy]

(no idea how to restore them though, if copying the file is enough, never did it)

[outer storm]

well great I reinstalled and I still get the same thing

[zinc veldt]

any recommended browser to use with clawd?
maybe something that helps with captchas etc?

[dense saffron]

Hi everyone should the bot know his name after i changed it in the IDENTITY.md file? BOOTSTRAP.md is still there

[outer storm]

well this is odd, A new session was started via /new or /reset. Say hi briefly (1-2 sentences) and ask what the user wants to do next. If the runtime model differs from default_model in the system prompt, mention the default model in the greeting. Do not mention internal steps, files, tools, or reasoning.

[strange crow]

Config validation failed: channels.slack.dm.allowFrom: channels.slack.dm.policy="open" requires channels.slack.dm.allowFrom to include "*"

Any fix for this?

[median ivy]

if you don't move it I suppose it will use the same stuff from ~/.clawdbot from the previous install, maybe it's that

[outer storm]

I fully deleted that as well

[median ivy]

I'd say only one possibility remains, 0% quota . Since you did the oauth thing again when reinstalling the authentication token for that should be valid (but then, it should display an error...)

[minor scroll]

Hello guys, i'm trying lots of options but I can't get the gateway dashboard to work, with password, token, anything I get the same when trying to connect

disconnected (1008): pairing required

any ideas?

[outer storm]

i’ll try again tommorow and see if results are different

[glad ibex]

I don't seem to have claude-connect in the skills folder😅

[ornate forge]

ive got clawdbot up and running but i seem to have missed a step as it doesnt know who i am or who it is..

is it possible to rerun the Wake up, my friend! onboarding again? i have the feeling theres some personality setup i skipped?

[fallen token]

Im setting up Clawd bot on my laptop, (windows ) i thought i was pretty tech savvy have used all the llms but running into non stop problems. Whats the best tutorial out there for windows downloads

[wheat gyro]

I am trying to setup the android node... In the docunent it says there is android app. But I am not able to find any android app or android companion app for this

By android node app, does it mean chat app (whatsapp, telegram, discord)?

[thick river]

Also looking for the apk

[mental mica]

android is still very WIP, there's no published app yet

you can build it from the codebase if you want but there be dragons

[wheat gyro]

https://github.com/clawdbot/clawdbot/tree/main/apps%2Fandroid

To folks who wanted apk they can build from here
Instructions are in Readme.md

[worldly rivet]

Hi, I need help! I want to use my clawd bot in the Mac app with the canvas. But I don´t have it. I only can use it from the dashboard. I have installed clawd bot locally. No external connections yet. What can I do? thanks in advance!!

[glad field]

Anyone know how to configure the gateway token? I deployed on railway, getting this error on the gateway dashboard:
disconnected (1008): unauthorized: gateway token missing (open a tokenized dashboard URL or paste token in Control UI settings)

[old hornet]

a nyone use it with llamacpp yet?

[mint brook]

hi guys im having problem with clawdbot, does anyone has the same problem? he is sending me these msgs:
exception TypeError: fetch failed sending request
The AI service is temporarily overloaded. Please try again in a moment.

[hallow horizon]

Means exactly what it says, the service is overloaded and you need to wait

[honest ridge]

means ... you llm is busy ... wait until it has capacity or chnage the model / provider.

[whole root]

hey has anyone had a good experience with real time voice? im using twilio and openai tts and its kind of bad. starts good then quality nosedives

[shadow saffron]

how do i get clawdbot to do something it thinks is risky
'my safety precautions wont allow me to do this' ?? lol

[hallow horizon]

Different model

[whole root]

anyone try other models like GLM, minmax, Kimi k2, or deep seek?

[zealous raptor]

that's called a refusal and stop doing risky things 🙂

[old hornet]

want to try some of the new falcon h1r models

[shadow saffron]

if its not actually risky at all ?
its being silly

[obsidian fog]

Having an issue where if I do "clawdbot config" and try to add Obsidian as skill, I get an error saying "brew is not installed". I'm on Windows, so what am I supposed to do?

[pastel pike]

How can I get Gateway Token ?
I closed the token tab

[old hornet]

disconnected (1008): unauthorized: gateway token missing (open a tokenized dashboard URL or paste token in Control UI settings) might be in the same bod as mrshadow, also trying to figure out how to use llama.cpp and local models

[sour slate]

@old hornet, the fastest way to get your problem solved is to follow the instructions here: https://docs.clawd.bot/help/faq#im-stuck-whats-the-fastest-way-to-get-unstuck

@pastel pike, the fastest way to get your problem solved is to follow the instructions here: https://docs.clawd.bot/help/faq#im-stuck-whats-the-fastest-way-to-get-unstuck

[old hornet]

TY!

[sleek marsh]

Is the “Agent failed before reply: session file locked” a Docker only issue or everyone gets it?

[mental mica]

thats a docker issue where you're mounting its .clawdbot workspace as readonly

[odd swift]

lmk if anyone has a fix

[tulip dune]

hey guys quick question. will clawdbot run on windows 11 iot. I have been trying to get this to work on my machine for the last week. im trying to figure out whats going on the only thing i can think of is my OS. Before i go and wipe my machine and install linux i figured i should ask y'all first.

[warped dew]

Hi guys, I'm running on anthropic api on my linux server. keep getting (no output). Anything i can do to fix that ?

[odd swift]

that means there’s an error

[warped dew]

on anthropic side ?

[cursive pumice]

I'm really struggling getting clawdbot to function very well using Gemini CLI. Is this normal at all? Constantly getting "no output" if I cancel a request or stop it from doing work. Can't get any verbosity on thinking and it hangs a lot. It doesn't seem to be able to fall back from pro to flash on usage limits being hit. Barely able to get my laptop to talk to it (it took about 4 hours yesterday because it wanted "pairing" but provided no way to do it, ended up just using ssh). Am I doing something wrong or is this just how it goes right now?

[old hornet]

even turning auth off still geting the same error.

[elder hazel]

Heartbeats don’t seem to be firing. Anyone else dealt with this?

[mental mica]

gemini sucks

[sour slate]

@elder hazel, the fastest way to get your problem solved is to follow the instructions here: https://docs.clawd.bot/help/faq#im-stuck-whats-the-fastest-way-to-get-unstuck

[fluid void]

Guys im going insane here, I installed clawd in WSL like the docs told me but now i can't access the web ui from my PC.

Anybody knows how to fix it? The docs don't explain how to solve this issue

[sour slate]

@fluid void, the fastest way to get your problem solved is to follow the instructions here: https://docs.clawd.bot/help/faq#im-stuck-whats-the-fastest-way-to-get-unstuck

[fluid void]

that didnt work

can’t establish a connection to the server

[random pilot]

Please help me with this error:HTTP 401: authentication_error: invalid api key

[radiant elbow]

@olive stag responding to your question from #general there's a skill for nano banana pro. i added a veo skill and i think there's a few others that also do veo. check clawdhub. Search is currently broken unfortunately so you may have to just load all skills and manually search in page.

[old hornet]

thanks for the help, fixed the pasword/auth token issue but now trying to figure out how to use local installed llama.cpp models

[atomic kelp]

switch to Sonnet, they said, it'll be less expensive, they said.... 😂

could have bout a MacMini and been running from Claude Max subscription already ...... stuborn and still on Windows

Claude Connect only worked with Mac, correct?

[barren sparrow]

What is the best setup to have Clawdbot with Obsidian, when Clawdbot is hosted on Cloud? Having hard time to sync all, using a git plugin in Obsidian.
Should I move to Notion? Any other recommedations?

[lime lodge]

DOES I need cloud storage to run clawdbot on my windows system

[zinc mica]

You can't have both setup token and api key. That's what caused the bearer error for me. Also if you try with the setup token once and it fails you need to get a new one before you try again.

[stiff tusk]

Just asked Sonnet 4.5 for a 40 y.o. song lyrics... got some major bs about copyright infringement for an answer. Now I'm definitely not subscribing to Claude 😄

[jolly flower]

anyone run into models.providers.ollama.api: Invalid input when setting up Ollama?

[meager helm]

no, there are repos for transfer. But it'll never be the same regarless of how fluent they make it.

[floral sand]

Switch to a subscription or MiniMax-M2.1 if you're running out of subscription quota.

[floral sand]

It's refused to give me known public domain stuff in the past.

[stiff tusk]

imagine one day you'll need some vital info and get something like "i can't do that because it violates some ancient copyright law"

[gray kraken]

has anyone tried to install Clawde on a chromebook?

[lilac ibex]

Hey guys, have you managed to get clawdbot to work with Antigravity?

[fluid void]

anyone knows how to solve this issue? it got stuck here and nothing is happening

[twin snow]

does anyone has a good set up of local models that work well this clawdbot ? i have tried openai-oss , and qwen3-coder , and both sucks , help !

[grand tartan]

anyone have experience setting up ollama models with clawdbot?

[sweet moth]

i’m sorry that’s what i’m trying to figure out too clawd consumed $100 in last 12 hours at this point a human could do all this so gotta get off antrhopic api

you having luck with claude max subscription?

i’m thinking at these cost levels would a new mac mini with 32gb do 90% of the work and then use sub agents with api access for better programming models?

[fathom fog]

I want to talk with the ai on discord but it asks for approve and got a code, what do I do?

[opal sorrel]

Hi friends — new here but loving my new crustacean friend. A couple questions I’m hoping for answers to:

1. The macOS app seems to spawn the gateway as a background process, so it doesn't inherit the Full Disk Access permissions that you give it, which makes it impossible for it to read and respond to iMessages. How have people addressed this?

2. Is there a TestFlight link for the iOS app, or are people building it from source?

🙏🥰

[sour slate]

@steady marten, the fastest way to get your problem solved is to follow the instructions here: https://docs.clawd.bot/help/faq#im-stuck-whats-the-fastest-way-to-get-unstuck

[mental mica]

1. grant it the permission separately when it requests it
2. its still WIP in source

[sour slate]

@fathom fog, use #1459642797895319552 for help. The fastest way to get your problem solved is to follow the instructions here: https://docs.clawd.bot/help/faq#im-stuck-whats-the-fastest-way-to-get-unstuck

[opal sorrel]

Thanks. It asked for a bunch of perms but not FDA. Will retry. You recommend building from source or installing the cask from
Homebrew? I did the latter maybe that’s the issue.

[mental mica]

mac app is stil extremely WIP

build from source

with code signing

[opal sorrel]

Ok perfect will do. Had Clawdbot build me a janky wrapper that I could grant FDA to and that works for now, but wanted to move to what’s being worked on community wise. Seems wiser long term 🥰

[stark badger]

hi there!

i'm a non technical user, someone can help me to install and set clawd ?

is there any .dmg to install on mac?

[weary elm]

Ok perfect will do. Had Clawdbot build

[atomic kelp]

i have not been able to figure out how to link Claude Max subscription to Clawdbot specifically for Windows. Claude Connect works fine for Mac users.

[trail moat]

How the hell do you get Minimax to work? I set up everything correctly and still no output.

[harsh lodge]

Hello. Is there a formal way to add my clawdbot to the server so it can give me daily summaries? I have trouble keeping up with the notifications.

[sour slate]

@harsh lodge, your question is answered in the server FAQ in our Server Guide

[sullen knot]

how to integrate deepgram, anyone got success with STT & TTS?

[uncut burrow]

hi guys i'm just playing around with clawdbot on a ec2 from AWS but it's seems so slooowww even with gemini flash

do you recommend to go to railway or hetzner vps ?

[fierce heron]

Guys how do I add whatsapp channel ?

I am new to this stuff

[lavish ginkgo]

Has anyone been able to get the browser extension working? is there a better way to do this?

[fluid void]

What do i do if i get this?

Error: No tool-capable OpenRouter free models found.

there is no way this is true

[balmy spindle]

I have the same error when using tailscale. Only way I can resolve this is by clawdbot config set gateway.controlUi.allowInsecureAuth true which doesn't feel great -- althought still prob "secure"

[hasty mural]

Hello everyone sorry if this is a noob question can’t seem to find an answer. Keep getting an error saying that I’ve hit my rate limit, even though I am using Claude Pro.

[weary dew]

My clawdbot keeps breaking its own tools, and doesn't know how to fix it

[fast kayak]

Hey all, if anyone needs a hand setting up clawdbot just lmk

[brisk valley]

Damn i respect the husstle

[thick wind]

keep getting this error- any known fixes?

sorry completely new to clawdbot specifically, first time trying

[lethal linden]

Setting up Clawdbot in a VPC on AWS. Anyone opt for that method?

[brisk valley]

u have npm installed?

[fast kayak]

you know it

[signal plume]

anybody using discord having trouble getting a response, my bot is connected and it can see the mentioned messages and i can force a response from the webchat but it fails to respond in discord directly

[thick wind]

yes. I think it may be a permissions issue. Let me see if I can troubleshoot*

had to run sudo chown -R 501:20 "/Users/username/.npm"

all gucci

[dry mica]

Hi! Anyone facing this issue
: « LLM request rejected: messages.7.content.3.tool_use.input: Field required »

[zealous raptor]

google?

[ionic kraken]

hey All, I am facing this issue when talking to the bot. If it uses a tool to edit a file or something, it then goes completely silent, and I have to prompt it to say "hey did it work, are you done" etc. Always the same, for instance it edits the clawdbot json file and then just goes silent on me

[rocky orbit]

You’ve hit rate limit and are unable to use your LLM sub for a period of time

[rocky orbit]

Did it really edited the file? Sometimes when asking to do tasks it can take some time

[spice heart]

Anyone figured to install local tts instead of elevenlabs?

[ionic kraken]

Yeah it did actually edit the file. Does it a lot

[rocky orbit]

I didn’t install eleven labs and it used TTS automatically

[ionic kraken]

This is typicallly when it just stops talking 😛

It then promises to talk to me after tool use, but never does LOL

[rocky orbit]

What’s your channel @ionic kraken ? Can you check gateway logs and send it to #1459642797895319552 ?

[spice heart]

What? How lol

I have added elevenlabs api and on telegram it just sends me voice notes, burning tokens

[rocky orbit]

I don’t really know tbh I’ve just asked him to send vocals yesterday and it did via TTS lol

[ionic kraken]

Webchat (clawdbot-control-ui)

Issue: Text I generate after tool call results doesn't reach the user. Text before tool calls works fine. Pure text messages (no
tools) work fine.

Reproduction:

1. Ask me to edit a file
2. I say "Doing X now" → tool runs → I say "Done"
3. User sees "Doing X now" but NOT "Done"

The gateway logs above show mostly voice-call activity and restarts. The webchat connection logs show normal connects. No errors
related to the silent-after-tools issue.

[spice heart]

Im sure it must be using something? Let me dig in

[rocky orbit]

I suggest not using the web UI to discuss with your bot, it’s far from being the best channel

[lone tangle]

hi when trying bird I get this error:

{
"exitCode": 1,
"timedOut": false,
"success": false,
"stdout": "",
"stderr": "📍 CLI argument\n❌ Failed to post tweet: Authorization: You have reached your daily limit for sending Tweets and messages. Please try again later. (344) (344)\n",
"error": null
}

Any fix?

My clawd runs on a raspberrypi but runs the bird on my windows (my windows desktop pc is connected as a node)

[rocky orbit]

Bird is not made to send tweets

X api is blocking it, only to read them

[ionic kraken]

The same behaviour happens on Telegram

[ionic kraken]

II'll dig around in the code and see if I can debug it

[rocky orbit]

Yes use you local agent

[topaz pike]

are you publishing sboms/signatures for all signed artifacts yet?

[mental mica]

no

all the apps are still in unfinished prototypes or betas

[tired locust]

how do you get clawd to just do things? like at specific times?

do you have to ask it to set a cronjob or anything like that?

[zealous raptor]

cron or heartbeat

yes ask your clawd

[lone tangle]

Everytime my Clawd is reading tweets, he sends me the tweets and then he gets a system message which is telling Clawd that the "async command is finished and tell the user about the result" which causes Clawd to AGAIN tell me tweets
anyone have an idea on how to fix? its really annoying

[sour slate]

@lone tangle, the fastest way to get your problem solved is to follow the instructions here: https://docs.clawd.bot/help/faq#im-stuck-whats-the-fastest-way-to-get-unstuck

[true mesa]

I'm trying to use the magic one-line installer that works everywhere and I get the following error message. I'm installing on a Windows virtual machine under VMWare, I'm running Powershell as an Administrator with elevated permissions.

[*] Installing Clawdbot@latest...
node.exe : npm error code ENOENT
At line:1 char:1

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

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

[ionic kraken]

Mystery solved, when Clawdbot edits the clawdbot.json file it triggers a gateway restart that aborts the response. Maybe the restart for wait for in-flight responses to complete.

[glacial jay]

Anyone know if Clawdhub is down? The skill keeps giving me "{"code":"[Request ID: d9ee5c08035210e9] Server Error"}
Error: {"code":"[Request ID: d9ee5c08035210e9] Server Error"}"

[rocky vigil]

Getting limits all the time moved down to Haiku. Still too many tokens, any suggestions

Paid for more tokens and tier 2, do i have to wait a week?

[heavy rapids]

Hello everyone, I can't seem to find how to send a message on WhatsApp to 2 recipients?

clawdbot cron add \ --name "Morning status" \ --at "2026-01-12T18:07:00Z" \ --session isolated \ --message "Summarize inbox + calendar for today." \ --deliver \ --channel whatsapp \ --to "[phone]"

I have tried the following for 2 numbers:

--to "[phone],[phone]"

Instead of --to, I even used --target but no luck. Can you help me out?

[zinc mica]

Has anyone used elevenlabs API on free tier to give their clawdbot speech capabilities and how token heavy is it?

[floral sand]

Does anybody know what causes "no output" in TUI?

[rigid vigil]

Has anyone got the Claude Max Plan to work with oAuth without going through the Api key. I know there has been some issues recently. I haven't got it to work yet. Thanks

[rough galleon]

i did and then it stopped working entirely and can't get it to work again lol

[rigid vigil]

What have you been using since then for your model?

[past vault]

Agent failed before reply: Copilot token exchange failed (someone please help with this error?

[whole root]

party is over on max plan :

[junior glade]

Does anyone have himalaya working with ubuntu install?

[minor scroll]

thanks for responding

[rough galleon]

just tried uninstalling and reinstalling clawdbot and trying with codex but its still not runnign for me now lol

[fringe dirge]

Clawd wanted me to pass this bug to the devs, but didnt see the best place to put it but here it is.
Version: Clawdbot (npm-installed, latest as of 2026-01-26)

Issue: WhatsApp pairing codes are being automatically sent to contacts who message the linked WhatsApp account, not just the owner.

What happened:
1.⁠ ⁠A non-owner contact messaged my WhatsApp
2.⁠ ⁠Clawdbot automatically generated and sent them a pairing code
3.⁠ ⁠This was logged in ⁠ ~/.clawdbot/credentials/whatsapp-pairing.json ⁠

Expected behavior: Pairing codes should ONLY be sent to owner numbers defined in config.

Actual behavior: Pairing codes are sent to any contact who messages, regardless of whether they're an owner.

Security impact: High. Anyone who messages the WhatsApp account could potentially receive codes to link devices.

Workaround: Manually clear the file: ⁠ echo '{"version":1,"requests":[]}' > ~/.clawdbot/credentials/whatsapp-pairing.json ⁠

[nimble valley]

I’m using Gemini api and telegram and my bot just won’t reply some times or take super long to reply. Anyone fix this? I heard a few ppl w this problem

[zinc mica]

You have to remove the API key, only have the 0Auth setup token. If you have used the setup token and have failed you have to generate a new one.

[rigid vigil]

This is currently working now and hasn't been restricted by anthropic?

[zinc mica]

Implemented last night at 3AM EST, currently running it this way.

[rigid vigil]

Great, how can i remove the current api key? I tried to generate a new one but that didnt seem to work, maybe i missed a step?

[sacred swallow]

Hi guys

I started getting this suddenly - was working fine in the morning / I have restarted gateway, did update of clawdbot but no success

“I am still receiving errors that prevent me from executing checks.

The tools are requiring parameters (sessionId) that are normally handled automatically in this environment, and they are not accepting standard commands like ls -la. This indicates the tool system is still malfunctioning and has not recovered despite the gateway being connected.

I am physically unable to check the files at this moment.”

[rigid loom]

why is opus hallucinating so much in clawdbot man

purely making shit up

[zinc mica]

you get the setup key in your local terminal you are using. Open another instance and use claude setup-token . DON'T startup claude code in terminal.

[zinc mica]

uou'll need to auth in browser and then the setup token will be in terminal.

[zinc mica]

then access whether you're hosting clawd.

[zinc mica]

Then send: clawdbot models auth remove --provider anthropic **then paste: **clawdbot models auth paste-token --provider anthropic then when prompted paste the set up token:

[zinc mica]

then verify: clawdbot models auth list **if successful restart the clawd gateway: **clawdbot restart

[balmy jungle]

I setup Clawd and right away had API credit issues, which is odd since I have the Anthopix Max plan. So they I tried to change to change to Claude Sonnet 4 while I figured out api issues. Now I am stuck here
Agent failed before reply: Unknown model: anthropic/claude-sonnet-4

Suggestions on fix? My api issues should be resolved but somehow not sure how to fix this one

[novel spruce]

i did that today on a ec2 machine with 2 cores and 8 gig memory. running ubuntu. not very fast. but works.

[shadow tree]

I would like to use nim nvidia api, anyone know how? thanks (im still lost after reading everything..

[zinc mica]

did the same but with 1x 16GB

[shadow tree]

technically it is openai api but just the different endpoint..? idk if im right

i just dk where and how to set it

[lethal linden]

Excellent! Good to know. Any particular hiccups you ran into that if you had to do it all over again, you'd make sure to take a different approach?

[shadow tree]

or is there a place with relevance info that i can check?

[rigid vigil]

Thanks for this! I got an error though - error: unknown command 'remove'

[shadow tree]

i tried setting the clawdbot.json but it doesnt work

[next goblet]

has anyone set up using deepseek api? i get no errors when running the fix command but my chats are not generating replies and my deepseek dash is not showing any requests, so i think there is something wrong with how im labeling the model version in the json file?

[zinc mica]

that just means you don't ahve that command, try the same with clear.

[gentle nexus]

Anyone know of a good thread or convo about setting clawdbot up with just ollama local. Been going around in circles. I burn all my credits on cluade every 5 hours

[rocky orbit]

I was away today so unless there was an update, it's not compatible with Deepseek as of right now

[next goblet]

damn 🙁

[rigid vigil]

ok, i tried clear but i got same error. How can i get this command? Thank you!

[novel spruce]

worked pretty smooth. if i would run it 24/7 i would probably choose something different than ec2. i had good experience with contabo vps in the past.

[novel spruce]

how is the speed for you ?

[gentle nexus]

Trying any tooling one like qwen or glm. Clawdbot get really confused and burns my credits. Really stuck with it.

[lethal linden]

Really appreciate the insight! I'll be sure to let you know how it goes

[faint dirge]

Hello, I'm getting this error, can you help me? I'm trying to use GPT 5 Mini.

[rigid vigil]

i did this and it seemed to work..clawdbot models auth order clear --provider anthropic

[radiant dome]

I posted this to help before seeing this section. New user, got Clawd up on AWS (Ubuntu) but can't seem to get it working on Macbook. Gateway won't start, having trouble (I think) with Claude oauth to my Claude Pro account.

Is anyone available to help $$. I know your expertise is valuable and I'm wasting hours and hours here. Sorry if this is not allowed. 🙂

[hard shore]

hi guys I just did my first clawd deployment and it went well. However my tokens on openai went bananas 😄 is there any way to use a light weight model I can use locally on a VPS alongside clawd? Otherwise I'll have to set up a second partition on my gaming pc with an RTX3090 like Ollama I guess.

[distant grove]

Hello i am trying to setup gmail with clawd and it kept saying gog not installed. even though it is installed. i am using windows . i have tried every possible fix

[gentle nexus]

Yep where I am stuck just need a backup. Burning credit and allowance in minutes.

[next goblet]

is openrouter considered the best route to go with to keep costs down and compatible?

[zinc mica]

hmm, it's version dependent. You could try to override the key, should delete api key and ask for set up token: clawdbot models auth paste-token --provider anthropic

[hard shore]

Really have to be careful otherwise you'll be homeless in no time lol

[zinc mica]

honestly don't hate it, I'm going for more of a real assistant feel.

[gentle nexus]

Yep, had xai working on Friday, burnt through 30 buck, done 20 bucks on api creds with clause, open ai has run out so have to wait 5 hours. Every so often can’t even use opus as get 5 mins convo and done

[hard shore]

I think the only way to use this thing efficiently is to use ollama hosted locally. But this will need proper hardware.

I had ollama running on my RTX3090 before and it worked slow but ok.

[gentle nexus]

I have ollama runnig here, but I can’t get it working. It’s only for basic task as well.

[hard shore]

What GPU?

[gentle nexus]

Just. 3060

[hard shore]

My guess is it just loads forever

Damn it AI hardware masterrace wasn't a bubble ffs

[gentle nexus]

Works ok though. Don’t need fast . Just bit by bit. Can chat good. Very small coding tasks and it’s ok. Like single file python jobs. It works and reasoning is ok.

[hard shore]

so your issue is connecting it to clawdbot?

I read somewhere you have to edit the clawdbot config .json and use the open-ai connector but I'm still researching

[gentle nexus]

Getting clawdbot to talk to ollama, it think it is but it’s not. Gone over my clawdbot.JSON so many times.

[hard shore]

damn I'll probably face this issue then as well

[gentle nexus]

Gonna be something stupid.

[molten current]

I'm trying to get clawdbot configured with llama.cpp and don't even know where to start. Ollama isn't shown as a provider.

[hard shore]

gonna set up a new m.2 ssd on my pc tomorrow with ubuntu + clawdbot + ollama to have it air gapped to the rest of my pc

[gentle nexus]

It’s all manual, only gone by the docs. Can get it to open up an instance sometime but others it says it’s a local model but it’s lying. lol. Love to see one who has a working clawdbot.json file with ollama.

[hard shore]

search the threads in #1459642797895319552 maybe? I found some info there

but as I said I'm not there with my install yet. Just did VPS and OpenAI which was a token call desaster 😄

[gentle nexus]

Gone through them all, unless I have made a mistake. Been going at the for 18 hours now and are burnt out.

[hard shore]

damn then I'd pause and start grinding again tomorrow

[gentle nexus]

If you find anything tag us please, I’ll do the same. Gonna sleep. Great tool, just can’t afford 100 bucks a day.

[pure zodiac]

Hey, do you know how I can talk to the AI chat here on the server? I have a few questions before signing up, but I keep getting the message "Clawdbot: access not configured. Ask the bot owner to approve with:
clawdbot pairing approve discord <code>

[rocky orbit]

you can ask claude code or codex about it or go to #1459642797895319552

[ivory bridge]

Anyone tried add qwen coder free tier API into clawd?

[sand shadow]

I'm using Claude and got a LLM request rejected: imput length and max_toekns exceeded...context limit >200000, decrease input length or max_tokens and try again...I can't send anything to clawd though because I always get that message. Any thoughts on how to fix that?

[thorny lark]

Hi everyone!
I’m using clawd.bot and I have a native macOS application that requires a lot of repetitive manual work.
I would like to automate these tasks with clawd.bot.
Is it possible to integrate it with a macOS app, and how can this be implemented?

[pine elbow]

just curious are most ppl trying to install clawdbot devs or normies ?

[terse flicker]

ok i have 4 mac mini's im going to set up, whats the best starter guide?

[broken nebula]

Anyone having issues with slack? I have it setup via the stream connection as an app, but when i DM to it, its to the APP and not in the normal DM area, and the history keeps randomly resetting. has anyone got it to work the same as any other user?

[distant grove]

If you are on windows and Clawd kept saying no gog installed for gmail services , just direct clawd to use Tailscale ( i dont know how secure this is ) , it works

[pulsar grotto]

Hey everyone, anyone else dealing with the repetitive anthropic token expiration issue?

what's the best way to automate renewing the token?

[molten zealot]

Does perplexity work with Clawdbot?

[jade pike]

Having the same issue. So annoying

[molten current]

After digging and surmizing, llama.cpp is not yet supported by Clawdbot. I'd use Ollama, but I'm on Strix Halo, which Ollama's ROCm doesn't yet support; also their Vulkan support is compile-in only, and still stumbling quite a bit.

Guess I'll check back in with this project when my little edge case is resolved, one way or another :/

[wraith latch]

Hey guys when i tried to install i get this error, Windows Install can someone help

SecurityError: (:) [Invoke-Expression], PSSecurityException

[junior glade]

What is the best way to get email working with Clawdbot. I have go in circles forever trying to get himalaya to work with no luck.

[harsh spoke]

I'm trying to install gmail and google services for like 3 hours without luck on Windows clawdbot...I'm very desperate 🙁

[languid hamlet]

Hey folks, need some help with our Clawdbot setup on macOS.

Context

• Machine: macOS (CURACAO’s iMac), user clawdbot
• Node: v22.16.0
• Clawdbot: 2026.1.24-3 (885167d)
• Main agent now runs on Claude Opus 4.5 and works fine.
We’re stuck on two things:

1. ClawdHub / skills completely broken (Node 22 + “undici”)

Whenever we try to use ClawdHub, we get this error:
npx clawdhub@latest install sonoscli

...

Error [ERR_MODULE_NOT_FOUND]: Cannot find package 'undici' imported from /Users/clawdbot/.npm/_npx/5cb439d7d313769b/node_modules/clawdhub/dist/http.js
Node.js v22.16.0
Same story with the globally installed clawdhub CLI.

[faint dirge]

[clawdbot] 400 Your organization must be verified to generate reasoning summaries. Please go to: https://platform.openai.com/settings/organization/general and click on Verify Organization. If you just verified, it can take up to 15 minutes for access to propagate.

[junior glade]

Got many hours in as well wit no luck as well.

Tried windows and gave up on that install. Now on Ubuntu which bot works but can't get email setup.

[void apex]

i keep getting error 1006 random disconnects without explaination

[karmic gale]

so with the official mobile app taking a brief hiatus from development… has anyone figured out a way to access your Clawdbot via realtime voice on the go?

[languid hamlet]

anyone have a issue with the new node update not connecting the skills

[astral garnet]

voice messages via discord work great, create a voice skill for faster-whisper and piper, or use elevenlabs

[coarse jacinth]

Why would someone use Mac Mini when this thing can run on RaspberryPi?

[karmic gale]

looking for real-time back and forth not async… hoping to interact while walking

[astral garnet]

$$$$

[chrome pier]

im getting this a lot too and so are others it seems.. i had to deactivate telegram.

[astral garnet]

good question. i havent researched it that deep. it probably isnt easy with current api costs

[mellow heron]

Imessage integration is I think the answer.

Or one of them.

[coarse jacinth]

meaning? They've got too much dough?

[fair coral]

Any Linux distro recommended over another of the least hassle?

[chrome pier]

i know, it's effin ridiculous tbh. it became a trend to use that over priced hardware for this. Any old computer can run CB wonderfully.

[mellow heron]

I set it up on a macbook pro m1 , but I haven't been able to get a new apple id with the gmail I setup to isolate it. It was denying me the apple id. So I don't have imessage setup yet.

[mellow heron]

Also, you can use it to drive the machine with peekaboo. So it is like a person operating a mac. So depending on what you want to do that might be a bonus.

Haven't tried all the things yet myself.

[ashen patrol]

How can I set the audio message to use Groq Whisper isntead Openai?

[wanton tangle]

i’m running mine on a trashcan mac pro lol

[mellow heron]

I'm just not thinking it is ridiculous. But it just depends on what you try and do. If you want it to act like a person operating a machine, then you don't want a slow machine where you have to wait for it to do things. And also you might want to run local models.

[astral garnet]

uh just ask clawdbot to do it lol

[void apex]

what did deactivated telegram do ?

[terse hinge]

hello everyone, i'm having issues I want to send audio to clawd via telegram to be transcribed, the whisper-ai is enabled and configured. but the bot say it dont have permissions to use it 🙁

[coarse jacinth]

Thank you... so iMessage is probably the only USP and a reason I'd consider it... maybe Airdrop etc? Otherwise - a bit of an overkill?

[pseudo kindle]

Realtime voice

[mellow heron]

Could be, but again. The question is, what are you trying to automate? And how are you going to automate it. If you can set it up and all the resources you want are external, that's great. Ultimately I would like to have a bunch of power somewhere. I have a 5090 in my pc. So maybe for me I use the mac for the bot and just call out to local models on another machine.

I'm not suer the lowest end mac mini is what I would want, but I don't know. Luckly I had something to start with so I can see what I think and move later.

But getting off the online models. What would be really cool. But I don't think it is good enough yet. Mabye just like, automated stable diffisions or audio generation or something locally, but not the default bot model.

[jaunty stag]

Is Clawdhub ok for everyone? My Clawd seems to be having troubles searching it... gets a timeout.

[void apex]

i fucking hate this shit man, everyone is tweeting hyping this up and I cant even get mine to work

[jaunty stag]

How far have you gotten?

[void apex]

I was chatting with it - using it, just got it to start remembering key details about me then boom out of nowhere 1006 error after 1006 error. Disconnected gateway for no reason

[jaunty stag]

What are you running it on? Is it definitely still running?

[void apex]

it is currently down, i have wasted too much time on it

[astral garnet]

hey

you have to turn off auth on customUI or something in config

[sharp summit]

If I messed up setting my Claude max sub token on setup, what’s the best way to wipe it and reconfigure. I’ve tried a bunch of stuff and it defaults to the old (wrong one). Do I just have to do the whole wizard setup?

[astral garnet]

try this:

"gateway": {
"port": 18789,
"mode": "local",
"bind": "lan",
"controlUi": {
"allowInsecureAuth": true
},

[jaunty stag]

I found re-running the wizard wasn't too problematic. It remembers a lot of the key details and you get the option to either change or leave as-is, so that might work for you.

[astral garnet]

also, if you're on macos i found you had to run ssh tunnels to make it work

took me a long time to get remote and local gateway working, changed tokens, etc.

[void apex]

not on mac

[astral garnet]

are you local or remote gateway?

[void apex]

allow insecure auth?

[astral garnet]

test it and see if it fixes it

[hearty dove]

Anyone know if there is way to speak with it? I'm trying elevenlabs and I can't find a way to get it hooked it. It's even got me building a HTML app for it

[sharp summit]

Ok I’ll try that ty.

[astral garnet]

• All WS clients must include device identity during connect (operator + node).
  Control UI can omit it only when gateway.controlUi.allowInsecureAuth is enabled
  (or gateway.controlUi.dangerouslyDisableDeviceAuth for break-glass use).

[astral garnet]

yea, create a voice skill and tell it when you send a voice message it must follow the voice skill, write it to memory and soul/memory

[mint bloom]

Hi everyone, I am trying to run clawdbot with my own ubuntu server, I am able to have it running and ask questions, but looks like it is not able to connect to the skills that is showing in the skill page, how can I enable it for the agent to use?

[pulsar grotto]

My clawd was able to write a script to automate refreshing it

[astral garnet]

what does it say when you ask it where are its skills?

[jaunty stag]

I had a similar problem. It seemed like the format of the config file was different than what Clawd was expecting. I told Clawd all the skills I wanted to be enabled, and the rest disabled, and got Clawd to fix the config.

[wanton sandal]

GM

cant setup minimax with openrouter api

already burned my brain

⚠️ Agent failed before reply: Unknown model: openrouter/auto.

[astral garnet]

no luck in #old-models ?

[mint bloom]

@astral garnet

what skills do you have? can you check the current weather? or look up places in google?

You
23:31
A

I don't have direct access to real-time data sources like weather APIs or the ability to browse websites such as Google. However, I can help you with a wide range of tasks using my built-in skills and knowledge.

Here’s what I can do:

General Capabilities:
Answering Questions: Providing information on various topics.
Generating Text: Writing articles, stories, scripts, emails, reports, etc.
Providing Information: Offering insights and data on specific subjects.
Offering Advice: Giving suggestions or recommendations in areas like personal development, career advice, learning strategies, etc.
Translation: Helping with translations between different languages.
Creative Writing: Assisting in creative endeavors such as poetry writing, scriptwriting, brainstorming ideas, and more.
Specific Tasks:
Weather Information: I can provide general guidance on how to check the current weather using reliable sources like Weather.com or OpenWeatherMap.
Places Lookup: While I can't browse Google Maps directly, I can help you find information about places by suggesting reliable websites or APIs that you can use.
If you need assistance with any specific task or have a question, feel free to let me know! How can I help you today?

trying to ask clawd to fix the file .clawdbot/clawdbot.json and fix the skills

I am using ollama with qwen2.5 14b, does that matter?

@jaunty stag

Sure, I can help you with that. Let's take a look at .clawdbot/clawdbot.json and ensure the skills are correctly configured.

First, please share the current content of the clawdbot.json file so I can review it and suggest any necessary fixes.

If you have the file handy, paste its contents here. If not, you can also describe the specific issues or desired changes you want to make.

is it suppose to do this?

[lone agate]

hi, so I installed it, how do I launch it?

(copilot gave me a bunch of ps commands that returned nothing so far)

[hallow horizon]

Where are you at in the documentation that you're stuck?

How you launch it depends on how it was installed, so there is no one answer to that question

[lone agate]

I ran the ps command at the clawdbot web site : "iwr -useb https://clawd.bot/install.ps1 | iex" on my windows 11 pc

[hallow horizon]

I can't find where you're seeing that command, but you should have a clawdbot binary on your system path now - close and reopen powershell if it doesn't work

[lone agate]

haven't got to the documentation but it's second / last on my list 😄

I took it from here:

[sterile brook]

Having a really weird issue here. I installed Clawdbot in a fresh Ubuntu VM (on a Proxmox machine) and went through the set up, using my Anthropic key and setting up Discord. I was able to do the full TUI through the VM's shell, and conversing via Discord is working.

But I can't for the life of me access the control UI! I see that it's serving on the 127 local gateway, but since the VM is headless I can't directly check that on a browser. So I tried the LAN IP, and I just get a refused connection. Is there some trick or step that I missed?

[hallow horizon]

If the VM is headless you can use the curl command

To test http://127.0.0.1

From that machine

Connection refused means firewall likely

[stray glen]

hey guys, i wanna put in a new api key in openrouter in my existing clawdbot terminal.. what command do i need to do to setup a new openrouter api key?

anyone can help me out here?

[jaunty stag]

Mmm... that sounds like the LLM isn't getting whatever system prompt Clawd uses, so it's unaware of Clawd's config and directory structure, etc. Maybe tell it that it is an installation of clawdbot, and it should be able to access those files? Otherwise, you might try using a more tried and true model.

[lone agate]

is it normal that powershell closes automatically when installing clawdbot?

[jaunty stag]

You may be able to find it in the Control UI and update it there? Otherwise, I would ask Clawdbot

[stray glen]

looking for it in control ui but can't find it anywhere. where's the clawdbot

[sand shadow]

/compact fixed this for me.

[terse hinge]

I'm trying to enable whisper, the bot is trying to do it but no luck this breaks the json and I have to edit it back

[sterile brook]

Hmmm, so I was getting err_connection_refused when I tried from a different machine, but with curl it's just "could not connect to server" so not sure what's up

[jaunty stag]

Wherever you talk to Clawdbot

[stray glen]

thing is my credits are over lmao
thats why i need to add a new api key to keep chatting

[astral garnet]

That’s why.

[jaunty stag]

Does adding a new API key renew your credits? Are you using multiple accounts?

[stray glen]

yea multiple accounts

[mint bloom]

can someone please share the skill section of .clawdbot/clawdbot.json, just want to see how it looks

[astral garnet]

You need to stop using small local models.

[terse hinge]

"skills": { "install": {"nodeManager": "npm"}, "entries": { "nano-banana-pro": {"enabled": true, "apiKey": "-2wrDcIAPtZO-_WABaF8szs"}, "sag": {"enabled": true, "apiKey": "6e7d13345ced31"}, "openai-whisper-api": {"enabled": true, "env": {"OPENAI_API_KEY": "skVYvr0_z78g> } }

edited as it was with my API's they have no $$$ but still haha

[tired pollen]

Hey all, I have spent a few hours trying to setup clawdbot to use a local mistral model (using ollama). I have the model running on my mac mini, and in another terminal window I set up clawdbot but just cannot figure out how to get it to use my local model. I have tried much recommended to me by chatgpt but it's giving hallucinated advice that is not matching reality.

Would anyone be able to point me in the right direction?

[astral garnet]

You cannot use this with 30b models and be successful

[mint bloom]

😭 alright..

[astral garnet]

The context window is already going to break most of them then add in tool calls and it’s too much

[drowsy nova]

Hey guys, I might need your help D:

Setup: Fresh Hetzner VPS, followed the deployment tutorial, built from source with docker compose build, onboarding completed successfully. SSH tunnel is working.
Problem: Dashboard shows disconnected (1008): unauthorized: gateway token mismatch
Tried:

Using token from onboarding output in URL
Using token from .env (CLAWDBOT_GATEWAY_TOKEN)
Restarting container
Verified both tokens match

Tahnk you! ❤️

[mint bloom]

@terse hinge , please delete your api keys and disable them

[mental mica]

They weren’t real full keys

[terse hinge]

yup theyre not real

just examples

[mental mica]

Real keys are way longer

[jaunty stag]

I have mine set to explicitly enable / disable, so it looks roughly like this:

  "skills": {
    "entries": {
      "apple-reminders": {
        "enabled": false
      },
      "bear-notes": {
        "enabled": false
      },
      "goplaces": {
        "enabled": true,
        "apiKey": "<REDACTED>"
      },
      "local-places": {
        "enabled": true,
        "apiKey": "<REDACTED>"
      },
      "1password": {
        "enabled": false
      },
      "apple-notes": {
        "enabled": false
      },
      "bird": {
        "enabled": true
      },
    }
  },

(But much longer )

[paper dust]

how do i setup the application

not the terminal

[hallow horizon]

⁉️

[flint nova]

does anyone know how i can fine tune claw or give it instructions? i want to turn it into a moderator bot that can underrstand when a user asks a questiona dn reply to the user with a response, tho this is what i get:

its like suggestion what to say rather than responding to what i ask, where do i change this in the dashboard?

[jaunty stag]

Do you have both a gateway and cli container? If so, you might need to confirm both have the same value in CLAWDBOT_GATEWAY_TOKEN. I had to do this.

Edit: The gateway container runs continuously, and the cli container is just for running the cli, so it's not running continuously just when you run commands. But it's possible my setup is different from yours.

[steep nimbus]

I cannot for the life of me get clawd to remember to use the skill creator skill when creating or editing skills unless I directly insist he use it every single time.. has anybody figured out how to get him to do this?

[jaunty stag]

Have you tried telling Clawd to update the relevant file / instructions so that whenever it's creating or editing skills it will use the skill creator skill?

Not being funny, but like, so far the majority of times so far that Clawd has come back to me with some kind of error or issue, I have just asked it to fix it, and we fix it together, and then we're off and running again!

[hazy cloud]

Which models are you guys running locally fro clawdbot to use?

[stray glen]

hey guys, i wanna put in a new api key in openrouter in my existing clawdbot terminal.. what command do i need to do to setup a new openrouter api key? i wanna use a new openrouter account with loaded credits

[jaunty stag]

Is it an option for you to top up just enough credits on the current API key to ask Clawd this question? 😅

[dim sonnet]

I'm currently trying to setup clawdbot to work with LM Studio on an M4 Pro (24GB of memory). Unfortunately, I haven't been able to successfully get it running despite my best efforts. I am trying to test glm-4.7-flash and glm-4.6v-flash

[languid hamlet]

could someone please help me i cant add skills because of the following error,

Node 22 → undici error
Node 20 via nvm → same undici error
npx clawdhub@latest

[stray glen]

fck no, just loaded 200 bucks on a new account lmao

am i rekt or what?? no way to change the api??

[jaunty stag]

Nah, there's got to be a way, I just don't know it, sorry.

@stray glen Try looking in ~/.clawdbot/clawdbot.json under model.providers and see if the API key is set there.

[stray glen]

ill check, ty fam

[wraith seal]

hello chat, i am new

unfortunately clawdbot is no longer responding, and i want to keep my config

[jaunty stag]

hello new

[novel crystal]

try using agentmail.to

[wraith seal]

could anyone please assist me? thanks

[stray glen]

where do i open this? in terminal?

[jaunty stag]

Most likely. Depends a bit how you have it installed.

[jaunty stag]

How did you install Clawdbot? It may just need to be restarted or something, and that will also give more info on where your config might be. It could be you just make a copy of ~/.clawdbot/clawdbot.json

[north fulcrum]

Ask Claude code or codex, much better than asking here

[stray glen]

dont have it lmao

[wraith seal]

I installed via curl. I copied over my old config (I updated my values indeed) from my old instance, and now it doesn't work

[dim sonnet]

Lol. I've even tried setting up clawdbot with claude code. Claude code modified my ~/.claudebot/clawdbot.json and ~/.clawdbot/agents/main/agent/models.json files and ran clawdbot gateway restart.

[wraith seal]

Chat bubble pops up. Times out after 2 mins.

[plain sigil]

My ClawdBot is running on an M4 Mac Mini, seems to think it is just a chatbot. We started with me giving it telling it that it was my assistant and i wanted to give it info and context for projects. Hours later, I asked a question about one thing and it was like "what is that?", then I had to re-explain it because it was just agreeing with me in the telegram chat, but never actually saved anything

[north fulcrum]

Are you using telegram?

[junior glade]

I got it working. Man the bot was not following the github example for himalaya configright. I had to copy and past example and put info in and then it made 1 correction and it work ed.

[wraith seal]

WhatsApp.

[prisma vault]

Any way to stop things like "[[ reply_to_current" from appearing at the end of the output? It happens from time to time

[wraith seal]

With a separate phone number (eSIM on my phone)

[dim sonnet]

What model are you running?

[north fulcrum]

Can you ask Claude code or codex to solve it for you?

[jaunty stag]

Can you check logs or output to see what error might be happening?

[wraith seal]

It cannot figure it out even lmao

[wraith seal]

No errors. At all.

It just says it timed out, and that's it

[north fulcrum]

Ask it to check the offset if there’s one in Clawdbot for WhatsApp. Like the message offset

Not sure if WhatsApp has one. You need to check logs but Claude code can do that

[storm sentinel]

When I ask clawdbot to do something simple like "ls" in a discord server channel it works fine. If I ask the same thing in a DM it says it doesn't have exec permissions. There are no special permissions configured for the guild server and I'm not sure how to allow clawdbot to run commands on the server when I ask via DM.

[jaunty stag]

This is Clawdbot permissions, not Discord permissions, to protect from being able to do certain things from any old platform. But you can ask Clawdbot to modify that config to allow it.

[obtuse ember]

Anyone set up clawdbot with a minimax code subscription? I am facing an issue where clawdbot doesnt give me an output when using the minimax m2.1 model through even though it is in my models list

[storm sentinel]

Yes, I know that, but in my clawdbot config I cannot see how to change it, and clawdbot has tried several things most of which caused a broken config.

[storm sentinel]

Oh, when I said "no special permissions for the guild server", I meant in the claude config file. That was not clear.

[atomic sparrow]

⚠️ Agent failed before reply: No API key found for provider "anthropic". Auth store: /Users/xxx/.clawdbot/agents/main/agent/auth-profiles.json (agentDir: /Users/xxx/.clawdbot/agents/main/agent). Configure auth for this agent (clawdbot agents add <id>) or copy auth-profiles.json from the main agentDir.
Logs: clawdbot logs --follow

I'm going crazy with this after changing from Anthropic API key auth to Claude Pro Max oAuth. Anyone knows how to solve it?

[storm sentinel]

I have this setup and working for me. My models list line looks identical to yours.

[obtuse ember]

Did you use the clawdbot onboard command to set it up? Or did you do some sort of manual setup?

[storm sentinel]

I did it manually. I have both minimax and anthropic models available and have set fallbacks and such. But minimax is my default.

[obtuse ember]

Can you share the link to the guide you followed on setting it up manually? if you have it

[copper raven]

my clawd never responds it just sends an empty message however it always fails to install the skills

[barren blade]

So, running my head into a wall and hope someone can help. I routinely get the fun “no output” specifically in the webUI. Just a few minutes ago, it wouldn’t answer my request, I’d have Claude beat it for a bit, it’d answer the request mid-beating, and when Claude was done, it wouldn’t answer again…round and round we go.

Spent all my antigravity/opus tokens, all my opus/opus tokens, was attempting to set gpt as primary and the whole thing annoyed me

[atomic sparrow]

⚠️ Agent failed before reply: No API key found for provider "anthropic". Auth store: /Users/xxx/.clawdbot/agents/main/agent/auth-profiles.json (agentDir: /Users/xxx/.clawdbot/agents/main/agent). Configure auth for this agent (clawdbot agents add <id>) or copy auth-profiles.json from the main agentDir.
Logs: clawdbot logs --follow

I'm going crazy with this after changing from Anthropic API key auth to Claude Pro Max oAuth. Anyone knows how to solve it?

[barren blade]

Are you out of tokens?

[storm sentinel]

I'm not certain, it would be pages on docs.clawd.bot (or the docs directory in github as I use those too). Here is what I have though - MINIMAX_API_KEY in .env, an auth.profiles.minimax:default key with provider and mode "api_key", and an agents.defaults.model.primary set to the model. There are some defaults built in that are enabled when you have the API env variable set apparently.

[fair coral]

Does running Clawdbot on a system where it can open a browser help it browse the web?

[sour slate]

The fastest way to get your problem solved is to follow the instructions here: https://docs.clawd.bot/help/faq#im-stuck-whats-the-fastest-way-to-get-unstuck

[copper raven]

no my openai key is good

[verbal gorge]

anyone had their Claude account suspended for using their Max account instead of API?

[kind star]

never and using it locally and on 4 different remote services

[atomic sparrow]

⚠️ Agent failed before reply: No API key found for provider "anthropic". Auth store: /Users/xxx/.clawdbot/agents/main/agent/auth-profiles.json (agentDir: /Users/xxx/.clawdbot/agents/main/agent). Configure auth for this agent (clawdbot agents add <id>) or copy auth-profiles.json from the main agentDir.
Logs: clawdbot logs --follow

I'm going crazy with this after changing from Anthropic API key auth to Claude Pro Max oAuth. Anyone knows how to solve it?

this is driving me nuts

[wraith seal]

I am going insane. The bot is just not responding. My quota is fine...

[hazy cloud]

I was able to get it setup using qwen2.5-14b-instruct-1m as the model. However its very slow generating responses

[atomic sparrow]

I fixed it. If anyone ever faces the same issue, locate which agent is running throgh your gateway and then run clawdbot agents add [name-of-your-agent-here] and just fix the auth through there

Maybe it's nothing crazy - but somehow I fucked up something that made me go in loops, so if anyone ever faces it, there you go.

[sharp pollen]

Try messing around with the TUI. It’s a faster way to debug this

[wraith seal]

Nothing shows up in the TUI. At all.

Fresh, base config. No changes.

[sharp pollen]

Connecting to the right session?

[hallow hinge]

so with the claude code + opencode debacle, is it safe to use a claude max plan with clawdbot?

[cobalt sonnet]

curious which debacle you mean, as a max user 👀

[hallow hinge]

well, people getting banned using claude code max sub with opencode

[meager mortar]

after running a few weeks, i built up a lot of skills i am not using, adding weight to the context. periodically its a good idea to have your bot list out all its SKILLS

[wraith seal]

Indeed.

[sharp pollen]

Are gateway logs of any help?

[ornate forge]

how do you build and install the mac app?

[azure totem]

"The setup costs $200/month for Claude Max plus whatever hardware you have lying around."

What does this mean? People are saying this online but from my understanding, Clawd cant use Max as you need credits to run Clawd?

[ashen patrol]

I asked, we can with a solution, but Is not the best.

[sacred parrot]

yo guys I'm having this problem, and I can't solve it no matter what! I've already tried a different model, I've already reinstalled it from scratch...
Has anyone had something similar and managed to solve it?

[steel silo]

Hello people, I'm running a local setup of Ollama + Clawd + Discord on a Windows machine without WSL and i feel like the bot has no permissions to read/write anything, even its own .MD files

[azure totem]

ive been doing this. use claude code and it will help you out

or i was doing this and having issues

[sharp pollen]

./scripts/sign-mac-app (something like that)

[hushed dawn]

Any luck on this. I'm having issues as well

[wraith seal]

Nope.

Literally no errors, strangely.

[sharp pollen]

What does /status say?

[golden sphinx]

anyone know why the bot is always sending this back to me

[pastel anchor]

anyone knows why i'm keep getting this error when start the gateway?

02:31:11 [telegram] telegram setMyCommands failed: HttpError: Network request for 'setMyCommands' failed!
02:31:11 [clawdbot] Unhandled promise rejection: TypeError: fetch failed
at node:internal/deps/undici/undici:14902:13
at processTicksAndRejections (node:internal/process/task_queues:105:5)

[cyan walrus]

Anyone have tips for creating a new iCloud account? I’ve tried a few times over last week and keep getting “You cannot create an Apple Account at this time. Try again later.”

Not sure if Apple is restricting it based upon my cell being associated with my main iCloud account or what heuristic.

[wraith jolt]

Hey ya’ll! I don’t know if anyone can help me, I keep running into token limit issues and now context overflow issue with clawd bot, how do i fix? I already increased my subscription to max

[sonic snow]

I used:
sudo clawdbot models auth paste-token --provider xai

[opal plover]

Are you running on Linux server with faster whisper?

[wraith jolt]

Can anyone help me please?

[jade pike]

How did that fix it?

[paper pilot]

⚠️ Agent failed before reply: All models failed (2): anthropic/claude-opus-4-5: session file locked (timeout 10000ms): pid=1 /home/node/.clawdbot/agents/main/sessions/d3086ba6-3e37-4431-a58f-3128493bbb71.jsonl.lock (timeout) | openai-codex/gpt-5.2: session file locked (timeout 10000ms): pid=1 /home/node/.clawdbot/agents/main/sessions/d3086ba6-3e37-4431-a58f-3128493bbb71.jsonl.lock (timeout).
Logs: clawdbot logs --follow

When are we fixing this guys 😔

[slow geyser]

Same issues... i'm using LM Studio and context at 16384,but Clawdbot sends massive inputs... my prompt "Hey what's up" input size 12K tokens lol WTF?

[north fulcrum]

System prompt has tools, skills, soul, agents.md etc etc

[paper pilot]

Why are we sharing the QR code

At least blur it

[north fulcrum]

@echo vector don’t share the qr code are you insane

[echo vector]

fine didnt think it did anything cuz it didnt work when i tried it

[paper pilot]

💀

[echo vector]

im fixing it hold on lol

[sour slate]

@echo vector, the fastest way to get your problem solved is to follow the instructions here: https://docs.clawd.bot/help/faq#im-stuck-whats-the-fastest-way-to-get-unstuck

[slow geyser]

Any way to reduce the input bloat? or do we have to assume almost 12K token inputs always?

[echo vector]

soo how do i get past this i installed it and everything im trying to get to the dashboard

clawdbot daemon restart zsh: command not found: clawdbot
@paper pilot @north fulcrum

[wanton tangle]

I gave my bot an email. Im wondering what would be the best way to keep him checking for any new emails or to have him respond when the email gets a new email

[astral garnet]

Probably a cron

[wraith seal]

Source .bashrc?

[atomic sparrow]

Anyone else slow as hell?

[wraith seal]

Elaborate.

[odd swift]

bro cool to see another ecom guy here lol

[unkempt patio]

Anyone know how to install on an older mac? i’m on monterey 12.7.6 (can’t update) it always fails at npm install

[vale elbow]

Anyone else having this issue?

[fresh moon]

hey guys

been having trouble trying to install clawdhub by my ai says theres something wrong with it

[valid veldt]

Anyone else having issues with Discord bot being incredibly unreliable, whenever sent to do some tasks like self improvement, it continuously stops responding prompting some kind of action through TUI, some simple some complex

[white crown]

Should web_fetch work on the standard sandbox docker image?

[sand loom]

Hey did you find a fix?

[vale elbow]

just uninstalled all my work, going to try a fresh install

[slow geyser]

I did using local LLM via LM Studio - went to smaller model to fix it also had to pu the context size to 32768

[ornate pelican]

No butter jokes! haha love it! Hey looking for a way to limit request rate limits! Any ideas:

[primal tree]

Does anyone know how to resolve this issue? I tried to change the model from opus to sonnet just via asking it in WhatsApp and now it’s broken

[sharp pollen]

/ models or / new

[formal fiber]

I need help. I've just copied my config over to my new ec2 instance. It detects the existing config, however when I go to configure channels, the Discord plugin is disabled when I go through the wizard. It says configured, plugin disabled.

[iron dawn]

how do you guys install the companion app on mac again I uninstalled it because it kept crashing but no idea how to reinstall it lol

[static oak]

Has anyone seen this tool calls error before? I'm using GitHub copilot and changed models to gpt 4.1 and started getting this error. Gpt 4.1 worked earlier for me, I was trying out different copliot models. I'm not sure how to fix this, I've rebooted and tried reinstalling the gateway with no luck.

[astral garnet]

Check out open patcher

[astral garnet]

Releases on GitHub

[tired locust]

i have a local ollama server setup, and clawd has used it before for subagents, but im trying to get the default session to use it, how do i set that up and then restart the session?

[twilit axle]

can someone recommend a skill to do broswer use well? particulary, i am looking to fill out signup pops on shopping websites

[unkempt patio]

can you elaborate please and talk to me like i’m a child or ill get confused

[astral garnet]

Search openpatcher for macos

[wraith seal]

He's trying to refer to OpenCore Legacy Patcher.

This is a very complicated series of setup you can go through in order to get a new version of MacOS.

If you think you're capable, use Tahoe. If you're not as confident, use Sequoia. If you're less confident, use Sonoma.

Ventura presumably doesn't work.

[mental mica]

Ventura does work, that’s what I run

With opencore

[acoustic cedar]

Hi everyone, can anyone help me with this, i use Macbook M1 and try instaling with 1 one-liner code provided in the website, but it keep says npm failed to install.

How can i fix this? thank you so much

[golden sphinx]

haha thats fire

[dusk ridge]

Does anyone know how to use the models I have already installed locally?

[static oak]

1 hour

[severe rampart]

I am getting a lot of tool call errors as well, with minimax-m2.1 and glm 4.7 flash? Minimax m2.1 worked great earlier! Not sure what is broken.

[digital cairn]

https://docs.clawd.bot/concepts/model-providers#local-proxies-lm-studio%2C-vllm%2C-litellm%2C-etc

[lethal olive]

how can i let clawdbot control my local device? it says to install node but what then

[digital cairn]

or if you're using ollama - https://docs.clawd.bot/concepts/model-providers#ollama

[digital cairn]

you'll need to be a bit more specific with your set up - where is it running? locally? vps? mac mini? something else?

[static oak]

Are you able to get anything working right now? All I get are tool call errors

HTTP 400: Invalid 'messages[135].tool_calls[0].id': string too long. Expected a string with maximum length 40, but got a string with length 442 instead.

[severe rampart]

Nope, everything is failing

[lethal olive]

im using it on a small laptop i own, just set up clawdbot through whatsapp with gemini ai api key, and aws hosting

i would love for it to search up some sites on my pc, just a simple task like that is what im trying to setup 🙂

[severe rampart]

Are you using OpenRouter? I wonder this is a case with OpenRouter models.

[static oak]

My issues are with GitHub copilot

[severe rampart]

Oh, I am not sure what broke then. I didn't manually update clawdbot in this time.

[lethal olive]

here this error is what im getting

[digital cairn]

maybe try this? https://docs.clawd.bot/cli/browser#remote-browser-control-node-host-proxy

[iron dawn]

I uninstalled all and ran a fresh new install with the one liner terminal and finished all the onboarding but the app is still not installed no idea why

[acoustic cedar]

Can someone kindly help me with this 🙏🏽🙏🏽

[digital cairn]

seems like you're running your whatsapp client on your PC, gateway + agent in AWS, if you want it to do something on your PC, you'll need to give it a way to invoke browser actions (or other actions ie cameras etc.) via nodes - https://docs.clawd.bot/cli/nodes

[lethal olive]

when i try to install clawd node i get this

[digital cairn]

404 indicates that's not a package that exists and that you can't download/run it- try running 'clawdbot nodes'

[digital cairn]

try running the script again with the --verbose flag for better debugging

[digital cairn]

ie 'curl -fsSL https://clawd.bot/install.sh | bash -s -- --verbose'

[acoustic cedar]

okay will try

thank you

[wooden mirage]

Ive seen a lot of discussion between folks saying to deploy this remotely vs on prem. Anyone done it on Coolfy yet? What are the tradeoffs remote vs local (Mac Mini etc)

[acoustic cedar]

hey mate, this is what i get:

[dusk ridge]

Thank you!

[digital cairn]

thanks! the error shows that you don't have permission to create anything under the /usr/local/lib directory - probably because you're not an admin on the macbook air you're running it on. Try running the alternate (non-root-friendly) CLI instead? 'curl -fsSL https://clawd.bot/install-cli.sh | bash'

[acoustic cedar]

Thank you, will try

[digital cairn]

this thing? https://coolify.io/

[severe rampart]

Running clawdbot update i.e. updating clawdbot fixed the tool call issue for me.

[wooden mirage]

mhm, just wasnt sure if I needed to run it on a machine with a graphics component is all

b/c the server I have has no GPU so I guess thats a thng to consider

[stable bolt]

Direct Access to the Application Form: The ability to navigate and fill out forms on the web.

Which skill do I need to enable for this activity ?

[silver imp]

HI people - does anyone have any skills for 'port management' when you're developing multiple apps on a VPS and want different default ports?

[unkempt patio]

ok so after 8 hours of figuring this out, finally got the dashboard and when i type, claws doesn’t say anything it just says “assistant”

[mint bloom]

Hi everyone, what is the cheapest model and provider is good for testing ?

[digital cairn]

just reading over it, this reads like an oss platform alternative to heroku/vercel/netlify/etc - they're all platforms; you'd deploy coolify on top of an existing server - those could be hosted anywhere. Unless you're hosting a model on a server you probably don't need a GPU component; there's a lot of model providers bundled already - https://docs.clawd.bot/providers/models

[eager ferry]

I just finished installing the clawdbot and initially i was using the free gemini account but it was giving me errors, how do I change it to another model using the web dashboard? I have a $20/month chatgpt plus plan but not sure if I can use clawdbot with that?

[drowsy steeple]

Hey seen u can use github copilot with this now, is that gonna be better than a $20 minimax plan???

[nocturne moth]

Anyone had much success using sonnet for clawdbot? I watched a video where they recommended opus but man that sounds expensive

[mellow heron]

well. the google client skill has eaten an hour of my time trying to setup oauth to access the email I setup with no love. Clawdbot is foisting me off on himalaya. So frustrating. I have not idea why it won't let me do the auth.

[rocky orbit]

Gogcli is not working?

[mellow heron]

no it is not. I can run the command line, but ...

[rocky orbit]

Sonnet is fine

[rocky orbit]

Ask your bot to help you

[mellow heron]

It has me set up and app and download the json oauth config and it goes through the permissions and then just hangs.

Yeah, it has been "helping" .. that and chat gpt , no love.

[rocky orbit]

Have you tried with Krill on #1459642797895319552 ?

[mellow heron]

no. maybe I will do that. I was trying to do the app password thing with himalya or whatever and my new free gmail I think doesn't have app passwords!

[rocky orbit]

You don’t need Himalaya for google

[mellow heron]

I need something that works.... is it because it is a free account?

[burnt quest]

this is what im currently seeing need some help and im not very pc savy only a gamer really

[mellow heron]

Ah.. service accounts... maybe that is the thing... in the help section. thanks.

[rocky orbit]

Restart gateway

[burnt quest]

how to do that

[mellow heron]

Ah, damn it, that is only for org accounts. This is so tedious.

[sour slate]

The fastest way to get your problem solved is to follow the instructions here: https://docs.clawd.bot/help/faq#im-stuck-whats-the-fastest-way-to-get-unstuck

[burnt quest]

this is what it tells me

seems to not still be working

[left spire]

I am trying to add my gemini api key and then use gemini with the tool. But I keep getting errors in every form way and shape... Which gemini model should I use for it to work?

[median ivy]

tried both pro and flash and they work but I've used the subscription

don't you have to just set the GEMINI_API_KEY env var ad it should work? which error are you getting? empty response?

[rocky orbit]

Go on #1459642797895319552 it’ll be easier

[left spire]

Yeah, I've set a gemini api key variable. I just get empty responses. And when I try it through the terminal, I always get unknown model errors. I've tried everything from gemini 1.5 all to 3.0 pro

Should I try it gemini CLI rather than gemini api?

[median ivy]

unknown model? never happened to me

[left spire]

I'll terminate the VM and start from scratch again. So I can share the correct error. I've changed so much by now...

[median ivy]

you can try, it works for me but again using the subscription

[left spire]

yeah, I have a sub. Made an app for myself (not clawdbot related) which works flawlessly

[median ivy]

are you using clawdbot onboard to configure the key? I'd also check the content of .clawdbot/clawdbot.json to see what is currently set in the configuration

[left spire]

Yeah, I did it through the clawdbot installer first. And then I also tried it as a global var.

Will try again, gimme 10m

[wispy spruce]

Have an issue where cronjobs are shown inside jobs.json however when I ask CB it says there is nothing in cronjob and also cronjobs dont run?

[upbeat oak]

Getting this while installing through "curl -fsSL https://clawd.bot/install.sh | bash"
So I tried "curl -fsSL https://clawd.bot/install-cli.sh | bash" it said, clawdbot installed but the onboarding command says command not found: clawdbot

[fossil lichen]

use node 22
24 and clawd is buggy

[upbeat oak]

The "curl -fsSL https://clawd.bot/install-cli.sh | bash" installed node 22, it said clawdbot installed but the onboard command says not found sadly

[fossil lichen]

do nvm ls

[digital cairn]

can you run it again with --verbose?

[fossil lichen]

youre on n24
clawdbot is on 22

[digital cairn]

ie curl -fsSL https://clawd.bot/install.sh | bash -s -- --verbose

[upbeat oak]

Yes I did run it, but still onboard command not found

[upbeat oak]

Any easier way to get to n22? Tried installing nvm but not successful on it

[upbeat oak]

It's saying clawdbot installed but not working onboard command

[digital cairn]

no, it's saying the command clawdbot is not found, meaning it's not in your active path

[digital cairn]

if you ran install-cli.sh with no flags, can you confirm there's anything in the ~/.clawdbot directory?

[random thunder]

Anyone know common solution to all models failed no available auth profile ? I have opus 4.5 as main and gpt 5.2 as fallback with max plans for both companies

Impossible that both my Claude max and OpenAI max plan ran out at same times i assume its a straight forward bug to reset or something?

[upbeat oak]

Okay so I just installed node 22 on my system using nvm and tried installing again; everything is working now perfectly
Thank you so much @fossil lichen @digital cairn ❤️

[twilit moat]

Hi does anyone know if ClawdBot would be capable of creating a browser based linkedin automation? Find companies in the niche I would specify as keywords and then find people working for those companies to send them a personal message? All based on browser automation?

[random thunder]

Yes easy man, just connect it to clawd browser and it does whatever a human can do inside a browser. This is part and parcel for it

[median ivy]

with a bit of work I'd say yes

and I'm thinking about headless browser, like lightpanda (did some experiment yesterday)

[twilit moat]

Ahh sounds great! I know this would be easy locally but I'm trying to run the bot inside a VPS so installing a headless browser and then writing commands seems like a lot of work 😄

[random thunder]

It’s not a lot of work. Literally log in with your credentials on the browser clawd opens which is chrome with clawd profile

Then copy and paste what you wrote here and ask it to run every hour

It’s 30 seconds of work

Oh yes Vps harder

[median ivy]

the only roadblock you could hit is the human verification thingy.....

[random thunder]

You won’t get it because it’s browser use and playwright so it acts on top of browser like a human with a mouse basically

[twilit moat]

Perfect you think any security concerns at all running the browser locally? Like I'm not sure how much access would the clawdbot would have over my chrome's data.

[median ivy]

oh yes, if you use the gui definitely

[random thunder]

If I’m honest I went all in, full access to my MacBook tofl

Degenerate but it’s so incredible I couldn’t resist

Don’t do that. But if you do, life is easy

[digital cairn]

buddy, running it over your local chrome is giving it access to your entire life

[median ivy]

and it's not necessary for most tasks

[fossil lichen]

my use case is
going over graduation videos on youtube and making a list of all the uni, graduates, majors.

[random thunder]

Be atleast 10% responsible and use clawd profile on chrome on your otherwise fully given access MacBook

[twilit moat]

Okay guys I'm going to try and create the headless browser remote VPS and report back how it goes 🧀

[median ivy]

can't yet take screenshot but the rest is great, the agent can navigate the rendered DOM perfectly -> https://lightpanda.io/

(10x less ram, 10x faster yo)

[upbeat oak]

For some reason I'm not getting any reply here:

[fossil lichen]

check logs

also what model

[upbeat oak]

Using 5.2 pro I think

[fossil lichen]

scroll
not the heartbeat ones

[random thunder]

Guys sorry to repeat but does anyone have ideas about this

Anyone know common solution to all models failed no available auth profile ? I have opus 4.5 as main and gpt 5.2 as fallback with max plans for both companies

Perhaps @digital cairn
Impossible that both my Claude max and OpenAI max plan ran out at same times i assume its a straight forward bug to reset or something?

[median ivy]

have you tried onboarding again? I suppose you did

[wind flicker]

Anyone have ideas on how to get clawdbot to solve captcha?

[random thunder]

Ive been using it all day with opus which makes sense it hits limits but gpt $200 plan is even more usage then Claude max and unlikely to have been hit

[warped zephyr]

hi, super new to this. I am having issues connecting my claude max account, just getting blank messages when using auth token. did anthropic patch 3rd parties connecting?

[fossil lichen]

agent start and end too quick
something is wrong on your model end @upbeat oak

[oak galleon]

what is with the no output bug? literally nothing is working, tried every fix

[digital cairn]

judging by this, I think maybe you've hit limits on anthropic and the fallback to GPT isn't configrued correctly?

[hard valley]

hi guys all i've heard is how clawdbot controls everything but my doesn't it keeps telling me how its not allowed to do anything I'm trying to figure out why my Experience is so different than everyone else's

[random thunder]

I’ve used gpt already it’s configured via copilot bridge or whatever it’s called with OpenAI oauth

[sacred swallow]

Can you try /reset ?

It’s annoying for me as well- yesterday it send me the text file as attachment on whatsapp and now today it’s saying it can’t do that action

I used reset to break its hallucination but still end up with similar experience like yours

[wind flicker]

Same question actually

[random thunder]

Reset no good

Maybe from terminal restart?

[sacred swallow]

Even restart doesn’t work for me

It keeps saying can’t access my files and cant access browser so something might be broken

[fossil lichen]

are you on sandbox by any chance?

[median ivy]

is this related to /elevated maybe? https://docs.clawd.bot/tools/elevated

[sacred swallow]

How do I check that? I reinstalled and did the normal onboard process

[hard valley]

i feel like we need a file that tells it what can and can't do and it should check it each time to remind itself it a demi god

[upbeat oak]

Tried using different model but still the same issue sadly :/

[hard valley]

i just saw a youtube use clawd to control googles antigravity and when i tried mine said this

[bitter crane]

Anyone using Gemini cli auth? My bot doesn’t answer my questions nor do his cronjob we setup yesterday. Yesterday he was working well. I’m running it on my old MacBook Air M1 xd

[analog barn]

hey guys whats the best node manager npm / pnpm or bun ?

or it doest matter

[oak galleon]

anyone else just getting no output regardless of what model they use?

[digital cairn]

npm is good enough for 99% of your use cases, you likely won't care or notice much of a difference between npm/pnpm/bun

[icy timber]

If i want to setup clawd in an old mac mini, is it ok to user my apple account or should I create a new one just for that?

[hexed fractal]

Bot getting stuck in same loop using Kimi.

[floral sand]

Does anybody know when all the code is going to change from clawdbot -> moltbot and how that's going to work and be coordinated?

[lavish swan]

heeyy everyone, I got a question with clawdbot: is it possible to use a local llm instead the Opus 4.5 API ?

[gusty axle]

how can i restrict Clawd executing the /rm /mv /cp/ and etc, only by user approval, so in future i wont loose some important documents?

[bright marlin]

Does anyone know if there is a list of valid values for "openai-codex/X" models somewhere? I saw discussions were people have been talking about "openai-codex/gpt-5-mini" but I have been trying to get a mini version of the model working for a while and no value I can think of or that clawdbot itself can suggest seems to be correct.

[floral sand]

It is.

[dull delta]

https://docs.molt.bot/gateway/local-models

[dull delta]

seems like you want a allowlist https://docs.molt.bot/tools/exec#allowlist-safe-bins

[grand epoch]

whats the cheapest and best way to use anthropic models?

[dull delta]

You should be able to /model list and get the list there (https://docs.molt.bot/concepts/models)

[raw yarrow]

is their a difference between mr lobster and moltbot?

[obsidian bone]

Failed to resolve Discord application id..

but my bot token is correct and "Server Members Intent","Message Content Intent" is turn on

[bright marlin]

It does list **openai-codex/gpt-5-mini **but then if I set that value

"Error: Unknown model: openai-codex/gpt-5-mini"

[bright marlin]

Ok, it's

openai-codex/gpt-5.1-codex-mini

[trim sparrow]

was anyone able to authenticate w/ claude pro account and have the ai response in the 'default' gateway? I'm only able to have it response when I put an api key

I assume that connecting a seperate UI like signal or something won't solve it right? my friend also has this exact problem

[iron dawn]

only get these messages all of a sudden LLM request rejected: messages.181.content.1.tool_use.input: Field required

[bright marlin]

I was wondering,
I have seen people stating things like "Clawdbot worked on it all night and I woke up with the full project done".

Those are probably exagerated, but in general, is there anything I need to learn from the docs or from articles on how to fire long running tasks (OpenAI DeepSearch style)? To me Clawdbot seems to just answer quickly with whatever it comes up first. More or less the quality I get is the same as the non thinking models and certainly far from agents

[oak portal]

can i use my clawd bot with my chat gpt pro subscription for unlimited tasks ? how does it work

[wind flicker]

Hey guys which are the MUST HAVE skills to get on clawdbot?

[chrome pier]

yes. Haven't tried it myself but i saw the option during setup, you choose to use ChatGPT subscription (OauTH style) or API. I'm currently using my API because I want to track how expensive CB can get. So i setup CB's own API key and wow, i was just using the 4 mini models and i racked up 15usd chatting w/ it for a few hours (just doing some troubleshooting).. not even heavy lifting stuff. so, be careful, it may hit your limits in your pro sub.

[oak portal]

do i want new ai accounts for it if im using it to do 1 of my companies ?

[chrome pier]

yah well i don't know about this.. it seems to pretend it works in the BG... but im not seeing any updates or signs of work in the bG... check out my convo with CB..

[bright marlin]

yeah, sadly pretending is what LLMs are good at.
The issue is that if it's unreliable in invoking skills or tackling tasks, at that point using something like n8n is probably more effective.
I'm surprised because gemini-cli, codex, etc.. are much more reliable on this, so I expected more or less the same level given it relies on the same models

[stiff tusk]

has anyone been able to configure CB MacOS Companion app to work with remote gateway via ssh?

[chrome pier]

...i dunno what everyone was raving about the "always working in the background" - i've yet to see that fa-real. atm, it seems to be task triggered only UNLESS you actually create a cron and by that, of course it's gonna work in the background but that's a different story.

A true "Treat it like a real human assistant" is not what it is, as far as I'm experiencing. I hope I am wrong coz it would've been very cool. Coz at this rate, AgentZero would be a far better dedicated agent.

[chrome pier]

that really depends on you buddy and how much you use it.. coz even ChatGPT Pro has limits .. while it has unlimited message caps, I would assume, clawdbot is like Codex instance, accessing the LLM via CLI etc, so it's like an API thing and that has limits.

I code a lot on my GPT account so , with the rate of CB burning through my API, i'd assume it'd push my main account to its limits.

But if you're not coding or eating up much of the API kind of access, i guess you're fine.

[random thunder]

So auth problems, how do we reset it? is there a way to re-enter claude and gpt auth tokens after its set up and is broken?

[oak portal]

how much usage u think u get for the coding then ? i dont code but would likie a idea of what should setup with...

[chrome pier]

difficult to tell. coding does burn a lot due to the large context and I always hit my 5hr and weekly limits on my GPT PLUS subscription .. Pro sub is way more generous though. Since you dont code, then your codex/cli/api credits and benefit are just going to waste haha. So go ahead and use your PRO on it, and monitor your usage limits at https://chatgpt.com/codex/settings/usage

[unkempt vector]

hey which group can I post some logos for molty?

[analog barn]

is running local ollama any good ?

[austere bluff]

Hi i am running clawdbot on a hetzner 2cpu 4gb instance. I see that the clawdbot freeze time to time then after ~30 min i get the reply back. Is there a way to debug and see what's going on while its freezing?

[cyan marsh]

curl -fsSL https://molt.bot/install.sh | bash

[sturdy hamlet]

If I get an api at aistudio.google.com can I use that to see if it works?

free plan*

[dreamy parrot]

Hello! Does anyone knows how to update to lastest main branch commit? I just keep getting version 2026.1.24-3

[austere bluff]

yeah you can use aistudio api key

[warm thorn]

Can anyone spot why only Opus works, and if I switch model to Kimi or Minimax, there's no response ❓

  "env": {
    "MINIMAX_API_KEY": "x",
    "MOONSHOT_API_KEY": "x"
  },
  "auth": {
    "profiles": {
      "anthropic:default": {
        "provider": "anthropic",
        "mode": "token"
      }
    }
  },
  "models": {
    "mode": "merge",
    "providers": {
      "minimax": {
        "baseUrl": "https://api.minimax.io/anthropic",
        "apiKey": "x",
        "api": "anthropic-messages",
        "models": [
          {
            "id": "MiniMax-M2.1",
            "name": "MiniMax M2.1",
            "reasoning": false,
            "input": [
              "text"
            ],
            "cost": {
              "input": 20,
              "output": 60,
              "cacheRead": 1,
              "cacheWrite": 4
            },
            "contextWindow": 200000,
            "maxTokens": 8192
          }
        ]
      },
      "moonshot": {
        "baseUrl": "https://api.moonshot.ai/v1",
        "apiKey": "x",
        "api": "openai-completions",
        "models": [
          {
            "id": "kimi-k2-thinking",
            "name": "Kimi K2 Thinking",
            "reasoning": true,
            "input": [
              "text"
            ],
            "cost": {
              "input": 0.6,
              "output": 3,
              "cacheRead": 0.1,
              "cacheWrite": 0.6
            },
            "contextWindow": 262144,
            "maxTokens": 8192
          },
          {
            "id": "kimi-k2.5",
            "name": "Kimi K2.5",
            "reasoning": true,
            "input": [
              "text"
            ],
            "cost": {
              "input": 0.6,
              "output": 3,
              "cacheRead": 0.1,
              "cacheWrite": 0.6
            },
            "contextWindow": 262144,
            "maxTokens": 8192
          }
        ]
      }
    }
  },

    "defaults": {
      "model": {
        "primary": "moonshot/kimi-k2-thinking"
      },
      "models": {
        "minimax/MiniMax-M2.1": {
          "alias": "minimax"
        },
        "anthropic/claude-opus-4-5": {
          "alias": "opus"
        },
        "moonshot/kimi-k2-thinking": {
          "alias": "kimi-thinking"
        },
        "moonshot/kimi-k2.5": {
          "alias": "kimi"
        }
      },```

[sturdy hamlet]

I'm trying to test it for a discord bot and it just wont work and it would say no output

[sweet moth]

ditto

[scarlet shell]

just get ollama local working with glm-4.7-flash, its just too slow running locally

[analog barn]

yes I am running in vps and it was consuming al the ram

[tribal pebble]

Anyone succesful with connecting their macbook as a node to gateway on a VPS?

[tribal flame]

hello guys, anyone having problem when autheticate the Anthropic setup-token

OAuth/token status
- anthropic
  - anthropic:claude-cli expired expires in 0m (claude-cli)
  - anthropic:default static

I already tried to login via claude setup-token I already been redirect, but the auth token still expired

[fringe hollow]

Day 1 for me, I am a bit at a loss (maybe because I don't understand the architecture fully) - I have everything set up and running fine (as long as I don't hit the api block, which happens every hour or so...) on an ubuntu linux machine. I would like clawdbot to occasionally control stuff on my windows desktop, though. There seems to be support for something called a "headless" node, which I think I have set up correctly (it does show up as a connected instance on the gateway dashboard). I have also installed the chrome browser plugin on my windows client - however, this seems to depend on a specific service running on 18792 which I do not have. Is it possible to active this with the headless node at all?

[hot belfry]

hitting an install problem due to the package rename... i guess this will get fixed but can I do anything to help?

In ../extensions/googlechat: "clawdbot@workspace:*" is in the dependencies but no package named "clawdbot" is present in the workspace

[oak portal]

i accidentally just canceled setup after syncing my telegram and downloading imsg... how do i fix ?

[mossy ginkgo]

Hello, I have two Clawdbot Telegram bots in one group, but how can I make them interact with each other? I have already set the group policy to open with a group allow list, but somehow the bots still cannot see each other’s messages.

[unique void]

have you tried asking clawdbot? 🙂

[fresh vine]

hittin the same issue

[boreal sedge]

Hey guys, how to get my claude setup-token? Thanks

[green ravine]

anyone know how to solve this? telegram channel. also getting disconnected from gateway pretty frequently

[wraith wedge]

Hey guys! Are there any good guides on how to set up MoltBot onto a Hetzner VPS?

[pulsar brook]

Is Gemini 3 Pro much slower than the other models with Clawdbot? Mine is pretty darn slow compare to others I've seen on Youtube etc.

[gray light]

same issue