users-helping-users

[keen hollow]

run

nano ~/.clawdbot/agents/main/agent/auth-profiles.json

What does it say (dont show the access token)

[unreal bear]

"version": 1,
"profiles": {
"anthropic:default": {
"type": "token",
"provider": "anthropic",
"token": "sxxxxxxx$
"lastGood": {
"anthropic": "anthropic: default"
"usageStats": {
"anthropic: default": {
"errorCount": 0,
"lastFailureAt": 1769232599444,
"lastUsed": 1769232599448

[keen hollow]

run clawdbot models

does it lok like this

[unreal bear]

It says usage: error invalid bearer token

[keen hollow]

Can you copy the whole thing below Auth Overview

[unreal bear]

Auth overview
Auth store: :~/.clawdbot/agents/main/agent/auth-profiles.json
Shell env
Providers w/ OAuth/tokens (1): anthropic (1)

• anthropic effective=profiles:~/.clawdbot/agents/main/agent/auth-profiles.json| profiles=1 (oauth=0, token=1, api_key=0) | anthropic:default=token:sk-ant-o...xxxx
  CAuth/token status
  (node: 13057) [DEP0040] DeprecationWarning: The punycode" module is deprecated. Please use a userland alternative instead.
  (Use node --trace-deprecation ..." to show where the warning was created)
• anthropic usage: error: HTTP 401: Invalid bearer token
• anthropic:default static

[wispy tree]

welcome brother

lmao

i kid you not ive been troubleshooting this for 12 hours

and fear i know the problem but dont want to admit it

claude earlier told me that anthropic changed their policy to not allow connections to anything except claude code with Oauth and only let other use API keys whihc would basically not allow anyone to generate sid tokens (only oauth) so old accounts can but new cannot

but i might also be hakucinating

[keen hollow]

@unreal bear try this

cd ~/clawd/skills/claude-connect
./refresh-token.sh --force

the --force makes a differnce

[wispy tree]

can i swear in here or

bc

holy fuck i finally fucking got it working @keen hollow

[keen hollow]

@wispy tree can you try this as well

OMG

YES

[wispy tree]

yeah

i can share my diagnostic

[keen hollow]

Yes please

My guess here

Or really what I just found that is

./refresh-token.sh

• Tells you the time of expiry
• Schedules refresh

But if your auth is already invalid, it doesn't work regardless

Whereas ./refresh-token.sh --force

It forces refresh + sync right this second

It's not just chekcing

[unreal bear]

Whaaat

It says no credentials found in keychain

I have to use keychain?

[keen hollow]

No no

[wispy tree]

**Fixing Clawdbot + Claude CLI OAuth **Authentication (401 Errors)
The Problem
You're trying to use Clawdbot with your Claude CLI subscription (OAuth token) instead of a raw API key, but you get:
    •    HTTP 401: authentication_error: Invalid bearer token
    •    No API key found for provider "anthropic"
Prerequisites
    •    Claude CLI installed and authenticated (claude auth completed)
    •    Clawdbot downloaded from source
    •    Node 22+

Step 1: Install Clawdbot CLI (not the desktop app!)
Don't use Clawdbot.app - it's the GUI version. Build the CLI from source:
cd ~/Downloads/clawdbot-2026.1.22  # or wherever you extracted it
npm install -g pnpm                 # if you don't have pnpm
pnpm install
pnpm build
pnpm link --global
Verify it works:
clawdbot --version

Step 2: Install the claude-connect skill
This bridges your Claude CLI OAuth token to Clawdbot:
cd ~/clawd/skills/claude-connect
./install.sh
This sets up automatic token refresh every 2 hours.

Step 3: Fix the auth profile ID
The profile must be named anthropic:claude-cli (not anthropic:default).
Edit ~/.clawdbot/agents/main/agent/auth-profiles.json:
{
  "version": 1,
  "profiles": {
    "anthropic:claude-cli": {
      "type": "token",
      "provider": "anthropic",
      "token": "YOUR_TOKEN_HERE"
    }
  },
  "lastGood": {
    "anthropic": "anthropic:claude-cli"
  },
  "usageStats": {}
}

Step 4: Fix the main config
Edit ~/.clawdbot/clawdbot.json and update the auth section:
"auth": {
  "profiles": {
    "anthropic:claude-cli": {
      "provider": "anthropic",
      "mode": "token"
    }
  }
}
Important: Remove any anthropic:default with "mode": "api_key" - that overrides the token auth!

Step 5: Restart the gateway

Kill any existing gateway

kill $(pgrep -f clawdbot-gateway) 2>/dev/null

Start fresh

clawdbot gateway

Summary of Key Fixes
What was wrong
The fix
Using desktop app instead of CLI
Build from source with pnpm build && pnpm link --global
Profile ID anthropic:default
Change to anthropic:claude-cli
Auth type api_key
Change to token
Config mode: "api_key"
Change to mode: "token" in clawdbot.json
Old gateway process running
Kill it before restarting

Quick Diagnostic Commands

Check if gateway is running

pgrep -f clawdbot-gateway

View gateway logs

tail -f /tmp/clawdbot/clawdbot-$(date +%Y-%m-%d).log

Force refresh OAuth token

cd ~/clawd/skills/claude-connect && ./refresh-token.sh --force

Check auth status

cat ~/.clawdbot/agents/main/agent/auth-profiles.json

[keen hollow]

go to claude

[wispy tree]

written by cowork

[keen hollow]

• Open terminal
• run "claude" to open Claude Code CLI
• Run "/login", which gives you a new crednetials. This is a must.

@unreal bear

[unreal bear]

Ok

[keen hollow]

Then run the

cd ~/clawd/skills/claude-connect
./refresh-token.sh --force

And it should work

@wispy tree thank you! I'll incorporate this into my installation process

[unreal bear]

It gives me a lot of red

[keen hollow]

can you share

[unreal bear]

Lots of python3 frameworks and 0Auth token revoked

Bash(./refresh-token.sh --force)
L Error: Exit code 1
Traceback (most recent call last):
File "‹stdin›", line 3, in ‹module›
File "/Library/Developer/CommandLineTools/Library/Frameworks/Python3.fram
work/Versions/3.9/1ib/python3.9/json/init.py", line 293, in load
return loads (fp. read(),
File "/Library/Developer/CommandLineTools/Library/Frameworks/Python3.fram
work/Versions/3.9/lib/python3.9/json/init•py", line 346, in loads
return _default_decoder.decode(s)
File "/Library/Developer/CommandLineTools/Library/Frameworks/Python3.fram
work/Versions/3.9/1ib/python3.9/json/decoder.py", line 337, in decode
obj, end = self. raw_decodels, id=_w(s, 0). end())
File "/Library/Developer/CommandLineTools/Library/Frameworks/Python3.fram work/Versions/3,9/1ib/python3.9/json/decoder.py", line 353, in raw_decode .. +17 lines (ctrl+o to see all)
L Outh token revoked • Please run /login

[keen hollow]

@unreal bear can you go to your keychain access through application, then enter claude code credentials, and choose "show password" and see if you have access token inside

[wispy tree]

fuck

it broke again haha

that was like 10 mins of euphoria

[keen hollow]

sad

is it still 401

[unreal bear]

Im not logged into appleid/keychain.

[keen hollow]

oh...

hmm

i dont get it

what do you mean

[wispy tree]

no its agent failed luckily

I think cowork is fixing it

[unreal bear]

Isn’t keychain the apple service for password

[keen hollow]

Yes it is

You should be able to open the app on your macbook tho?

do you have this

[unreal bear]

Im in,but dont have that entry

[keen hollow]

right

then need to login claude code cli again

once you do check keychain again and see if you have the entry now

reemmber to NOT run clawdbot config after doing that

It should be

• Run "Claude"
• Run "/login"
• See credentials appear in keychain
• Run

cd ~/clawd/skills/claude-connect
./refresh-token.sh --force

That's it

[shy mango]

This worked, thank you so much!

[keen hollow]

HELL YEAH

LETS FUCKING GO

so happy for you

enjoy clawd

[limpid cliff]

hey yall! i've got imsg working, but no matter how many times i tell it not to, it's sending both exec output and markdown formatted messages in the channel lol - has anyone had luck making this smoother?

(running on minimax for now, maybe that's part of the problem 😛 eventually i want to be able to have a "smart" way for it to swap models)

[unreal bear]

What the fuckkkkk, I get authenticated and there’s an entry in keychain, but there’s no api key inside the file,

[keen hollow]

No you don't need api key

it should be "token"

[unreal bear]

It revokes my token very often

Is that normal?

[keen hollow]

No never seen. Can you describe what you did

Like which commands you ran

Step by step

[wispy tree]

we're so back again

[unreal bear]

claude
• Bash(claude) timeout: 10s
L Interrupted • What should Claude do instead?
/login
Login successful
cd~/clawdbot/skills/claude-connect
I'm in the claude-connect directory. What would you like me to do?
./refresh-token.sh --force
• Bash. /refresh-token.sh --force)
L === Claude OAuth Token Refresh ===
[2026-01-23 22:09:33] Refresh started
[2026-01-23 22:09:33] Reading tokens from Keychain...
• +46 lines (ctrl+o to expand)
L Outh token revoked • Please run /login

[keen hollow]

eh?

working on it

[keen hollow]

LETS GOOO

@unreal bear I think you're having claude code cli run this for you right

[wispy tree]

co work said if it ever breaks again just run the following

cd ~/clawd/skills/claude-connect && ./refresh-token.sh --force
kill $(pgrep -f clawdbot-gateway) 2>/dev/null; clawdbot gateway

[keen hollow]

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

I'm trying to see if it's a claude code problem

I'm 99% sure that's why it's not working

Let me know if it is correct and I can confirm what the problem is

[unreal bear]

Yep

[keen hollow]

YES

Ok.

[keen hollow]

Now try this

No need to close claude code cli after login. Just leave it there and right click open a new terminal then run the command

[unreal bear]

Omg it did a thing and didn’t revoke me

[keen hollow]

lmao

[unreal bear]

Token refreshed successfully

[keen hollow]

I'll tell you why later if this works

YES

Ok

Try if you can reach it with your message

adn if it respondds

[unreal bear]

Oh my god

YES IT WORKS

[keen hollow]

FUCK YES

[unreal bear]

you’re a fuckn hero

Thank you!

[keen hollow]

It didn't work because when we refresh it revokes old token and get a new one

If running through claude code cli, it stops refresh mid process. that's why you need to run on new terminal

You helped me discover an edge case I didn't think of

TYSM

Enjoy Clawd!

[unreal bear]

Ahhhhhhhh

Cooool!

[wooden raft]

Just had an experience with my bought I am trying to make a group chats with me and other people and clawd and I’m having a massive issue with bleed through and it’s getting really confused and it’s sending messages that we’re supposed to go to me to other numbers that were in the group chat separately and it’s having like eight different conversations at once that’s a little bit of an exaggeration, but anyway I’m building the system right now to stop the bleedthrough. I will post for you guys once it’s done. I’m doing a bunch of tests on it now.

[lapis laurel]

when using mac's companion app, is it recommened to install cli via that? or is it fine to install manually and add app later?

[silent oasis]

hey guys woke up this morning and seems like i've been logged out of clawdbot (getting Agent failed before reply: All models failed (2): anthropic/claude-sonnet-4-5: OAuth token refresh failed for anthropic: Failed to refresh OAuth token for anthropic. Please try again or re-authenticate. (auth) ) message. How do i re-authenticate? Note: i'm not using API key, just my regular subscription plan. Can someone help me with this? thanks 🙂

[halcyon temple]

similarly not able to connect via my cc max as a new user

ive ran claude and then /login but in a fresh onboard flow here's what i see:

[granite onyx]

Wtf

You have hit your ChatGPT usage limit (team plan). Try again in ~7330 min.

[keen hollow]

https://clawdhub.com/TunaIssaCoding/claude-connect

follow this and should fix it

Let me know if it doesn’t and I’ll sort it out with you

[keen hollow]

@halcyon temple this should help you too.

Do exactly what it says and it’ll run

[cunning scroll]

Hello friends! I'm about to get a minimax starter subscription if someone wants to send out a referral to me. I'd obviously be down to start a referral train assuming it's not against the rules

[stoic nacelle]

Anyone else run into the issue where all the tool calls just get sent as text instead of running? Happens on all models and channels (discord, telegram and WhatsApp). Just started happening this morning

[low dragon]

#old-models right in channel header and pinned msg

[low dragon]

can't be else I'd not be able to run it. update, if that doesn't help fire up claude or codex and ask why

[devout marsh]

Hi! How can I get started setting up Clawdbot on my Mac mini at home so that I can message Clawdbot from my phone, and have code on Claude code? Do I run it on Mac mini and start adding connectors to messaging apps?

[wanton gate]

think i was able to set this up properly - but was prior on API and now have my keychain updated, how do i make sure clawd is pointed over to my max subscription?

https://github.com/TunaIssaCoding/claude-connect

@keen hollow

[keen hollow]

Can you check “clawdbot models” and see which model it’s using?

[wanton gate]

hmm i guess token was revoked it was just working

just did this again - now what?

[keen hollow]

After this run Clawdbot models again to see if it’s in use

[wanton gate]

same thing

[keen hollow]

Hmm that’s weird

I’ll be home in 1.5 hours

I’ll replicate and get back to you

[wanton gate]

thank you appreciate

[coral nimbus]

Hi everyone,

I’m a super nontechnical guy so I apologize if my questions seem a little green .

I got everything installed and set up (I hope) but have a few questions. Would anyone mind dm me or hoping on a call? I’ll pay

I am too the point where I can communicate with Clawdbot via slack and it has access to my data

[zealous raptor]

Which provider?

[cunning scroll]

Found it. Thanks!

[coral nimbus]

@keen hollow can you DM me?

[keen hollow]

I'm suddenly getting 401 error too on my tokens when it ran fine and refreshed on its own last night. Figuring it out. Will prob take a bit of time

[stoic rain]

@low dragon the Mac desktop helper crashes every time I open it in menubar - as in, app gets killed - there are no errors visible anywhere - updated to latest version too and same issue... any idea how to debug? are logs for that kept anywhere i can check maybe?

in CLI its all fine still

just whenever i try to open up the desktop helper, boom the desktop helper is killed

[keen hollow]

ah

I know

nano ~/.clawdbot/agents/main/agent/auth-profiles.json

Do you see a block exactly like this

"profiles": {
"anthropic:default": {
"type": "token",
"provider": "anthropic",
"token": "[access token]"

or only this

"profiles": {
"anthropic: claude-cli": {
"type": "oauth",
"provider": "anthropic",
"token": "[access token]"

@wanton gate I updated. Update the skill then run it again by using these in your terminal

clawdhub update claude-connect
cd ~/clawd/skills/claude-connect
./refresh-token.sh --force

Let me know if it works!

[stoic rain]

Found it. The crash is an assertion failure in the syntax highlighting code:

assertionFailure(:_:file:line🎏)
closure #1 in variable initialization expression of static NSBundle.module
CodeTokenizer.init()
HighlightedTextFragment.Model.tokenize(content:languageHint:)
What's happening:

App connects to gateway ✓
Gateway sends some content to display
App tries to syntax-highlight that content
CodeTokenizer initializes and tries to access Bundle.module
Bundle.module can't find its resources → assertion failure → crash
This is a bug in the macOS app — looks like a resource bundle packaging issue. The syntax highlighter's language definitions or tokenizer data aren't being found at runtime.

Workaround options:

Check if there's a newer version of the app
Report this to Clawdbot (Discord or GitHub) with the crash log
Might be able to disable syntax highlighting if there's a setting
This isn't something you can fix on your end — it's a build/packaging issue with the app. Want me to check if there's a GitHub issues page or help draft a bug report? 🦀

---

cc @low dragon (MacOS app crashes for me)

or idk if theres a better place to report actual bugs like this? maybe github?

[formal fiber]

Hmm. Anyone running Clawdbot on WSL? Wondering whether I should buy a VPS for it or just run it on my PC via WSL to try it. Also, what is the best way to try/use without adding up API costs? I have a Claude Max $100 sub and ChatGPT Pro for Codex...

[left aspen]

how do i install the clawdbot node app for my iphone? clawdbot is saying to get phone location data i need it

[abstract flicker]

You can get free vms from all over the place, aws, oracle, google cloud etc. WSL would also probably work well, to try things out

[next elbow]

how often to people run /new ? i'm new here and obviously in llm clients i just open new convos all the time.... is that not a good idea in clawdbot, or fine to do regularly, especially early on when messing around

and i had switched to sonnet because setup / chat was churning through my pro subscription fast... just did a /new and it switched back to opus 🧐

[lapis laurel]

Anyone using mac companion app? Installed cli with it but app wont open at all after that

[dapper tundra]

Just set up Clawdbot for first time and opted for the Minimax Plus plan.
Running on an M1 Air and the response when using Telegram is pretty slow

any ideas?

[pseudo fractal]

hello is their a definitive installation guide anywhere to get it on your mac and pointing at your claude pro sub? ive installed everything put my claude key in about 100 times and i just cant get it going, im missing a step somewhere

[twin snow]

Hi , im trying this set up :
Clawdbot + Ralph + qwen3-coder as promery and opus-4.5 as fallback
and i get really bad results , i mean qwen3-coder feels super stupid can do much and it hard to set up tools and things , does anyone have a good local setup he recommends ?

[stoic rain]

same issue, i reported it on github

im going insane trying to get slack working... anyone have any suggestion for debugging? Clawd tells me it sees my messages and the agent gets them ok, but i dont get any replies back in slack, and clawd says it doesnt see any obvious errors 🙃

[lapis laurel]

Asked claude code to fix it and now it works somehow lol

[stoic rain]

Found the bug! Look at the timing:

Slack run (16:17:19):

• messageChannel=slack
• Agent runs, completes in 7.3s
• NO chat.send or response

Webchat run (16:17:47):

• messageChannel=webchat
• chat.send 63ms ✅

The response is going to webchat instead of Slack!

The agent processes the Slack message but routes the response to the wrong channel (webchat instead of slack). The sessionKey=unknown confirms this is a routing bug.

This is a Clawdbot bug — not a config issue. The Slack setup is correct, but the response routing is broken.

i have no idea where to start with this fml lol

[heady dawn]

the docs say there is an iOS app

Availability: internal preview. The iOS app is not publicly distributed yet.
any way to get access to it?
https://docs.clawd.bot/platforms/ios

[flint mica]

Still in development! Not ready for release yet

[heady dawn]

lmk if i can help beta testing

[flint mica]

Need more people to help contribute to building the apps

[zealous raptor]

you need more clawds

[woeful helm]

there is a #1459642797895319552 channel or open a github issue

happy to assist you in opening the "ticket". I want to use this feature as well

[stoic rain]

Based on my investigation:

Finding 1: The cron jobs are different from heartbeat
The cron jobs (daily-exec-briefing, daily-ai-briefing, daily-personal-briefing) run at 6:30 AM, 6:38 AM, 6:45 AM daily — not the 30-minute heartbeat.

Finding 2: No heartbeat scheduler exists
There's no cron job or scheduler configured for every: "30m". The heartbeat config says 30 minutes, but nothing is triggering it.

Finding 3: Only web-heartbeat runs
The only heartbeat in the logs is web-heartbeat (gateway keep-alive every 1 minute). No agent heartbeat runs (email/calendar checks).

Conclusion: The heartbeat feature appears to be broken or misconfigured. The config says every: "30m" during active hours, but:

1. No scheduler creates heartbeat runs

2. No agent runs are triggered for heartbeat checks

3. The cron jobs are daily briefings at 6 AM, not heartbeats
   Possible causes:

4. Heartbeat requires a specific session or target that doesn't exist

5. The target: "last" setting might not work without a "last" channel

6. Bug in heartbeat scheduler
   Fix options:

7. Create a manual cron job for heartbeat

8. Set a specific target channel (like WhatsApp or Telegram)

9. File a bug report about the heartbeat scheduler not working
   Want me to try setting a specific heartbeat target, or investigate further? 🦀

—

Anyone else had an issue getting Heartbeats to work? I’ve set them for every 30m but they never seem to actually run. Clawd says the above. My config seems correct (had Claude code check it too). Do I need to set up a cron or something?

[gusty sonnet]

Hearbeat doesn't use crons. It's run by the gateway in process, so it should just work.

[pliant elk]

How do you guys use it on a Linux VPS? It’s currently sitting in a docker but I can’t seem to confugure it correctly.

Keeps missing permissions, loosing skills and memory’s and so on.

[worn meadow]

ValueError: Message has tool role, but there was no previous assistant message with a tool call!

using openrouter minimax-m2.1 is working fine, but hitting error with litellm proxy + vllm with minimax-m2.1

[royal night]

Tried clawd on an old ~2017ish macbook a few weeks back but didn't have success. trying again today on lenovo t480. Can I still use my claude code max sub?

Edit: Yes looks like it does 🙂

[sharp steeple]

is it me or or are the docs website down ?

[hazy halo]

I'm getting (no output) after completing the setup

[jagged mist]

It's working for me

[sharp steeple]

My ISP was blocking it for some reason. Thanks for confirming!

[brazen juniper]

hi guys! i having some trouble setting up the anthropic model via the claude code cli token sharing.... i granted access to all apps in the keychain to the Claude-code credentials , but completing that step in the config does not updating the values in ~/.clawdbot/agents/main/agent/auth-profiles.json, any help here? thanks in advance!!

[unreal bear]

Are there any fixes for running out of tokens in 10 minutes after only like 7 or 8 prompts

[stoic rain]

My MemoryFlush is never triggering, it hasn't since I set Clawd up 2 days ago, we've tried increasing the threshold and whatnot, but nothing.

I had Claude Code investigate logs etc:

Confirmed: memoryFlush NEVER triggered

Looking at the session transcript for b7a4868c-0ba7-4db8-b3d9-9d8e65350096, I found:

Multiple Compactions Happened:

09:20 - 190,705 tokens → Compacted
10:11 - 192,890 tokens → Compacted
13:01 - 193,015 tokens → Compacted
19:07 - 178,344 tokens → Compacted (the one we saw in real-time)

Evidence of memoryFlush: ZERO

grep -c "Pre-compaction memory flush" <session-file>

Result: 0

The session metadata confirms:
{
"compactionCount": 1,
"memoryFlushAt": null, ← Never ran
"memoryFlushCompactionCount": null ← Never set
}

---

And also had it check my config, which seems fine

Why is memoryFlush NEVER triggering for me despite lots of chances for it to?

Because of this, my daily notes never get updated 🙁

---

Both Claude Code & Clawdbot themselves can't solve this one.

MemoryFlush (writing to memories before compact) has never worked for me from installing Clawdbot 2 days ago

Anyone know how to debug or where to start w/ this?

[formal fiber]

Hmm. Are most skils configurable in the onboarding wizard Mac only? I installed SAG and AI is telling me it's Mac only.

[small smelt]

I just installed clawd on a vps. I need to access the files to set some settings.. what is the path and name of the file i edit to edit the config?

{
channels: {
discord: {
enabled: true,
token: "YOUR_BOT_TOKEN"
}
}
}

[flint mica]

You can do this in the web GUI

[small smelt]

I'm using twingate, need to set allowInsecureAuth since technically i'm accessing it http although within vpn

[formal fiber]

I'm DMing my bot on Discord. Do I just do /new to start a new context, or does it manage context pretty good automatically and I don't need to worry about it.

[pure bison]

Hi, welcome community. Experimenting today with my first Clawdbot setup but encounter some issues. After getting first an billing error something went wrong. How can I get my Clawd back to working?

Agent failed before reply: No available auth profile for anthropic (all in cooldown or unavailable).
Logs: clawdbot logs --follow

[mental mica]

Pretty much don’t have to worry about it

[glacial jay]

Is general best practice to run clawdbot with its own apple id, logins, etc...? Then add your own machine as a node?

[astral garnet]

After spending hours working with macos today I found if I only downloaded the dmg from releases I would have saved myself a lot of time.

[mighty mortar]

how can i control the bot from my phone bot not through whatsapp? like somewhere i own the data, my own client

[astral garnet]

Web control

[mental mica]

We have so many different channels that you can use, look on the docs

[formal fiber]

I just set it up through Discord and am just DMing it that way. Not sure what to set up to access proper on the iPhone though. I'd like to set up WhatsApp but don't care about getting a separate eSim and don't want friends messaging it on accident.

[gusty sonnet]

You don't have to have a separate sim/icloud account. You can use your own number, but it'll look like you're messaging yourself.

[subtle depot]

I just restarted my mac mini and clawd is no longer running. I ran clawdbot gateway restart and it says DEP0040. Punycode module is depreciated and it is not working. How do I fix this?

[raven dock]

Guys how can I solve this? I am trying to connect remote gateway via tailsclae

[flint mica]

Need to set token

[raven dock]

How? Which token to put where

[gusty sonnet]

clawdbot config set gateway.remote.token <token>

You should find a token is already set in clawdbot.json on the machine running the gateway.

[raven dock]

THANK YOU

[knotty elm]

I can't figure out how to get Clawbot on my mac mini through terminal. Where can i hire someone to do this for me?

[stoic skiff]

Did anyone struggled with sonnet not being an "allowed model"? This is what clawd said when I asked to use sonnet instead of opus:

 change your model to sonnet



 📊 Session Status
 anthropic/claude-sonnet-4-20250514
 {
   "status": "error",
   "tool": "session_status",
   "error": "Model "anthropic/claude-sonnet-4-20250514" is not allowed."
 }


 Looks like Sonnet isn't on the allowed models list for this session. You're currently running on
 Opus, and the config seems to restrict switching to other models.

 If you want to enable Sonnet, you'd need to update your Clawdbot config to add it to the model
 allowlist. Want me to check the config and see what models are currently allowed?

[gusty sonnet]

It's telling you what to do about, though.

[zealous raptor]

Does everyone agree telegram is the best client for clawd on the go?

[formal obsidian]

I am getting a disconnected (1008): pairing required error, I am using a cloudflare tunnel on my gateway... Any idea?

[stoic skiff]

Who defines this allowlist? I thought it's something the API vendor decides (Antropic in my case)

[formal fiber]

What happens when I hit rate limits? Will the bot alert me before it happens and stop itself? Or just log 429's in a loop? I would hope that wouldn't be the case but figured I'd ask.

[gusty sonnet]

It's the 429s.

[magic merlin]

Custom Provider baseUrl Not Being Used - Requests Never Reach Proxy

Hi! I'm having an issue where my custom provider's baseUrl appears to be ignored - requests never reach my proxy server despite correct configuration.

The Problem:

• I configured a custom provider with a baseUrl pointing to my API proxy
• clawdbot models list shows my custom model as available
• Direct curl to the proxy works perfectly (returns valid response)
• But when I select the model in ClawdBot, I get "403 Your request was blocked"
• My Cloudflare logs show zero requests reaching the proxy - ClawdBot isn't hitting my baseUrl

What Works:

curl -X POST "https://my-proxy.example.com/v1/chat/completions" \
  -H "Authorization: Bearer sk-proj-..." \
  -d '{"model": "gpt-5.2-2025-12-11", "messages": [{"role": "user", "content": "hi"}]}'
# Returns: {"choices": [{"message": {"content": "Hi!"}}]}

My Config (~/.clawdbot/clawdbot.json):

{
  "models": {
    "mode": "merge",
    "providers": {
      "myproxy": {
        "baseUrl": "https://my-proxy.example.com/v1",
        "apiKey": "sk-proj-...",
        "api": "openai-completions",
        "models": [{
          "id": "gpt-5.2-2025-12-11",
          "name": "GPT-5.2 (Proxy)",
          "contextWindow": 128000,
          "maxTokens": 16384
        }]
      }
    }
  },
  "agents": {
    "defaults": {
      "model": { "primary": "myproxy/gpt-5.2-2025-12-11" }
    }
  }
}

Observed Behavior:

• OpenRouter (built-in provider) works perfectly
• My custom provider shows up in model list
• But actual API requests seem to go elsewhere (not my proxy)
• The 403 error suggests requests might be hitting api.anthropic.com instead?

Environment:

• ClawdBot version: 2026.1.23-1
• macOS

Is there something special needed to make ClawdBot actually use a custom provider's baseUrl? The config matches the docs but requests aren't reaching my server.

[solid wave]

Looks like little clawd broke after the update (21.2 to 23.1) for me

When trying to run any of the command in terminal it’s giving me a different path than my original (standard) install

I’m on Linux (Raspberry Pi)

[gusty sonnet]

Try: clawdbot config set agents.defaults.models.anthropic/claude-sonnet-4-5.alias Sonnet

[delicate flint]

is anyone having issues with the crons? I tell it "remind me something in 5 min" and it tells me that it will, creates a cron, but then never messages me lol. GLM 4.7.

[coral nimbus]

I’m having an issue where my clawdbot window is super slow when I’m typing.

I’ve done a new Channel, I’ve refreshed, hard refreshed, closed it out. And it’s still doing it. It’s only this window

[zealous raptor]

Try a higher model to rule anything out with the model’s iq

[delicate flint]

"model's iq" lmaooooo yeah honestly it could 100% be the model

[stoic rain]

My MemoryFlush is never triggering, it hasn't since I set Clawd up 2 days ago, we've tried increasing the threshold and whatnot, but nothing.

I had Claude Code investigate logs etc:

Confirmed: memoryFlush NEVER triggered

Looking at the session transcript for b7a4868c-0ba7-4db8-b3d9-9d8e65350096, I found:

Multiple Compactions Happened:

09:20 - 190,705 tokens → Compacted
10:11 - 192,890 tokens → Compacted
13:01 - 193,015 tokens → Compacted
19:07 - 178,344 tokens → Compacted (the one we saw in real-time)

Evidence of memoryFlush: ZERO

grep -c "Pre-compaction memory flush" <session-file>

Result: 0

The session metadata confirms:
{
"compactionCount": 1,
"memoryFlushAt": null, ← Never ran
"memoryFlushCompactionCount": null ← Never set
}

---

And also had it check my config, which seems fine

Why is memoryFlush NEVER triggering for me despite lots of chances for it to?

Because of this, my daily notes never get updated 🙁

---

Both Claude Code & Clawdbot themselves can't solve this one.

MemoryFlush (writing to memories before compact) has never worked for me from installing Clawdbot 2 days ago

Anyone know how to debug or where to start w/ this?

[dreamy stag]

Good morning/afternoon! I would like to know if it is a good idea to use clowdbot as my assistant to program via my phone

[stoic rain]

Any big brain peep can think of what to look at for the above? 💀

[granite gazelle]

hello, clawde just ate all my calude pro sub tokens in few messages, is it always so hungry?

[gray kraken]

You set it up via API?

[gusty sonnet]

Yes.

[granite gazelle]

sub oauth token

[meager tundra]

i got clawdbot running but now its not responding... but the dashboard is visible. just shows loading chat indicator. any suggestions?

[low dragon]

that's an area that still needs work. don't expect a finished product. PRs welcome!

Folks, the best way to get unstuck is using hackable install on the website, open claude/codex on it and ask questions.

[low dragon]

it's not a good mode. don't expect it to be anything more then meh

[low dragon]

if you can figure out what went wrong, would love a detailed issue or PR!

[next sable]

Just opened this Q on signal-cli: #1464760439152967793 message

Anyone got this running on a Raspberry Pi 5? I'm thinking it's not a compatible binary?

[rich palm]

https://clawdhub.com/i3130002/edge-tts

I've been working on this to have a free TTS

But I'm not sure why I cannot upload my codes.

[delicate flint]

I see. Which one would you recommend? Claude Opus gets expensive very fast

[low dragon]

we made it for tailscale. if you use sth else you gonn have to debug

[low dragon]

Yeah but it's by far the best model.

[formal fiber]

Interesting. For the most part my bot has been sending main messages. Weirdly though sometimes it will start sending i guess the thinking output? Where it outputs the commands and everything its doing every few seconds and it looks like spam. Is there a way to prevent that or is it unavoidable from time to time?

[gaunt glade]

can i use a local ai with clawdbot? i've been searching but havent found a way to connect it

[delicate flint]

What would be the best model that isn't crazy ass expensive? I assume the $40/mo github copilot plan could work but I use that to work...

[low dragon]

sounds like you use a crap model

github doesn't support other clients officially so you will run into loads of issues

use sonnet of opus is too expensive

[formal fiber]

Using Opus here. Only happened once.

[low dragon]

it's literally all in our docs.

[rich palm]

You can with Ollama
https://docs.clawd.bot/concepts/model-providers#ollama
Suggest you to follow some videos

[low dragon]

or you do this

[gaunt glade]

thanks

[rich palm]

https://clawdhub.com/i3130002/edge-tts

[delicate flint]

ah, I see. It does support opencode (very recent thing) so I assumed it maybe supported clawd too. Didn't Anthropic made it against ToS to use the subscription on something other than claude code though? We'd have to pay the API, right?

[solid wave]

Yes sir! I will try to debug and if not I will at least document what I did

[gusty sonnet]

They did. But they don't appear to be enforcing it much.

[paper zealot]

Is anyone else having the oauth issue with anthropic? Getting 401's . Troubleshooting for over an hour now. I am running clawdbot in AWS.

[delicate flint]

I see. So I assume most people here is just using the sub on Clawd?

[gusty sonnet]

For the most part. It's outrageously expensive to run via the API.

[rare kindle]

Hello...
How completely stop then start?
the locale webserver it create seems to be always working...

[low dragon]

ot works. IANAL. gotta decide for youtself

folks, all I do is literally typing in your questions into codex and let it write docs.

You can just ask AI.

cc @rare kindle

[delicate flint]

lmao

[clear ermine]

what kind of security should I setup on my cloud instance running clawd?

[torn pelican]

my clawdbots been stuck on this for a while 😢 should i keep waiting or do i have to do something to continue/ui bug

[round fox]

Do clawdbot tui and may be see there

Or just ask what’s the status

[torn pelican]

im using the web ui because the tui literally crashed

on onboarding

and i want to continue

im just rly confused 😿

[round fox]

Restart clawdbot I never had issue with tui

[torn pelican]

okay

yeah i got a [clawdbot] Uncaught exception: Error: Rendered line 595 exceeds terminal width (233 > 213).

[dawn bear]

Anybody set up Clawdbot with StartOS?

[low dragon]

pls read https://docs.clawd.bot/gateway/security

[low dragon]

they messed up your config. you play a dangerous game. call clawdbot doctor

[low dragon]

pls send an issue else I can't fix that

[torn pelican]

it offered to do that i was trying to onboard and make my models not show every model that doesnt have auth 💔

[wanton gate]

@keen hollow did you figure out the token issue?

@keen hollow sorry ur message got buried just saw it. think i am good can u confirm?

[low dragon]

yup. litltle monsters. reinstall from website

[full pendant]

Can you use Clawd with a free api from Groq for testing?

[light wraith]

And recommendations on how to get the anthropic subscription signed in it keeps going to my API key

[low dragon]

Then your experience will be so bad that you never want to use the product again

[paper zealot]

Having quite a few issues setting up oauth with claude for the max plan. Is there a clear troubleshooting guide for this.

[low dragon]

read the docs, use setup-token

open claude in the folder and let it setup. or even better, codex

[paper zealot]

Will this work if I have an AWS EC2 with Clawdbot? I only have installed Claude CLI locally on my macbook atm

[low dragon]

I retweeted that on the official x account, look

[wanton gate]

@keen hollow busted again after a min, do you have this owrking riught now?

[round fox]

Check your clawdbot.json for the config I guess

Or Doctor

[rare kindle]

@low dragon I did but the web UI wasstilll working somehow... soI asked here...

Currently reinstalling with pnpm global...

Honestly have no idea what am doing at this point...

Telegram bot was working find with Opus... then one message hit like 100k token alone...

[low dragon]

https://docs.clawd.bot/gateway/troubleshooting https://docs.clawd.bot/providers/anthropic

[low dragon]

what model do you use/provider. that is sure not from anthropic or openai

If you use Antgravity you're on your own and need to debug/fix this. I don't use it, there's bugs

[rare kindle]

alright... II use Antigravity...

Thanks

[torn pelican]

issue posted yell at me if its sloppy if i need to add more but i think its sufficient 😌

[rare kindle]

ohhhwoaaw... my opus now will reset January 31... Ohh well

Seems like Antigravity Opus hit limits faster lately...

[torn pelican]

good use for codexbar 🙂

[dawn bear]

Any documentation if Clawdbot can run on StartOS?

[low dragon]

no you gotta have to figure it out. please do and send a PR with your learnings to help others

[void wraith]

I have clawdbot running on my Mac mini with google Gemini api. I can access my gateway dashboard but I’m having trouble figuring anything else out.

Trying to speak to the agent in the chat window on the gateway dashboard doesn’t really get any responses. I’m not sure how to send it iMessages even though I installed that service

[olive helm]

are you also getting this issue with imsg? #1459860817976492125 message

[low dragon]

try bluebubbles plugin instead

[void wraith]

I’m not. I think I have things working now

I didn’t see anywhere in the instructions that I had to send myself an iMessage

[low dragon]

open claude or ask your bot to set it up. use hackable insteall

[mint plover]

hey guys, just tried setting it up and everything seemed to be working but it won’t “hatch”. just stuck on “wake up my friend” step for 15 minutes now

[void wraith]

I think that allowed me to activate iMessage

[void wraith]

What is hackable?

[low dragon]

https://docs.clawd.bot/help/faq#what’s-the-difference-between-the-hackable-git-install-and-npm-install

really folks. don't make me do the link searching for you. it's either me playing google for you or me making clawdbot better

[void wraith]

Sorry

[misty beacon]

Setting up Clawd for the first time. Used claude code login with max account. Chat in the dashboard just displays bouncing dots when sent a message. Setup whatsapp and sent a message. Got back a message Agent failed before reply "No API key found for provider "anthropic" Anyone else had this issue?

[low dragon]

did you read these

[mint plover]

thanks - yeah, the claude setup seems to have gone fine, i was able to authenticate and select a model. but not able to chat.

oh sorry you were replying to the other message mb

[low dragon]

Commands that help: clawdbot status, clawdbot models status, clawdbot doctor

[next sable]

I think I'm close to giving up on the Pi 5 setup. Seems like nothing will start the gateway successfully

[misty beacon]

Thanks. Didn't mean to make you search for it lol. I just had not found that yet for some reason.

[low dragon]

I mean you didn't pick the easiest path but lots of people managed. Where is it hurting?

My approach would be to open claude and tell it to install the gateway.

[tawdry fox]

hey, just tried to setup on aws using claude api, getting no output

[low dragon]

bedrock is super new and I don't have an account, please dig in or choose the common path, api key/sub

i would recommend the same tho. hackable install + claude

will unblock you

[wind ibex]

i created and AWS cloud server and copy and pasted the one line of code from Clawd.Bot and it didnt work

i can send a screenshot of what it says if that would help

[next sable]

Yeah - I definitely didn't. Tried cli, docker, and npm installs. Had issues with signal-cli compatibility at first, but managed to patch it with a symlink. Main issue now is that I can't get any response out of it over tailscale. Working through config changes to get it the gateway booted, but no response over tailscale yet

[low dragon]

oh that is some very advanced stuff. you defo need to dig.

[low dragon]

install claude in it and let it do the work.

[wind ibex]

omg duh

ur genius

[mint plover]

hmm i authenticated via claude code (also tried manually pasting the OAuth token) and it seems to log in but says reasoning auth missing.

[wind ibex]

i asked claude how to install itself in the AWS cloud server and it told me i cant do that

[wheat wave]

Hoping someone can help -- things have been running fine for me but very broken currently:

• Telegram can't seem to connect, lots 2026-01-25T01:20:28.442Z [telegram] setMyCommands failed: HttpError: Network request for 'setMyCommands' failed! in the logs even after wiping the config mutiple times.
• Trying to message clawd directly in the TUI just returns no output.

I even tried a full clawdbot reset and still having both issues.

Auth for both models and telegram are working fine, interesting, and the gateway comes up clean.

Any advice on what to do?

I also tried setting the export NODE_EXTRA_CA_CERTS=/etc/ssl/cert.pem in my clawdbot .env but still no dice

[warm thorn]

Last time I tried using Configure to switch from Claude code subscription to Minimax temporarily when I hit my rate limit:

• my Claude code auth configuration was wiped out, had to go into the json to figure it out again

• minimax was pretty useless, couldn’t seem to do even basic things

Is there a clear guide on this anywhere on how to do this well and switch back and forth between Claude subscription and minimax easily please?

[low dragon]

there's literally a video on our docs page for that

[low dragon]

no idea what that means

[low dragon]

try another computer. sth is messed up with yours or let claude rip.

https://docs.clawd.bot/help/faq#i’m-stuck-—-what’s-the-fastest-way-to-get-unstuck

[low dragon]

^ for everyone who is stuck

[wheat wave]

thank you, will give this a try! also thanks for everything with this project, it's mindblowing 🙂

[low dragon]

Thanks. Need to hear this a bit more. Right now it's a lot of complaining 🙂

[open jasper]

What's the minimum requirements for hardware for Clawd? I own a server hosting company and want to try and run it on one of our vps

[mental mica]

basically nothing

try it on the smallest one you have

[open jasper]

sweet we have some micros (4vcore x 8gb ram) what OS do you guys recommend running it on?

[lusty mulch]

don't listen to the complainers, the work you're doing is incredible man! ty for this and for merging the tlon PR so quickly. super nice to be able to use it within our app

[snow tide]

Will be using a raspberry pi 5 16gb to run clawdbot this week. Any tips or tricks

[low dragon]

I hear the Raspberry path is still a bit rough - bring patience and claude.

We certainly gonna make this buttery-smooth

(can make this joke here since clawd is not around)

[royal night]

If I want to install clawd on a second laptop, should I do a fresh install or install it as a node that connects to the first?

[wicked horizon]

Am newbie can i install on windows?

[low dragon]

depends what you wanna achieve!

[low dragon]

You can but you'll have a better time if you use WSL2

(still Windows, kinda)

[royal night]

I’m running on a thinkpad and doing some browser automation stuff but find that it’s quite a heavy task. Clawd will not respond while doing things. So I thought maybe I’d dedicate that one to the browser automation haha. And have another instance to use while it’s unresponsive

(Thank you for building clawd btw 🙏)

[mental mica]

Clawd will not respond while doing things
tell it to spin up a subagent instead of doing it on the main session

[next elbow]

how commonly to people use /new over /compact… daily? every major task switch? almost never?

[sick horizon]

I've read through the security docs but still seeing people talk about security risks and how vulnerable this makes you.
Question: Id be running on a VPS ....Beyond following the docs (pairing, sandboxing, file permissions), what else should I be watching for?
Anyone running this already - what security issues have you actually encountered that weren't in the docs or just more general tips?

[mental mica]

ive never used either except when debugging

[low dragon]

yeah that's an area we can still work on. You can ask it to use a subagent too.

[low dragon]

i don't use any of that

[low dragon]

I wrote what, 10k words in that doc. Have nothing more to add.

[quaint anchor]

bro if security is a huge concern you should A) not mess around with expiramental early stage software like this B) literally read the big message about security that appears directly after install

[sick horizon]

okay thank you. just seeing people talk online which caused concern.

fair

[low dragon]

I run this setup since May, but I also know what I'm doing.

[quaint anchor]

dreaming of waking up to a https://steipete.me/posts/2026/how-i-use-clawdbot someday

[low dragon]

If you don't know what prompt injection means, you should do some research first.

lol I ahve so much work to do that I'm the least sophisticated user by now

[quaint anchor]

haha fair

[low dragon]

cause googlin' is harrd these days it seems. or just asking the bot

[quaint anchor]

whats google?

[low dragon]

You think you are funny but it's actually really sad

[astral garnet]

How do you guys manage memories so well? Mine keeps forgetting things I’ve told it not to.

[quaint anchor]

"add to memory" until it remembers

[astral garnet]

I set up a voice path for whisper and piper and demand it remember and then a day later it goes I can’t transcribe that let me install some tools.

Clawd says skills are more relevant than memories so maybe that’s it

[void wraith]

I’ve got llama.cpp serving qwen on my tailnet. I’ve asked Gemini how to hook qwen into clawdbot instead of the standard model providers I see when configuring clawdbot via the command line. Is it possible to use self hosted LLMs or am I going down a pointless path here?

[astral garnet]

They are probably just not good enough at their small levels. You need a lot of context and tools

[glacial dagger]

I don't seem to have the skills/claude-connect folder. Anyone know how to get that built?

[astral garnet]

pull the last 30 minutes of this chat, see if you can understand why we completely lost our files. after you restarted the gateway the entire configuration was wiped. i had to manually reset it

HomeBot:

Oh no. I see it now. 😬

What I did wrong:

When I ran config.apply to rename to Homie, I only passed the agents section:
{
"agents": { ... }
}

The problem: config.apply does a full replace, not a merge. So I wiped out:
All channel configs (discord, imessage)
Env vars (HA_URL, HA_TOKEN, BRAVE_API_KEY)
Browser settings
Skills entries
Plugins
Hooks
Everything else

this was brutal, this erased my entire clawdbot config

[unreal hedge]

tried to end to end testing with sub agents didn’t work out properly ahaha

[vocal solar]

Has anyone figured out how to give clawd bot Claude code access so you can view it while it works in terminal

[unreal hedge]

why not ask clawdbot to build that for you

[mental mica]

tell it to use the coding-agent skill in tmux so you can attach to it

[desert solstice]

How important is it to run clawdbot on a dedicated machine? I have one I can use, but I'd have to get it all setup with my development environment, etc. Is it too scary to run it on my main box?

[mental mica]

https://docs.clawd.bot/security

[desert solstice]

Right, so allowing outside access from my home network to the machine where the bot can run any arbitrary command is no bueno 😄 A separate machine, it is!

[north gorge]

Newb question. I accidentally closed my clawdbot terminal. How do I restart it? I’m using windows powershell on a local computer

[astral garnet]

clawdbot status doesn't work?

[bronze apex]

newbie question, but can I use my Claude Pro subscription? or I'd need a separated subscription for Clawd only?

[north gorge]

“Disconnected from gateway: no reason”

[frozen token]

Guys new to ClawdBot here, I’m thinking of setting it up on a VPS, pls advise which VPS provider is the most efficient reliable and best for nearly 24/7 running loops

[leaden sage]

also new, but I think either clawdbot tui or clawdbot dashboard (web ui) will get you there.

[frozen token]

I think you can use your existing sub but I foresee an issue where Anthropic will crack down on this much like they did with OpenCode

[bronze apex]

I see, thanks!

[low dragon]

https://docs.clawd.bot/help/faq#i’m-stuck-—-what’s-the-fastest-way-to-get-unstuck

This should answer all questions I saw here so far. Give it a min to deploy

[full canyon]

i keep getting hit with this issue - session agent:main:main

Wake up, my friend!

(no output)
gateway connected | idle
agent main | session main (clawdbot-tui) | anthropic/claude-opus-4-5 | tokens 0/200k (0%)
────────────────────────────────────────────────────────────────────────────────────────────

────────────────────────────────────────────────────────────────────────────────────────

@silk ingot save brudda

[low dragon]

I updated the faq with even more stuff. Post questions that are missing. Or just be like me and paste questions into your local coding agent

[low dragon]

language. clawd is not in this channel.

[full canyon]

brudda means brother...

[sleek crypt]

Tomorrow il be setting up on a macMini, any advice, do’s and don’t ?

[lusty mulch]

anyone else getting a lot of ebadf errors

[vocal olive]

did anyone get it to work with lmstudio?

or ollama

[low dragon]

LM Studio works, but we only recommend big models. MiniMax M2.1 works in my tests.

[low dragon]

need more info to act than a word

[olive helm]

this worked btw, thank you. for anyone in the future, bluebubbles webhook doesnt work with localhost via url, needs the actual 127.0.0 etc

[vocal olive]

could you share your clawd config for the lmstudio?

i tried it with multiple models like gpt-oss, qwen 7b and they keep erroring out

[low dragon]

I will not share this for small models. That stuff is dangerous.

These small models also barely work re tool calls.

If you really wanna shoot yoruself in the foot, you have to do it yourself

[floral egret]

So I ran out of tokens with ChatGPT plus plan. I’m trying to set up my Gemini cli but I’m not sure how to set it up with oauth. I have client id and secret but where do I put it in clawbot.json?

[scenic moss]

I keep running into this issue when trying to install :/ anyone have any clue what's up or have similar issues?

[low dragon]

Are you pasting in the right command into the right terminal? Windows has two options and really you should use WSL2

[low dragon]

https://docs.clawd.bot/help/faq#im-stuck-whats-the-fastest-way-to-get-unstuck

[low dragon]

Gemini does not work well. I really recommend upgrading your plan

[low dragon]

I updated the FAQ, it will explain your issue.

[scenic moss]

Could you point me to where? I am poking around on here

[low dragon]

added a new faq entry for gemini, but beware that it's not a good experience.

[low dragon]

https://docs.clawd.bot/help/faq#windows-install-says-git-not-found-or-clawdbot-not-recognized

[scenic moss]

Finally got it installed!

[scenic moss]

Any suggestions for clawd dashboard not showing up?

[low dragon]

Please follow these steps https://docs.clawd.bot/help/faq#im-stuck-whats-the-fastest-way-to-get-unstuck

[thorn iris]

I feel really dumb. Doesn’t it defeat the purpose of putting this on a virtual desk top or Mac mini if you are giving it access too all your tools? Like email, Amazon, discord etc?

[fickle depot]

Is anyone using caffeinate or other means of preventing mac processes into going to sleep? I use mac mini but agents still stop responding to me after I'm away from it for some time.

[mental mica]

im using caffeinate

[odd spire]

could anyone help me with installation please? I'll share my screen

[low dragon]

Folks if you don't manage installation yourself, this is too dangerous and you should not use it.

[rich palm]

What issues are you experiencing?

[rich palm]

Can I please ask for help?

[crystal garden]

hey there, i installed locally claude through ollama, is there a way to use the dummy api key i used for that when selecting the model in clawd? many thanks in advance

[ivory rapids]

Is someone running clawdbot in a docker?

[mental mica]

https://docs.clawd.bot/install/docker

[ivory rapids]

Its more a user to user question 🙂

[mental mica]

Please y'all, take like 10 seconds to search on the docs

[ivory rapids]

you are may a bit to aggresiv

I run clawdbot since a month in a docker

I just want to ask how people uses it in this way, e.g. how to mount folders with access (on my NAS) etc.

and if people use connected second brain folders and how to give access

[mental mica]

thats not what you asked though, you asked if someone was running in a docker

https://dontasktoask.com

[ivory rapids]

Step 1 - find people running it in docker -> Step 2 - connect and discuss how to handle a bunch of usecases
Do you see issues in that?

[mental mica]

go read that link i sent

i see many many issues in that, yes

[ivory rapids]

ok fair, let me read your second link

fair enough, i guess i often use openers like above, seems like bad practice

[cunning seal]

I’m trying to set up clawdbot as a Docker container. The gateway works and I can chat using already existing skills, but when I try to create a new skill I get the error:

"I'm sorry, but I'm still unable to access the skill-creator tool. I continue to receive the "Tool skill_creator not found" error."

Any idea what I might have done wrong?

[uncut quiver]

Why use Clawdbot? Any Use cases? What are it's advantages over querying directly in Claude Code, ChatGPT, Cursor etc.? Just asking to wire my mind on how to properly use it.

[rocky orbit]

#showcase-old

[rocky orbit]

Memories, api, skills, channels…

[dark kayak]

I started here. https://www.youtube.com/watch?v=Qkqe-uRhQJE

[uncut quiver]

This video has convinced me to buy a mac mini for clawdbot 😛

[dark kayak]

lol

I just installed it in windows and I am going to install it on my mac tomorrow.

[unique rune]

Man I can’t seem to get my telegram bot setup 🙁

Created it in telegram but not linking to clawdcode

[stoic rain]

HEARTBEAT_OK 🦀

My guy has started actually messaging me this today each heartbeat. Yesterday he didn't... idk why.

Any ideas?

[modern thistle]

i cant do anythin with whith this tool just installed , everything is switched off in the dashboard

no whatsapp no agent i need some video guide to make use of this tool i think

[small smelt]

I added telegram... why do i get this..
Unsupported schema node. Use Raw mode.

In the accounts... it is not working

[modern thistle]

yes totally lost in conifgs

i think only ceo can use this tool

[stoic rain]

yeah... as amazing as clawdbot is, the UX is a nightmare lol, and constant things breaking, etc

it feels very much an early tech preview - which is fine - it's free, open source, can't complain

it has inspired me to want to build my own much easier to use version

[modern thistle]

complain reason it takes my time it is free but i lose my time for nothing yet

[dark kayak]

You guys need to run the clawdbot Doctor that is the fastest way I found to clean of the install

But then again I did install using anti gravity so the install was very quick and easy for me.

[hallow plume]

i have an old thinkpad x1, 32 gb ram, 11th Gen Intel(R) Core(TM) i5-1145G7 @ 2.60GHz (2.61 GHz) that I am not using. Installed linux on it to try it out last year, never used it since then. It might be the right time to give it a quick boot see if it is any good to run this bad boy

[dark kayak]

It'll run fine on that lappy...

[wispy spruce]

yeah cant link up telegram here unsupported schema node,use raw mode too!

[quasi wadi]

Hi all. What node manager should i use to install skills?

[gray light]

"Config was last written by a newer Clawdbot (2026.1.24-0); current version is 2026.1.21-2. How do I fix this version mismatch error?"

I am on the latest git checkout version

[dim olive]

My telegram bot is able to be read by clawd but it won’t respond though it. I’ve never used telegram before, any advice?

[twilit palm]

where do i put this claude setup-token ?

[low dragon]

Folks there are videos that explain setup. If you are so lost, I do worry about you understanding the risk of this, but it's everyone's choice.

https://docs.clawd.bot/start/showcase

[low dragon]

you ignore it.

[signal needle]

seems to be stuck here, any insights?

[low dragon]

https://docs.clawd.bot/help/faq#im-stuck-whats-the-fastest-way-to-get-unstuck

If you can't make it work with this then this project isn't for you - it's for hackers, it requires a bit of effort and computer knowledge.

[signal needle]

noted

[stuck matrix]

does anyone know where to find google antigravity api key

whats the simplest way to use gemini claude? i have the pro version

[valid drum]

what model is best for clawdbot

i got limited with claude instantly

[prisma lintel]

Question - are bug reports best here or as gh issues? I've hit a bug in the Matrix E2EE impl, all works but the trigger to request verification of other devices doesn't happen with this error in logs: {"0":"{\"module\":\"matrix-auto-reply\"}","1":{"error":"TypeError: client.crypto.requestOwnUserVerification is not a function"},"_meta":{"runtime":"node","runtimeVersion":"25.4.0","hostname":"unknown","name":"{\"module\":\"matrix-auto-reply\"}","parentNames":["clawdbot"],"date":"2026-01-25T09:58:31.120Z","logLevelId":2,"logLevelName":"DEBUG","path":{"fullFilePath":"file:///home/linuxbrew/.linuxbrew/lib/node_modules/clawdbot/dist/plugins/runtime/index.js:247:55","fileName":"index.js","fileNameWithLine":"index.js:247","fileColumn":"55","fileLine":"247","filePath":"/home/linuxbrew/.linuxbrew/lib/node_modules/clawdbot/dist/plugins/runtime/index.js","filePathWithLine":"/home/linuxbrew/.linuxbrew/lib/node_modules/clawdbot/dist/plugins/runtime/index.js:247","method":"Object.debug"}},"time":"2026-01-25T09:58:31.121Z"} - I can't see a ref to that method except in historical commits so I figure that got refactored but the call site didn't?

[viscid cove]

im try install on windows im noob . getting this error node.exe : npm error code ENOENT can someone help

[hollow canyon]

hello, on macos, here http://127.0.0.1:18789/config the scrool is not working at all

[gray light]

Anyone else experiencing pretty high usage for codex models (mainly openai-codex/gpt -5.2)? I just asked it to do three things and my usage went from 97 percent to 70 percent. I am on the latest git commit. Curious.

[prisma lintel]

I'll +1 the scrolling brokenness on /config, on Firefox/Windows (backend Linux)

[lone geode]

Hi all, Krill seems to hallucinate about memory features. Have an issue where my clawd isn't writing daily memory files. Is that normal? Per Krill should happen daily at midnight

[slender wasp]

gm!

how yo get your setup token?

[cerulean crow]

same here anyone can help

[uncut grove]

did you just install ?

[dim olive]

Yes, earlier today. I’m pretty experienced with self hosting AI and various other AI services but I’ve just never used telegram before

[hollow canyon]

did you solved that?

[prisma lintel]

Nope, but it is definitely broken!

[hollow canyon]

does this problem persits all over setups?

[uncut grove]

okay in telegram what steps have you already taken as far as following the docs?

[dim olive]

I’ve created the bot in botfather, given the api key to the onboarding, clawdbot created a script and ot can send messages to me, and can definitely see messages in telegram when I prompt it to check, but doesn’t respond automatically in telegram, messaging in telegram does not start/respond the chat

[uncut grove]

so in telegram after you did the steps in bot father, did you also already get to the part where you create your bot and you name your bot?

[dim olive]

Yes

[cunning scroll]

I'm attempting to create my own apple id for clawdbot to use imessage, however, that apple Id requires a phone number, does anyone bypass it? The message app fails entitlements on open without a verified phone number

[uncut grove]

when it came to entering your id during the onboarding, what numbers did you put in there? your phone? did you put like your phone number or did you get the correct user id from telegram?

[cunning scroll]

My current phone number is obviously tied to my apple id

[uncut grove]

okay everyone i see a common issue where people are putting their phone numbers as the id. it is not that

[dim olive]

I didn’t get that option to enter my ID I don’t think

[cunning scroll]

The id is the email

But the messages app won't work (and therefore imsg won't work) without being able to sign into it

[uncut grove]

to get your correct id in telegram at the bottom right, hit the search bar and type @userinfobot

[strange crow]

Hello

(no output)
connected | idle
agent main | session main (clawdbot-tui) |
anthropic/claude-sonnet-4-5-20250929 | tokens 0/200k (0%)

I am on the T.U.I, and I'm just not able to get this to work with my Clawd Max subscription. Does anybody have any tips to help me manage this?

[cunning scroll]

Oops. I think I messaged when yall were looking at something else. My bad

[dim olive]

cb did get me to do that, and I gave it the results, but nothing works still

[uncut grove]

okay then maybe the gateway is not running

Here you go:

Clawdbot Gateway Not Running - Quick Fix

1. Check status:clawdbot gateway status

2. If not running, start it:clawdbot gateway start

3. Check logs if it fails:clawdbot gateway logs

4. Common fixes: • Missing config: Run clawdbot init first
   • Port conflict: Another process on the same port
   • Bad token: Verify your bot token in ~/.clawdbot/config.yaml

5. Restart if stuck:clawdbot gateway restart

6. Verify it's working:clawdbot gateway status
   Should show running + uptime

Want me to tweak the format or add anything?

From my clawdbot lol 🦞

[floral citrus]

Maybe something is currently having issues, having the same on new installation right now, tried resetting a few times

[strange crow]

Yes I am struggling with this. Please let me know if you find a fix

[torn pelican]

can i hav quick confrmation reinstalling hackable over brew install wont wipe config right

[uncut grove]

Correct—config is safe. Lives in ~/.clawdbot/ which is separate from the install location. Switching from brew to npm (hackable) won't touch it. 👍

[hollow canyon]

this app is useless if the scroll is not working

[uncut grove]

wym?

[hollow canyon]

in the config page http://127.0.0.1:18789/config the scroll is not working at all, not on chrome not on safari .config-layout class have overflow: hidden

[covert thunder]

Hi, I hoped the cron job would just chat to me, but it doesn't seem to do anything?

[strange crow]

HTTP 401: authentication_error: Invalid bearer token (request_id: req_011CXU4sNFdtyELkeaem1MfE)

Getting this when I use my Claude Max subscription.

Has anybody figured this out?

[sonic snow]

How do I get this setup with grok?

[gleaming wing]

I think it's broken. I've been trying to fix this for a couple of hours now and just gave up.

[floral citrus]

Same here

[meager flume]

hi i am trying to use the minimax m2, and i got this error.. any help please?

[uncut grove]

That error (Cannot read properties of undefined (reading 'includes')) usually means a config field is missing or malformed.

Common culprits:

1. owners or allowed not set as array:# Wrong
   owners: 123456789

Right

owners:

• "123456789"

2. Missing channel config fields - Some required field is undefined

3. Model config issue - The minimax model might be returning something unexpected
   Quick debug steps:

4. Check ~/.clawdbot/config.yaml syntax

5. Run clawdbot gateway logs to see full stack trace

6. Validate YAML: clawdbot config validate (if available) or paste config into a YAML validator

[covert thunder]

Cron doesn't work for me either

[outer hazel]

did anyone use clawdbot with game development? like unity etc? or a stack like clawdbot with claude agents and making a game?

[little forge]

Hey i'm struggling to set up clawdbot. I keep getting stuck when trying to hatch. it says gateway connected however theres a message saying

HTTP 401 authentication_error: Invalid bearer token

I've tried many times to rerun claude setup token to get a new token but still get stuck in the same place.

Any help would be much appreciated

[outer hazel]

are you setting up with AWS or on your personal mac?

[torn pelican]

how do i run the gateway in the background none of the daemon or gateway other than gateway itself is working

[little forge]

Sorry should have added i'm running on AWS ec2 ubuntu

[slender wasp]

is it possible to use your gemini sub or is API the only way to use nano?

[toxic ravine]

same question i have actually

[steady flume]

If I had clawdbot do browser based tasks for me on TradingView under my account, or another trading website requiring authentication that tracks my trades, can’t I be flagged for violating TOS by scraping? Or am I missing something (both don’t have api)

[meager flume]

thx i managed to make it work, but now I am having same problem as here: #1464810660545433744 message

[untold veldt]

Why do i get this error ?

[covert crest]

Any ideas how to give clawd an actual cell phone number so it could make calls on my behalf using OpenAI’s voice models?

[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

[amber mortar]

why does claude not give me a damn setup token

[verbal plinth]

hello everyone.
i have a question,
i have a large number of files i want to upload on web, for everyfile there is a form to be fullfilled, like
-name
-date
-information
-ashtag...

i want every file to be fillformed.
can i do it?
i was thinking to create an excell file and upload it and say, ok" upload me this files"

is it possible?

[river badger]

setup token doesn't seem to work for me - so far only API keys have worked, at least for me

[amber mortar]

did you try in terminal or powershell

[river badger]

terminal

[amber mortar]

really weired

not gonna pay for claude code api when i have max subscription

[river badger]

yeah exactly, I literally just did it and it burned through my $20 API credits LMAO

but no point when I'm paying for max

[amber mortar]

Lmfao

[river badger]

hopefully anthropic opens it up so clawdbot can access the setup token correctly

[amber mortar]

Opus 4.5 burns through credit

like nothing else

[wooden raft]

Do set up copy token

Instead of setting it up the traditional way

[amber mortar]

100% it used to be open, but they closed it

[wooden raft]

I was having the same issue but doing the set up token copy made it use my max sub

[river badger]

really?

alright, gonna retry

[amber mortar]

so instead of setup-token,set up token copy ?

[wooden raft]

Yes

Like when it asked

You get like 4 options

[amber mortar]

i gotta sign out first right

[wooden raft]

Just make sure to have your code ready

[amber mortar]

cause it just loaded me into cc

[wooden raft]

Just do clawdbot config models then select and then config with your set up token but it says like copy or paste the set up token instead of the 0auth and new token

[river badger]

tried reconfiguring, still no luck - clawdbot doesn't respond in terminal or in the webUI

[wooden raft]

Spin up a clause code

And describe what’s going on

It will check through your files

Sometimes it’s as simple as the JSON is wrong

[river badger]

that's what I've been doing for the past 30 minutes before coming to the discord, didn't work but I'll give it another shot

[wooden raft]

And doctor?

[amber mortar]

in clewd code i think everything is right, but cc doesnt give me the auth token needed

ima try signing into CC on my wifes mac

for auth token

WSL, Powershell no luck

[wooden raft]

Even when you did that clause setup-token?

Claude

[amber mortar]

Yeah if i do that

[wooden raft]

In the new terminal ?

[amber mortar]

it just loads up reg. claude code for me

[wooden raft]

Hm try the other one

Like instead of the copy

And then just make note of the code

Or that doesn’t even work to?

[next glen]

Hey guys. I’m having issues with the toggle. I can’t switch to yes. Kinda stuck

[wooden raft]

Did you restart your terminal?

[amber mortar]

now it worked

had to add claude to path

than it worked. Thanks!

[fleet silo]

Not yet started with it, but could clawsbot be setup on a Linux and then act as a media assistent?

I send a message with title of movie or tv show. It puts that on my notion list and checks if it’s streaming on any of my subscribed platforms. If not it sends it to radarr or sonarr ?

I do this now with n8n just wondering if this is more stable

[river badger]

wait, what did you do exactly?

[wooden raft]

Glad you got it working!

[amber mortar]

is there any difference what Opus 4.5 i select?

[wooden raft]

Do current default id you want opus

[amber mortar]

i'm on windows, so i had to add the folder claude is installed in to the eviroment variables

[river badger]

ah okay, you're not through yet

let me know if you actually get it working though

[glossy rain]

chat pls help🙏

I typed the first installing commnd and it doesnt show wizard

it requires sudo access but I`m launching it from administrator profile

[astral garnet]

Use the dmg on releases

[next glen]

I restarted and re-installed. Should I tightened permission?

[meager mortar]

Yes

[untold veldt]

What is raw mode ?

[unreal bear]

How do i generate an oauth token that lasts longer than 3 hours

[amber mortar]

honestly my Clewdbot is pretty dumb

[unreal bear]

Fact

Lots of hype

[amber mortar]

loosing context after a few messages. i installed a gemini key, and it forgot that it can generate images with it

[unreal bear]

Well if it doesn’t forget then it will eat all your tokens

[astral garnet]

maybe you're not using it right

[rocky orbit]

Which model?

[amber mortar]

Opus 4.5

[astral garnet]

context is an anthropic problem

[amber mortar]

right

which model is better to use?

[rocky orbit]

Save stuff in your memory files

[astral garnet]

#old-models

[amber mortar]

also is it possible to use a higher context model, and have it switch to opus for coding work?

[astral garnet]

#old-models and yes, use coding-agent from clawdhub.com

this ecosystem isnt built on coding as the front end. its more of an assistant that calls tools for coding.

[amber mortar]

Right

Thanks will look into that

[astral garnet]

the real benefits are its always on and listening, and can run cron jobs, webhooks, etc. so it actually is alive to intake data without launching "echo 'test' | claude -p"

[hasty charm]

brand new here so please let me know if there's a better channel for this. what are folks finding is the best solution for getting past captchas with ClawdBot?

[amber mortar]

Maybe if you get it to build a cap solver integration

[tawdry robin]

Does anyone have experience running multiple sandboxed clawdbots?

[astral garnet]

peter does but i dont think most of the users here have gone that deep yet.

[amber mortar]

huh

my clawd does not respond anymore it seems

[sacred thunder]

how to connect clawdbot to github?

[astral garnet]

#1459642797895319552

[amber mortar]

so i think i exited clawdbot. how do i restart it?

[elder portal]

Hi, i need help with ollama, qwen/local. Model is running (I can query it from ollama run..) however clawdbot shows missing.

[astral garnet]

these are basic commands and help that #1459642797895319552 as well as most llm's can walk you through

[indigo cave]

I hope someone can help me out, when i add a new agent to following this wizard https://docs.clawd.bot/start/wizard how do i open up the dashboard for this agent? to set it up with telegram and pairing code

[astral garnet]

anyone using webhooks for clawd yet?

[weary plank]

hello so i want to install clawd but i cannot decide on which provider should i use so now i have a yearly gemini ai pro subscription but if there are cheap alternatives i can also buy them what would you recommend for me?

[astral garnet]

whatever is free, see how many wrong answers you get then try another.

[hollow canyon]

isnt this suppose to be safer with local llm?

[lament socket]

any idea what im doing wrong for the ui to not open correctly
ui opens with
"This site can’t be reached
127.0.0.1 refused to connect."

[astral garnet]

make sure you have your token set

[lament socket]

|
— Onboarding complete. Dashboard opened with your token; keep that tab to control Clawdbot.

tried reinstalling from the beginning
same thing

[lusty mulch]

I see error [tools] exec failed: spawn EBADF when its trying to run most bash commands

[granite onyx]

Anyone else facing this ?

⚠️ Agent failed before reply: All models failed (3): openai-codex/gpt-5.2: No available auth profile for openai-codex (all in cooldown or unavailable). (rate_limit) | anthropic/claude-haiku-4-5: No available auth profile for anthropic (all in cooldown or unavailable). (rate_limit) | anthropic/claude-sonnet-4-5: No available auth profile for anthropic (all in cooldown or unavailable). (rate_limit).
Logs: clawdbot logs --follow

[lament socket]

maaayyybeeee this is my issue too 👀

[hallow remnant]

What are the system requirements? I’ve noticed that a lot of people run it on a „Mac mini“ is there a specific reason for that?

[mental mica]

Shiny toy syndrome (and iMessage support)

[zealous raptor]

you don't need mac mini

[hallow remnant]

Yea i know, but why so many people use it…

[hallow remnant]

Okay make sense^^

[lone tangle]

Hi I have a Google Subcription (for Gemini and Antigravity etc), and I have a z.ai subscription, can I use one of those LLM providers? Or is it recommended to use Claude Pro?

[hallow remnant]

Okay so no iMessage support, does it have any other limitations?

[sour slate]

@midnight notch, any discussion about various AI models should be taken to #old-models.

[hallow remnant]

And what are the recommend requirements to run it?

[silver imp]

You should be able to create a new session on a regular mac to get iMessage support

[silver imp]

The cheapest possible if you're just running the assistant. A $5 machine will do

But if you want to actually code / run programs on the machine then scale accordingly

[hallow remnant]

Yes, this is the point, wich requirements a recommend 😂

[silver imp]

going to sound silly, but ask an LLM. It totally depends what kind of programming you're planning to do

Most SaaS / Simple Websites don't require much at all

If you want to train an LLM, you may want a $5000 / month rig

But, assuming you're asking the question and have no idea what you want to build, go with the cheapest option

And then - if for whatever reason that's too slow, upgrade

It will be cheaper and better that way

[hallow remnant]

Yes of course i understand it, and i know that it depends. So I don’t want to train an LLM, like for normal stuffs. But softwares/backends with „harder“ backends and not „simple websites“.

Is for this 16gb ram enough? Or does i need more? So exists maybe any list for comparison? -_-

[silver imp]

small machines can do amazing things if you optimize

16GB of ram is plenty

most people have that on their laptops

and they run much more complicated software than Postgres DBs

[hallow remnant]

Okay thanks, perfect. 🙂

[astral garnet]

Pay for Anthropic max and you’re good to go.

Everyone spending 100s a month on television and cell phones but want ai for free

[misty gale]

Anyone know if it would be possible to use Venice AI Pro (oauth) instead of API?

[rotund dragon]

Does anyone know if it is possible to host Clawdbot on digitalocean app platform? it looks like it will build (im testing it rn) since it has a docker file. However I am assuming the setup will be totally messed up probably unless someone took the time to actually make clawdbot specifically able to be setup in such a configuration

ahhh okay yeah for example it doesnt support health checks so DO rejects it . I think what it would need is 1) mandatory 'secret' provided in ENV so only the actual deployer can access (since endpoint is on the web for everyone) 2) endpoints to be protected by this secret 3) a healthcheck endpoint that DO can ping

but needless to say this would be pretty awesome

[sharp pond]

Noob here!

After restarting the gateway
Telegram stopped responding
clawdbot channels status
Shows

Gateway reachable.

• Telegram default: enabled, configured, stopped, mode:polling, token:config

stopped doesn’t sound good
How do i troubleshoot this

[lusty mulch]

found the issue, had a skill loaded with a massive directory

[shadow oxide]

What's the best model to run Clawdbot ? Opus 4.5 keep having the context window compacted super fast. It keep forgetting everything literally all the time

[devout nacelle]

im getting this: HTTP 401 authentication_error: Invalid bearer token (request_id: req_011CXUPVVjSzeoinrY5oM7VP)

when connecting telegram, can anyone help?

[dusky monolith]

how do we start an actual talking session in the cli

[flint mica]

You don’t use clawd in the cli, you integrate it with discord, iMessage, telegram, WhatsApp, etc

[dusky monolith]

But when I ran the onboarding I was able to use it in the cli

[brisk valley]

Im using glm 4.7, seems a bit slow but reliable and cheap

[dusky monolith]

figured it out, you run tui

[astral garnet]

clawdbot gateway --help is a start

[torn nexus]

Anyone see this issue when (randomly?) the only answer I get is "An unkown error ocurred"

[stiff girder]

jonah can u help me out?

[sturdy spoke]

Hi, I'm getting this message LLM request rejected: messages.69.content.2.tool_use.input: Field required. Anyone ever got this? What's wrong with my lovely crab!!

[stiff girder]

brew update && brew upgrade clawdbot i did this command on terminal but it did not update my UI @flint mica

[coral yoke]

I have a weird issue with the UI, but i dont know if its because im screen sharing my Mac Mini and controlling it remotely.

Basically the config pages cannot scroll down. I have tried both native Screen Sharing on Macs, and RustDesk, but on both, I can't scroll down on the pages.

Scrolling works on every other webpage, so not sure why this is happening, is it a known issue or is it because I am screen sharing. Haven't managed to go use the Mac Mini directly with proper mouse & keyboard yet so not sure.

[spice lantern]

If you want to get the most out of Clawdbot you will want to be able to ssh into the box, play around with the configuration, tweak the settings and experiment. DO app platform feels like the wrong fit.

[hollow canyon]

install Stylus chrome extension and force this new style @-moz-document url-prefix("http://127.0.0.1:18789/config") {
/* Insert code here... */
}.config-layout,
.config-layout * {
overflow: visible !important;
max-height: none !important;
}

[astral garnet]

Update to the latest version it fixes that

The new version changed the webui a lot

I didn’t realize it and was on 1.23. Now 1.26 is a lot cleaner

[bold cosmos]

Updated to the latest version and it stops to work. Try to restart, it works for 15 mins and then shuts down. Anyone else with the same problem? On iMessage

[lone tangle]

"disconnected (1008): pairing required"

I am trying to access the Web UI via Tailscale Serve. I see the webpage but cant do anything because of this error. How to pair?

[kind hatch]

how often does clawdbot save to memory? or are you guys asking to update memory often?

[hallow remnant]

Which model are you using, or does it depend on the general use case of Clawdbot? Alternatively, which models are generally available?

[sour slate]

@hallow remnant, any discussion about various AI models should be taken to #old-models.

[graceful flax]

Anyone have a fix for macOS app build error? I can't seem to compile it. I get these errors:

error: external macro implementation type 'SwiftUIMacros.EntryMacro' could not be found for macro 'Entry()'; plugin for module 'SwiftUIMacros' not found
68 |
69 | extension EnvironmentValues {
70 | @Entry var mathFont = Math.Font(name: .latinModern, size: 20)
| `- error: external macro implementation type 'SwiftUIMacros.EntryMacro' could not be found for macro 'Entry()'; plugin for module 'SwiftUIMacros' not found
71 | }
72 |

[meager mortar]

Clawd is feeling to predictable. Little is done proactively, just on a cron schedule. Does anyone have tips to make it more able to prioritize and push things? Like, oh he likes topics x, y, and z but there is nothing going on with y, but z has this interesting story on X

[rocky orbit]

Soul memory and agents, talk to it a lot and give him max info

It’s an ongoing onboarding with your bot

[severe rampart]

Anyone using the free tier Gemini-3-Flash API for Clawdbot? Could it be comparable to or better than Minimax M2.1?

[lone tangle]

Where can I manage my Model providers and what model it should use?

[lunar holly]

npx clawdbot models

[meager mortar]

i do -- but do you ever get something INITIATED by clawd? Also how use Agents for not code related things, whats a couple examples?

[rocky orbit]

It will not initiate unless you ask the trigger is your message or cron

[meager mortar]

right that is my point

[rocky orbit]

For example it reads my emails via cron and suggest actions based on that

[meager mortar]

i want it scanning the horizon (news, socials that i pick) and prioritizing

yes, i have not set up email yet (just my choice not to). I am thinking about giving it is own email and i forward select things in -- as baby steps for email

[delicate flint]

those of you using gog, mine has a keyring passphrase protection, how do you handle it?

you can just export the passphrase and call it a day but I feel like that would be the dirty solution lol

[barren frost]

How can I interact with that bot?

[delicate flint]

which bot? clawd?

[barren frost]

Yeah, new here sorry

[delicate flint]

well, you need to install and set it up first... https://docs.clawd.bot/start/getting-started

[prisma isle]

Anyone see this a lot? Is it “normal”?

[sudden tide]

I've been unable to use my claude code token, anyone else have trouble with this?

anthropic:default: cooldown (22m) — Wait for cooldown or switch provider

this is what i see when running clawdbot doctor

[signal dawn]

Will I need a Mac Mini to use with iMessage?

[mortal tangle]

not nessesarly, read about BlueBubbles

[twin depot]

in trying to set up clawdbot i started with claude opus 4.5 with an API key and pro subscription - is it better to use an o auth setup instead of api key?

it seems i can't even ask a simple question like "are you here" once per minute on TG.

is the answer just to upgrade subscriptions? is it more efficient to use o auth with subscription instead of API key?

[mental mica]

Far better, yes

[twin depot]

i tried switching to gemini 3 preview where google gave me $300 in credits, but something broke and clawdbot wouldn't give any output.

[mental mica]

Gemini sucks for clawdbot

#old-models

[twin depot]

thanks i'll check in there

and the only way i've been able to try to change models so far is by re-running the setup wizard

is there any other way or is it necessary to go through the wizard every time

[sudden tide]

im having trouble getting my anthropic oauth to work, i keep getting a 401 error, anyone else have any luck with that?

[red hatch]

Hello, I am seeing this issue where clawdbot's responses do not show up live in tui or from Telegram. When I type a message in TUI, I get a (no output), however when I restart TUI, I see the response. In Telegram, I can see that clawdbot is typing something, but it never shows up.

[prisma isle]

This happens if an agent crashes and fails to release a lock from the current session. You can go to a different channel, e.g. if you’re talking on the gateway, switch to telegram, and ask it to scan for stale session locks and clear them,

[prisma isle]

Use the web hi directly, if the chat there is working ask it where your messages from tui or telegram are going, you might have messed with the routing config

[prisma isle]

Yes it accesses the currently logged in iCloud acct

[red hatch]

I am running on a VPS. Do I need to setup a tunnel to load the web ui ?

[prisma isle]

Well any channel that’s responsive like switch to discord. But yeah web ui only binds to localhost

[valid drum]

what model is best for clawdbot that doesnt cost like a thousand dollars

[atomic sparrow]

Hello, my claude token wasn't setup properly in the assisted onboarding - how can I fix it now?

[signal dawn]

Thank you.
I will give Clawd it’s own iCloud account.

Last question does Clawd need to be admin on the Mac?

[prisma isle]

No but mine is because YOLO

[red hatch]

I dont think its the channel issue, isnt TUI or Telegram responsive? Those two dont work for me, as in , the message clearly goes through to the bot but I dont hear back synchronously

Because clawdbot actually did respond however it just never got communicated back into the channel for some reason

[white hinge]

how come my gateway keeps getting disconnected (then it reconnects after like 20 seconds)

[atomic sparrow]

Hello, my claude token wasn't setup properly in the assisted onboarding - how can I fix it now?

[brisk valley]

run clawdbot configure > local > model

[amber tartan]

Anyone else having serious issues getting basic remiders to work with CRON? I am NOT a programmer, but I'm using chatGPT ina. separate window to walk me through setup and configuration and no matter how hard I try, I can't get CRON to work. Doesn' thave to be a big discussion here, someone can DM if they're having same issues or have a simple fix.

I should mention my primary model is Gemini 2.5 Flash - purely because I already pay for Gemini Pro. I'm open to- and considering - downloading an LLM locally on my Mac Mini M2 using Ollama....

[brisk valley]

i tried to ask my bot to set a cron, ill let you know if it works

[brisk valley]

quick q, for gemini you use api? Because i also have pro but i thought it's outside of the plan with the apis

[indigo cave]

Is there a easy way to make sure agents can see other agents messages in groupchats on telegram

[rocky orbit]

Agent?

[simple violet]

Question, when setting up on a dedicated MacMini are yall setting up the mac with you're personal icloud logged in to the machine or creating another account? Im a little hesitant on logging in with my personal account.

[indigo cave]

i got a groupchat setup with mulitiple agents setting. They can't see there messages just my own

[rocky orbit]

It’s in the bot settings in Telegram

[tender cosmos]

ok i need help.. been hours..

I’m running Clawdbot on a Raspberry Pi. Gateway is healthy and reachable locally.

Gateway bind: loopback (127.0.0.1:18789)
Accessed via Tailscale Serve at https://dashboard.<tailnet>.ts.net
gateway.trustedProxies set to include 127.0.0.1, ::1, and my Tailscale client IP
Local pairing via SSH tunnel (http://localhost:18789) works

Tailscale Serve URL never pairs

Symptoms

Control UI loads, but WebSocket closes immediately

Repeated logs:

cause=pairing-required
reason=not-paired
code=1008
forwardedFor=100.x.x.x
remote=127.0.0.1

Key observation
Pairing is origin-scoped. localhost pairs successfully, but https://dashboard.<tailnet>.ts.net can never complete first-time pairing. Tokens do not bypass this.

[brisk valley]

It might be related to this

[brisk valley]

have you tried asking claude code what's wrong?
And tbh, personally I'd just start over following a step by step guide that you can find on yt

[warm oriole]

Hi -- if I were to set this up using a vps (aws) would I have an interface besides discord to communicate with the bot?

kind of seems like it would be best for it run on its own hardware so you can actually view it working

are there any videos of people actually using it? I have seen a bunch of set up videos, but none show actually interfacing with it

[brisk valley]

you should use that to communicate with it, you also have a nice dashboard where you can check the logs
You will be able to interact with it trough ssh

[warm oriole]

ok awesome -- but is it true that if you set up its own device (mac mini or somthing) you would actually see it controlling the computer?

or is that not the case

ty btw

[brisk valley]

docs (dot) clawd (dot) bot /start/showcase for some video tutorials

[brisk valley]

I'm not sure as that's not my config

[wild niche]

My fallback is not working: Error: No available auth profile for anthropic (all in cooldown or unavailable). What should I do?? when I set openrouter as my primary model, it works fine, but fallback fails

[prisma isle]

yes personal. it has already alienated my father and gotten my sister pregnant

[mint quarry]

Okay, I'm sure this has been asked a million times. Anthropic Max plan usage is currently not possible, right? Only API

[final solar]

Facing same issue. CRON doesnt work. i figured out the model is failing to respond becaused of malformed function call. in our case CRON schema sent by the clawdbot is not detailed enough for gemini to understand how to populate the data for cron.
Output
{
"candidates": [
{
"content": {
"parts": [
{
"text": "Defining the Briefing Structure\n\nI'm currently structuring a daily 7 AM briefing. It will encompass trending world news and AI news. The goal is to build a recurring automated process. I'm exploring the best way to leverage cron for scheduling. This will enable consistent delivery of the briefing.\n\n\n",
"thought": true
}
]
},
** "finishReason": "MALFORMED_FUNCTION_CALL",
"finishMessage": "Malformed function call: "**
}
],
"usageMetadata": {
"promptTokenCount": 12345,
"totalTokenCount": 12345,
"promptTokensDetails": [
{
"modality": "TEXT",
"tokenCount": 12345
}
]
},

[twilit spire]

Anyone have an issue connecting using open ai models via their api? I get this regardless of the selected model: 20:10:54 warn agent/embedded {"subsystem":"agent/embedded"} unsupported thinking level for openai/gpt-5.2-chat-latest; retrying with medium .. doesn't seem to matter what I select for a thinking model. Just get (no output

[alpine stream]

Hey all - does anyone know how I can use an Azure openai endpoint? the docs doesnt seem to tell me how 🙁 i posted in the "models" channel but i think its not right, sorry :/

[desert ether]

Hey! I’m on Claude max, tried using the oauth token but I thjnk anyhropic patched it earlier this month… so now im using api key from an org (just me in it). But I keep getting rate limited on the 30k tokens/min limit- any advice?

[alpine stream]

Austin: you are probably on the wrong tier?

[errant ferry]

Reach out to anthropic to increase your rate limit. Also take a look at openrouter if anthropic take too long to respond

[errant ferry]

Do you have a vibe coding setup? If so pull the repo and ask opencodel/cursor/Claude code to investigate then build it and submit PR! 👨🏿‍🍳😜

[unique rune]

Hi, can someone please help. It looks like I have connected to my Telegram bot but I am trying to talk to it and it isn't working

[alpine stream]

Hah i like your idea! 🙂 i might consider - I just thought/hoped clawdbot already supported it, knowing it supports openai already 😄

[alpine stream]

Read what it says 😄 you already have a gateway running. stop it, and try again. perhaps try restarting if youre uncertain about it.

[unique rune]

I am not sure what that means man sorry, how do I stop a gateway?

[desert ether]

I see - what is your rate limit per min? Also do you use openrouter?

[alpine stream]

https://docs.clawd.bot/gateway#gateway-service-management-cli

[errant ferry]

Well I'm on the highest one as been on their APi since it launched, but it's based on spend too. To avoid all that pain use openrouter, but what your use case?

[river iron]

Ey anyone tried using clawd with nvidia nim platform?

[desert ether]

I would ask it to just read through my iMessage and reply to people and couldn’t do it

[next glen]

How can I access this dashboard while I’m using AWS? The website doesn’t open

[dusky loom]

Anyone else having issues trying to start new sessions? /new just gets ignored/deleted.

[light wraith]

I have been have a issue connecting with a VPS the connecting to my windows PC. When it does connect then I can't talk to the assistant on the dashboard or on telegram. I have a Claude max subscription so I have opened the clawdbot folder with Claude CLI and let Claude trouble shoot. But I'm unable to get everything to connect and the assistant to work. I'm only able to connect one or the other either the assistant works I'm able to communicate via the VPS or I can't talk to the assistant but everything is connected according to Claude and play write mCP tests.

[unique rune]

Claude: HTTP 403: OAuth token does not meet scope requirement user:profile. Does this mean my API for ANthropic isn't working?

[pulsar marten]

Hi everyone I've set up Clawbot on a VPS (AWS EC2) but I just can't get it to connect to my Claude Pro account (using an 0Auth token). Has anyone else had the same problem?

[unique rune]

I am having this exact issue now I think too

[pulsar marten]

I asked Claude and got this: 'I found the issue. This is a recent change by Anthropic. Anthropic has changed permissions on Claude Code OAuth tokens. They are now restricted to work only within Claude Code itself and cannot be used for external API calls. GitHub
Subscription OAuth tokens no longer work outside Claude Code. Webcoda Pty Ltd This happened around January 9, 2026.
Unfortunately, this means you cannot use your Claude Pro subscription with Clawdbot anymore. Anthropic blocked this loophole recently.'
Is this true? Are there any workarounds?

[unique rune]

oh what seriously

surely not?

[pulsar marten]

So annoying. I feel I was so close to setting up.

[lucid widget]

I was chatting with Clawd and then I got "gateway disconnected: service restart | idle" message. I tried many commands from docs and #1459642797895319552 channel but it's the same message "gateway disconnected: closed | idle"

FIXED THIS
I restarted server and run doctor commands

[unique rune]

Yeah same, surely this isn't the case though..?

[pulsar marten]

Hopefully someone here can help us

[unique rune]

I hope so

heaps of people would be using claude

[pulsar marten]

Claude's suggesting I use an APi key and add credits. I don't want to do that as I already pay for Claude Pro and don't want to shell out more.

[astral garnet]

I don’t think that’s accurate

[high wadi]

Hi guys Im just reinstalled mac os on an m1 macbook air and trying to get clawdbot working

[astral garnet]

I only have Claude code and I’m still working.

[high wadi]

it seems very unstable always crashing and not responding to telegram and I have to manually restart my laptop to get it to respond

[pulsar marten]

Are you runing it on a laptop/desktop or on a VPS?

[astral garnet]

Vm

With Claude code oauth

[pulsar marten]

Oh wow. I can't work out what I'm doing wrong then.

[pulsar marten]

AWS?

[neat swallow]

Same problem - found any solution?

[unique rune]

I am doing the same thing

Just followed the quick start guide and not working

[warm oriole]

oath is against claudes ToS

maybe thats it?

idk though I am noob

[brisk valley]

If you configured tailscale, you can just go to your tailscale ip:18789
Otherwise something easier if you're connected with SSH in vscode
You can run clawdbot status it will tell you your dashboard link, you can ctrl + click to open the link in the browser and vscode will automatically tunnel the connection

[dusky loom]

I used the token today on the installation

[unique rune]

Then how do you connect it? Need the API?

Then you gotta pay?

[lone tangle]

Hi yes I think I disabled the need for pairing (for the web ui) in the config file

when i am home i can look up what it was exactly

[paper zealot]

I have the same issue with Claude OAuth on VPS.

[brisk valley]

I agree it seems to be struggling with CRON, if you find a solution please update

[kind hatch]

the dashboard on chrome just doesnt work anymore, it only works on safari for me now so weird

@sharp parcel can u help me

[main hemlock]

did you redo your installation? maybe need to delete old token from chrome

[wind seal]

i installed tailscale and then, clawdbot says that it will checkit out, and then: gateway disconnected: closed

anyone knows how to get the gateway going?

[atomic sparrow]

Is there any workaround to not look like I'm talking to myself in Whatsapp? Because I've connected my account, it looks like I'm talking to myself (because it's actually a chat conversation with me/my number). Anyone figured how to solve this?

[full bolt]

Building Gemini API version of ClawdBot. Hmu if you wanna brainstorm features or extensions. I got the core architecture and telegram 2-way communication setup.

[nova oracle]

Is the railway template working? Which template should I use? Both didn't work for me

[thorn iris]

I just keep getting this error. I have made sure I’m no where near my usage. I have try reconfiguring my api and restarting my gateway and still nothing. Any help would be great. (It might be something easy. I’m very new)

HTTP 401: authentication_error: Invalid bearer token (request_id: req_011CXUpvt4EBGuExV32Y14GM)

[paper zealot]

Getting this same error.. assume you're using Claude Oauth?

[thorn iris]

Yes

Am I doing something wrong or do you think that the system is just down?

[paper zealot]

I know a few people have had the same issue. Heard that Anthropic blocked it but have had no confirmation either way. So waiting for someone with a bit more of an idea of the issue to weigh in

[thorn iris]

Okie thank you!!

[spring trellis]

What is getUpdates and why are calls to it failing with a 401 when I do clawdbot gateway?

Seems like something with telegram but I cannot figure out what.

[raven atlas]

can someone please help me setup the 1password integration i cannot get it to work I am so fucking confused and frustrted

[knotty umbra]

can anyone help me out? i am stuck on here when I try to set up

[raven atlas]

When this happened to me it was my API key that wasnt working because my billing wasnt setup

[west yarrow]

trying to follow the guide here for setup on railway, but whenever i use the one click deploy, nothing happens?

https://docs.clawd.bot/railway

[spring trellis]

I can't figure out why I am hitting 401s

[atomic kelp]

im on Windows and using Clawdbot to help code some incredibly challending investment thesis and dashboard projects, its costing me about $450 per day in token usage, HMU if you feel like laughing at me and sharing my pain!

[faint nexus]

Any idea why when clawd launches a chrome instance to browse its super slow/feels like limited memory or something compared to me manually launching a chrome to browse web?

[zinc citrus]

I completed the setup and it says onboarding complete but my Google browser says “This site can’t be reached” … refused to connect

[rocky orbit]

It’s an agent controlling your browser so it’s slow yes

[faint nexus]

It’s the resources etc loading that are slow like it’s throttled request or something compared to normal

[atomic kelp]

yeah, it’s running a constrained, instrumented Chrome, not your normal one.

[faint nexus]

Thanks yes that’s the feeling - constrained, make sense. I can’t get it to be faster though so maybe just accept it loads some heavy dependent assset pages so slow?

[paper pilot]

how do i have clawd "think" in my same timezone?

[atomic kelp]

make it do less. try these fixes: hit APIs or text endpoints instead of full pages whenever possible, block heavy resources (images, fonts, video, ads) in the automation context, reuse a persistent browser context so cache/DNS aren’t cold every run, avoid JS-heavy sites unless unavoidable (React/Next apps are worst offenders), or just accept first-load slowness as the cost of determinism

[deft storm]

Jace, your social media analyst sub-agent is still facing trouble connecting to the browser. It seems that despite our previous restarts, the browser control server remains unreachable. My apologies for this persistent issue, but we'll need to try restarting the Clawdbot Gateway again to get it working. ... over and over. ugh

[halcyon jolt]

hi, im having the following issue (npm error code ENOENT), can anyone help? things i have tried: running in admin mode, installed latest npm

[zealous raptor]

AGI can't handle timezones

[faint nexus]

Thanks @atomic kelp yes the js heavy thing react/nextjs app is it really slow with. Its also on subsequential requests for the same domain that’s why its so weird and feels like throttling

[sour slate]

@halcyon jolt, 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

[lucid mulch]

Hi, what do people use open router for with clawd?

[zealous raptor]

windows is finiky, use wsl for the best experience,

[lament heart]

What messaging frontend most people are using to interact with ClawdBot?

[halcyon jolt]

just tried to install wsl via powershell & its given me error 14098 'the component store has been corrupted' .. not having much luck :')

[spare sun]

Hi, anyone able to bypass Office 365 to access microsoft apps with ClawdBot?

[zealous raptor]

don't give up, you got this

[twilit bloom]

just booted up clawdbot and have telegram connected and running on AWS. I want to connect it to my personal gdrive to help with email triage (read-access only).

For security purposes would be be better to setup a new gmail and have my primary gmail forward to the new address (the one I'd give my clawdbot access to)?

How are other folks managing?

EDIT: I'm non technical so worried I might do something stupid and allow someone to get access to all my stuff (password resets, etc)

[astral garnet]

Read only is pretty safe

[solemn kestrel]

my clawdbot is set up in ec2 aws , not remote. i just see the ubuntu terminal in my browser. i want to see the dashboard on my main mac im on right now . what should i do ? i wanna see the visual dashboard thing like this

Krill not responding on help chats for atm, and im eager to get started !!

[astral garnet]

Alex turned his Mac back in

[jolly sage]

Do I need to pay for Claude to get the API token required for Clawbot?

[topaz pike]

yes lol

[jagged abyss]

theres a microsoft cli in clawd hub if your not using it already

[topaz pike]

anyone tweak the discord code so that each channel is a new session / self-contained? i keep seeing conversations and forgetfulness leaking between "sessions" and don't know if they are really sessions or not. any help?

[astral garnet]

Kinda common. I’m using webhooks and see the same

[deft storm]

not being able to search the internet for results from my CB due to: "ou are absolutely right. This is crazy, and I completely understand your frustration. My sincerest apologies for the roadblocks we've hit trying to get automated data collection from social media. It's clear that the standard web scraping tools (browser and web_fetch) are consistently failing due to the nature of your EC2 environment's browser setup and social media platforms' anti-scraping measures."

[jolly sage]

pro ?

[deft storm]

wild

[marsh charm]

Anyone successfully got xAI configured? I've not seen Krill be helpful with that yet.

[knotty umbra]

where do i run the "claude setup-toke" command to get the token for claude

[fleet totem]

anyone using openrouter / gpt-5.2-codex-xhigh as session model? seems like it broke my messaging

[stoic rain]

anyone know how to do this? i made a new claude acc just for clawdbot and want to switch to it - i re-ran onboarding, chose CLI again, but it skipped over it since its already configured w/ my old claude acc ... how to get it to reconfigure so i can use my new one?

[short willow]

Anyone having issues with whatsapp business for Clawdbot? in Brazil we have a extra 9 before our number and after area code, for some reason the link ignores that.

[rare wave]

accidently selected the wrong model when doing QuickSetup. Terminal will not let me change it. Is there an easy way to change it? or Should I just restart the QuickSetup process?

[sturdy flame]

any other VPS yall are using to get clawd up and running? AWS isn't working for me

[normal seal]

Can I realisticly run ClawdBot on a $20 monthly OpenAI or Antrhopic subscription or do I need to use a higher tier monthly or API?

[dire birch]

Hey there, I'd like to use https://docs.libertai.io/apis/text/usage.html as a provider, it has OpenAI compatible endpoints. Can anyone guide me please ?
#1465121377278230702

[strong phoenix]

Loving this so far! Can anyone point me to the android app and where do download it?

[paper pilot]

@sharp parcel where does memorySearch.provider go in the config for the agent?

(please let this work)

[cedar parcel]

did anyone have an experience using Google Gemini instead of Claude or open AI?

[formal fiber]

So I have a number of chron jobs that the agent apparently has scheduled. About 20 minutes ago there was supposed to be an end of day summary. Agent says the job fired but they didn't send anything back. Do I have to put chron job info in heartbeat.md? So prompt it to put job info there?

[brisk valley]

Pro you tip send the docs link to Claude / chatgpt chat and ask away

[odd patio]

I've been attempting to get a headless browser working with Clawdbot on my Mac. I just did a search and most of the results are related to raspberry pi. Has anyone got it working? No matter how I set the JSON file it still launches the browser window. I've tried chrome and I've tried brave and it's the same result.

[silk veldt]

Venice API - is anyone successfully using this? My gateway is idle and I just can't seem to get any reponse from Venice.

[merry crater]

so strange I can't seem to get my claude max setup with this

got everything setup but when i got to chat it doesn't work lol

[floral citrus]

My issue was the terminal being too narrow it didn’t fetch the token correctly 🤷‍♂️.

[merry crater]

wow, that was it

[rose wagon]

I’ve got an old MacBook Air I’m installing on. What’s the best safety practices on setting it up to control that computer? Create a separate iCloud account or use yours with limitations?

[odd swallow]

What local models are people having success with for tool calling etc.?

[grizzled steeple]

with a 512GB Mac Studio, is there any way to speed up the minimax-m2.1-8bit-gs32? i find semi-complex prompts can sometimes take 2-3 minutes to complete

[grizzled steeple]

minimax-m2.1-8bit-gs32 works phenomenally well

just a bit slow

[shell summit]

Anyway is able to run locally? Ive been trying for last few hours and been struggling

Im trying to use ollama and qwen2.5:14b-instruct-q4_K_M

I can run it on ollama but connecting clawdbot doesnt work sadly

[digital ridge]

im trying to set up clawdbot on my old omen PC i wiped it and have the required stuff I want it to acess installed. Im new to developer and AI stuff. would anyone be willing to help me set up the agent?

[torpid jay]

[clawdbot] You have hit your ChatGPT usage limit (free plan).

but im on a paid chatgpt account... (plus). please help!

[dusky loom]

anyone find a fix for Profile anthropic:default hit Cloud Code Assist format error. Tool calls will be sanitized on retry.

[lone tangle]

Hi I am a bit confused by the structure of the chat context in Clawdbot.

I don’t understand how the context is managed between sessions, is a new session like a new chat with no context of previous conversations?

[raven atlas]

guys holy shit, i cant connect anything even google it straight up doesnt work ive been torubleshooting for hoursand cant even get google to work

[woven steppe]

whats the best browser use skill folks have used with clawdbot? i think it has the peekaboo built in but are there other better ones? heard good thing about vercels but more in regards to having claude code use it. unsure with clawdbot

[glossy kernel]

bro is anyone facing the memory-core installation error?

[torpid jay]

how do yall add another gpt ersion

[near nymph]

Hello everyone, clawdbot has stopped to answer me in Telegram, the status is ok, what is the cause? is there something that i can check? Thanks

Update: it continues to send me the same last answer

[dark kayak]

Yes, you can coordinate two distinct Clawdbots (and even other agents) if they are properly networked. 🌚

The standard way Clawdbot handles this is via Node Pairing or Gateway-to-Gateway communication.

If you have:

Clawdbot A (Windows / Trading Rig)
Clawdbot B (Mac / AI Engine)
We can pair them so that Clawdbot A can ask Clawdbot B to do things (like "Run this optimization script" or "Read this log file").

How to check if they are already visible to each other:

On this machine (Windows), run: clawdbot nodes list (or I can run it for you).
If the Mac isn't there, you generate a pairing token on the Mac (clawdbot nodes pair) and paste it here.
Once paired, I (on Windows) can use the nodes tool to send commands directly to the Mac's shell or filesystem. It's much cleaner than relying on a shared S: drive for everything.

Want me to check if any nodes are currently paired? 🌚

[jolly sage]

Hi! Is there a setup guide for Z.ai GLM-4.7 with Clawd.bot? Thanks!

[ornate tinsel]

are many of the skills basically useless on windows :(

[sturdy flame]

Anyone know how much server space you need to get for using a vps?

[umbral yew]

Clawdbot is giving me this error:

LLM request rejected: input length and max_tokens exceed context limit: 171004 + 34048 > 200000, decrease input length or max_tokens and try again

How can I fix it? I’m on free AWS server

[dusky loom]

this is a black hole for the questions guys

[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

[mental mica]

jump in and help then, we're absolutely swamped today

[zealous raptor]

Reduce your input length it says so in the error message

[timid fractal]

claude is very expensive and i was looking at some of the chinese models to start using. has anyone got experience using them?

[sour slate]

Any discussion about various AI models should be taken to #old-models.

[ivory chasm]

What's up with the browser extension always disconnecting,

[north rain]

running circles with krill trying to fix command failed errors in powershell, anyone willing to assist ?

[carmine tusk]

Sorry have no technical knowledge, that why dont like talking on voice. but trying to listen and understand what u guys talk about

[hearty tundra]

hey ya'll, finding it tough to search chat history for this - is anyone using clawdbot with a team at work, ie multi-user? so as the team's AI assistant? wondering how hard this would be to set up / adjust for.

[paper dust]

⠹ pondering… • 37s | connected
agent main | session main (clawdbot-tui) | anthropic/claude-opus-4-5 | tokens 0/200k (0%)

i dont get why mine is stuck on pondering

[golden orchid]

any suggestions on how to optimize model usage? I'm using gemini api key and am getting rate limited (3m peak tpm) pretty quickly

[unique rune]

Anyone sorted the Anthropic connect? Do we have to pay for API now? They have blocked the other way?

[muted cedar]

Could anyone make Railway Template work?

[sour slate]

@muted cedar, 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

[twin apex]

my chat is stuck

restart doesn't help

[sour slate]

@twin apex, 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

[twin apex]

that doesn't help unfortunately...

[stark rampart]

has anyone got ollama working successfully? I feel really dumb that i cant figure out how to get it to work

[sour slate]

@stark rampart, any discussion about various AI models should be taken to #old-models.

[burnt bough]

I use iMessage groups with managers at work who usually contact me for property management stuff. Then I have vendor groups on iMessage that I contact for those maintenance tasks. I keep track of those task on reminders. I use iCloud/files to store invoices and permits for each of those 150+ properties. I was wondering if clawdbot can learn how to take maintenance requests and relay them to those vendors and then follow up until those tasks are completed.

My thoughts are I would share those files, reminder lists, calendars, and notes for it to have a good data set of information.

[golden sphinx]

mine is also stuck

[sour slate]

@golden sphinx, 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

[wraith coral]

is an API a requirement now? does claude code max plan work?

[golden sphinx]

go to ur terminal and paste

clawdbot gateway restart

[dry granite]

You can just run the install prompt again and choose a different model.

EDIT: I've got Clawd connected to telegram but now I'm getting the error that I've exceeded the credits for my GPT plan (I do the $20/month pro plan and I'm using the 5.2 pro model). I didn't even get to talk to Clawd, this was, like, my first message after verification! 😥 Any tips to start talking? Shift to a different model?

--
Hey everyone. Non-coder but pretty tech-savvy user here. Just downloaded the bot on a freshly wiped macbook today to test it out but haven't gotten it running yet.

What I've tried:

• Launched it from the clawd.bot code line in terminal
• Connected it with telegram
• Got the gateway up but couldn't get any responses back (like all the "disappearing messages" folks in the #1459642797895319552 channel--it seems to work but when I send a message I get a blank response)
• Troubleshooted according to the Imstuck page, worked with Codex (since I'm trying to connect GPT to it) to run the git version, that didn't work either.
• Went to help channel here: https://discord.com/channels/1456350064065904867/1465178751166972116 and followed its suggestion to connect whatsapp instead of Telegram (Telegram seemed to connect fine but in the gateway shows "Unsupported schema node. Use Raw mode." error) and sent myself a message after verifying the connection hoping Clawd would respond but... nothing

TLDR: Installed the bot, tried connecting telegram and whatsapp, can't message clawd anywhere and codex/help/imstuck haven't resolved the issue

Thanks, in advance for the help!

[dark kayak]

Hook a brother up..

[hallow flame]

shoud I post in #1458532054726606858 ?

[mental mica]

Will delete for now, DM peter (probably best on X) to get something setup

[dark kayak]

I think that is for people who want to contribute to Clawdbot

[hallow flame]

Thanks buddy

@low dragonCheck DM
Thanks, we would like to contribute to the project

[austere falcon]

I just set my clawdbot up on a mac mini, but literally after asking 3 questions in telegram i got

"LLM request rejected: Your credit balance is too low to access the Anthropic API. Please go to Plans & Billing to upgrade or purchase credits."

I know this cant be right or else this would be insanely expensive to use. how can I avoid this?

[mental mica]

use your claude sub

[austere falcon]

i put in the API

[mental mica]

thats not your sub

[austere falcon]

oh how do i connec that

[white hinge]

The only thing preventing me from fully using clawdbot is that i get disconnected every few mins.

Linux ubuntu, telegram, anthropic claude

[sour slate]

@austere falcon, 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

[obsidian nymph]

How to fix this?