users-helping-users

[nova solstice]

Yeah that's the one I visited and running into issue

[wintry ibex]

Odd it works for me

[nova solstice]

hmm weird, I see a few people in #1459642797895319552 channel also seeing the same issue. I tried multiple browser as well.

okay cool thanks for confirming it works for you, might be some network setting or my ISP doing somethign weird.

[wraith seal]

This is a Firefox specific issue.

Try Chrome?

[nova solstice]

I did try multiple browser, actually found the issue to be my ISP modem is blocking it.

[wraith seal]

This might be a funky SSL cert that Firefox is nitpicky about

[wraith seal]

Aaah.

Strange, who where they?

[vernal elk]

i've got a similar issue happening

[nova solstice]

Comcast lol

[wraith seal]

As expected.

[proud girder]

What is the command to uninstall everything on Mac OS?

[wraith seal]

moltbot uninstall

[nova solstice]

I'm on linux and tried installing with both npm install as well as install from git source. However, I don't see the executable moltbot and I've check that npm prefix -g is in my $PATH. I don't see moltbot executable inside $(npm prefix -g)/bin folder. Any ideas?

[wraith seal]

then using the package installer you used, remove it

[proud girder]

this is what it is showing 🙁

[wraith seal]

Do the directories moltbot and dotmoltbot exist?

[proud girder]

i honestly just run the "curl -fsSL https://molt.bot/install.sh | bash" and then i set up wrong now I want to re-do everything

am I doing this wrong?

[wraith seal]

Welp you uninstalled it, right?

If it's gone, it's probably dead

If you wanna redo it, run the same command blah blah blah

[proud girder]

no it is still on it, idk how to remove

[wraith seal]

What do you mean?

Are the directories still there?

[proud girder]

can I send you $25 and we can hop on a call pls

❤️

[wraith seal]

What 🥀

I cannot hop on a call rn bro.

[nova solstice]

nvm the git install dumps the executable to ~/.local/bin. Need to add that to my PATH. But weird npm install didn't work though.

[proud girder]

this is too confusing T-T

[wraith seal]

Run ls.

What do it show?

[proud girder]

I ran that thing again, but it is saying it is already installed

[wraith seal]

Okay so now it probably reinstalled it

[proud girder]

like it doesnt show me the original options like how to set up each parts

[wraith seal]

To set up, run moltbot onboard

[proud girder]

because i want to use imessage and it is not working so i wanted to re-do all set up

[wraith seal]

Ah, okay

Just run moltbot onboard

and you're set!

[proud girder]

like this?

sorry i know i am dumb 😭

[wraith seal]

Run source .zshrc

[wraith seal]

Ah.

Bro just run ls

[wraith seal]

Ah...

Now run rm -rf clawd

And run rm -rf .clawdbot

[blissful zodiac]

anybody getting this? ⚠️ Agent failed before reply: All models failed (3): anthropic/claude-opus-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) | anthropic/claude-haiku-4-5: No available auth profile for anthropic (all in cooldown or unavailable). (rate_limit).
Logs: clawdbot logs --follow

[proud girder]

nothing happened

[wraith seal]

Sir, that is called a ✨rate limit✨

[wraith seal]

Yeah that's expected output

Uhhh

[proud girder]

should I just factory reset the mac mini at this point?

[wraith seal]

Check Opus limit

[wraith seal]

Is anything important on there?

[proud girder]

no i bought this mac mini just to try clawdbot

[wraith seal]

Does it say that on the web portal though?

[wraith seal]

Ah.

Okay yeah probably do that

[proud girder]

kk ill be back in 10min to ask for help for proper set up, im super sorry with the dumb question ❤️

[wraith seal]

Don't worry lol. I'm here if you need me

[supple crown]

hi

Hi,

I’m running into an “agent behavior” issue with Moltbot/Clawdbot and I’m wondering if I misconfigured something.

Context:

• I give it a list of ~50 invoices to retrieve for my accounting.
• It has access to the browser (browser clawd) and my accounts (especially Gmail) to find missing invoices (searching Gmail, authenticating on vendor sites, downloading PDFs, etc.).

Problem:

• It retrieves 2–3 invoices, then stops and asks: “Should I move to the next one?”
• I reply “yes” and explicitly tell it to stop asking for confirmations and to run the task autonomously until completion.
• It says “OK,” retrieves another 2–3 invoices, then responds with something like: “I’ll continue in the background autonomously.”
• But then it actually does nothing after that. It looks like a normal LLM text reply (no looping / no continued execution) and it doesn’t keep working through the list.

My question: Why does it behave like a plain chat LLM (similar to chatting with ChatGPT) instead of behaving like an actual agent (e.g., like lovable.dev) where you give it a task and it continuously runs loops, checks its work, and keeps going until the task is fully completed?

[nova solstice]

Try following instruction here: https://docs.molt.bot/install/uninstall#manual-service-removal-cli-not-installed

[pastel chasm]

hi, having an issue with my moltbot where its keeps using ** bolding and comment blocks despite being in imessage, no matter how much i write to not use it. Has anyone figured out how to stop it? thanks in advance

[wraith seal]

Well, it requires lots of tinkering. In order to actually get the cool product you want you're gonna need to mess with config.

I haven't reached that level of fine, prestigious, superiority yet, but it's fine.

[wraith seal]

Just... Tell it not to use it.

[pastel chasm]

I have, told it every time, just doesnt stop

[wraith seal]

Check the memory. Does it actually jot it down in there?

[pastel chasm]

heres me asking it to add it to its soul and it immediately doing it again

also here in its memory ## Communication Preferences (Jan 28, 2026)

• NEVER use bold (text) or code blocks (''' or ```)
• Plain text and bullet points are fine```

[wraith seal]

This is standardized AI behavior most likely due the fact that they've probably been fed Markdown so many times they can't even comprehend not using it

[pastel chasm]

no way to like strip it before sending using a script or somehow converting it to imessage bolding?

[wraith seal]

Think. Claude. ChatGPT. Perplexity. Gemini.

[pastel chasm]

true...

[wraith seal]

[wraith seal]

Probably not. Don't add it in the soul though.

That's like their life force

Add it to ✨le memory✨

[pastel chasm]

okay sure ill remove it there

would it be possible to intercept the message before sendign to clear out the bolding? or no way

[wraith seal]

Probably not, and if there is, it would be

VERY

hard.

[pastel chasm]

okay, i guess ill just pray it gets a feature one day

thanks gabectl

[round thunder]

I tried to install calwdhub and it installed but I get an error about not having undict available. I am running node 25, is that a problem?

[wraith seal]

No prob

[wraith seal]

Wha?

Screenshot the full error message.

[round thunder]

Here you go, on my mac-mini

[wraith seal]

What's your package manager?

[round thunder]

npm

[wraith seal]

npm install undici -g

Run this.

[supple crown]

I understand it requires tinkering and changing the config, but my question is simple: what’s the point if, by default, it doesn’t work easily as an autonomous agent?

In 2026, if it mostly behaves like a plain LLM and I have to:

approve/restart actions constantly,

answer “should I move to the next one?”

then I’ll spend more time telling it what to do than doing the task myself.

So: is there a simple “autonomous agent” mode (similar to tools like ElevenLabs) where it can loop, self-check, and keep going until it completes a concrete task (e.g., 50 invoices) without constant confirmations?

If yes, what exact options/config should I enable to get that behavior — or is it really that hard to have a more or less autonomous agent?

[round thunder]

Perfect thanks.

[wraith seal]

Run sudo if it complains

[wraith seal]

Ngl you didn't have to use AI for that but

Just read the documentation.

You'll understand it more if you actually get a sense about it.

[sterile brook]

Probably a dumb question, but for the brave API, should it be the search API or the "search for AI" API? Not sure if molt would actually take advantage of the extra metadata.

[supple crown]

What’s the point then?
retrieving 50 invoices is a genuinely useful real-world work task. If AI isn’t meant to handle boring, repetitive, low value-added tasks like this, then what are we using AI for?

What I’m looking for isn’t “a more precise idea,” it’s:
Can Moltbot actually run in an autonomous agent mode (looping + self-checking + continuing until completion)?
If yes, which exact pages/sections of the documentation and which settings need to be enabled (approvals, timeouts, step limits, etc.)?

[wraith seal]

I meant you didn't have to use AI to generate your damn responses.

Yes, Moltbot can do exactly that. However, you actually gotta read.

I haven't exactly gotten to the process of going full automation.

Just tell it to summon a little baby agent which does all the invoice stuff, let it run, and when it's done, ask for it to message you.

[supple crown]

Sorry — I misunderstood. I’m in France and my English isn’t great, so I use AI mainly to translate properly (and it’s 5am here, so English is even harder).

Just to make sure I understand: are you saying I can ask Moltbot to create a dedicated “invoice agent” that will handle the whole workflow, run autonomously, and then message me when it’s done?

If yes: does Moltbot create/persist that agent so I can call it later for similar tasks, or is it only spawned for the current run? Also, does it do this automatically if it thinks it’s necessary, or only if I explicitly ask it to “spawn/summon an agent”?

[wraith seal]

It cannot persist those little agents, and you have to ask it to summon one. And yes, it will run autonomously because it must.

[supple crown]

Thanks a lot for your answers.

Okay, so if next week or next month I ask it to retrieve 50 new invoices, that agent won’t exist anymore. So I’ll need to explicitly ask it to “spawn/summon an agent” every time — is that correct?

If so, can I put this into a config file / an instructions .md file, like: “when I ask for tasks such as retrieving invoices, you must run autonomously and you must spawn an agent for that type of task”?

[proud girder]

hey I am back, im going for a fresh install with the goal of using imessage to communicate, what would be my first step

i am using mac mini

apple silicon

[faint tulip]

Wondering the same thing. I'm in a help thread with Krill or whatever but it's going very slow.

So are you saying just running a regular update on Clawdbot updates everything to Moltbolt. Because if true mine has not done it. I updated but everything still says clawdbot

[cosmic idol]

whats the best workaround to install it in a vps ? the direct install.sh flag many issues

[mild aspen]

Yeah its been a nightware with the preloaded VPS

[hasty harbor]

why there is no response, i have 0% of the the tokeen used
can anyone help me?

[steel cosmos]

what linux distro? I got mine working well on ubuntu 25 on a GCP VM

[kind hatch]

anyone getting banned using anthropic oauth for their clawdbot/moltbot?

[little pelican]

Hi - do you know if we can connect clawdbot to a phone number for imessage/whatsapp

[signal dust]

I continue to run into these and now my agent is stuck. Only replies with these.

Anyone else have this issue? Anyone have a fix?🙏

[slow geyser]

Anyone run into this? endless growing input context loop?

[proud girder]

does anyone know how to fix this?

[frozen fog]

has anyone managed to get vllm working, or anything that requires manual setup beyond setting the magic ollama variable?

[paper reef]

Can this run on a Vilros Raspberry Pi 5 8GB RAM? Have an extra- don’t see why it wouldn’t.

[zealous raptor]

That’s normal for a conversation until it compacts. You can manually /compact if you want.

[slow geyser]

but check the reponse sizes 18 tokens... many per min / obviously some issue.

[glacial abyss]

If you using cloud inference, I would think so

[manic solar]

The issue is a security policy restriction that I haven't been able to override via the configuration file.

Here is the breakdown:

1. The Tool Exists: I have the sessions_spawn tool, so I can ask the system to create a sub-agent.
2. The Block: When I try to use it (e.g., spawn coding or skill-creator), the system rejects it with:
   "agentId is not allowed for sessions_spawn (allowed: none)"
3. The Failed Fix: I edited your clawdbot.json config file to explicitly add an allowAgents list (including coding, etc.) and restarted the Gateway.
4. The Result: The system ignored that setting and still says allowed: none.
   Why?
   It is highly likely that there is a hardcoded environment variable (e.g., CLAWDBOT_SUBAGENTS_ALLOWED=false) or a master policy in this specific runtime environment that overrides the local JSON config file.

Impact:
I cannot delegate tasks to background agents. I must do everything (scraping, coding, writing) sequentially in this main chat with you.

I'm fighting with this since 2 days.
It can't make sub agents no matter what.
Anyone could help?

[daring vault]

Hi all anyone can help on AWS EC2 gateway setup I got the following message: Service: systemd (enabled)
File logs: /tmp/clawdbot/clawdbot-2026-01-29.log
Command: /usr/bin/node /home/ubuntu/.npm-global/lib/node_modules/clawdbot/dist/entry.js gateway --port 18789
Service file: ~/.config/systemd/user/clawdbot-gateway.service
Service env: CLAWDBOT_GATEWAY_PORT=18789

Config (cli): ~/.clawdbot/clawdbot.json
Config (service): ~/.clawdbot/clawdbot.json

Gateway: bind=loopback (127.0.0.1), port=18789 (service args)
Probe target: ws://127.0.0.1:18789
Dashboard: http://127.0.0.1:18789/
Probe note: Loopback-only gateway; only local clients can connect.

Runtime: stopped (state activating, sub auto-restart, last exit 1, reason 1)
RPC probe: failed
RPC target: ws://127.0.0.1:18789
gateway closed (1006 abnormal closure (no close frame)): no close reason
Gateway target: ws://127.0.0.1:18789
Source: cli --url
Config: /home/ubuntu/.clawdbot/clawdbot.json

Service is loaded but not running (likely exited immediately).
File logs: /tmp/clawdbot/clawdbot-2026-01-29.log
Logs: journalctl --user -u clawdbot-gateway.service -n 200 --no-pager

[sour slate]

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

[desert crest]

Evening, I assume its been asked but openrouter and kimi-k2.5 doesn't respond in clawd anyone have a fix?
shows as an unrecognized model

[latent hedge]

I'm slightly confused. Everywhere in the docs, it refers to terminal commands as 'moltbot <parameters>'.. but that doesn't work for the install I did yesterday for the first time... it only works if I type 'clawdbot'. Anyone else have that issue?

[wooden zinc]

#old-models message

[proud girder]

anyone know how to set up imessage properly? it is my one millionth time setting up idk what is wrong lol

[round thunder]

Ok, probably a dumb question. My mac-mini is set up and I have clawdbot connected to telegram and it is talking. But periodically it stops responding,. It is like the mac has gone to sleep. What I would expect on my laptop when I close the lid I guess.

Is there a setting I am missing?

[round thunder]

That I can answer. Say clawdbot the name got changed yesterday and hasn't flowed through everywher eyet

[rocky orbit]

Your Mac needs to be always on yes!

[crimson steeple]

After 7 reinstalls of this, I might have to throw in the towel soon..
I'm getting constant issues after extended usage including the gateway completely bricking itself, the ai having constant timeout errors, failing to fallback
All of this without very clear reasons why it is failing, I've spent more time troubleshooting it than using it at this point... and all of this from normal use

Running out of quotas I understand but it doesn't seem to support anti gravity running out of quotas and just completely pile up on queues

I suppose we're just not there yet I suppose..
Cron breaks by default unless you ask it to be fixed

Ah.. but sure, open a pr and fix it myself... but I'm not looking for that yet

I suppose it ain't for me..

[round thunder]

It is on. Is there a setting i have missed

[dire bluff]

How do we ensure our main agent always responds to the user when given a task that involves restarting the gateway - either implicitly or explicitly?

Because gateway restarts always kill the agentic session, and the model never responds after it does a change that updates the moltbot.json (which triggers gateway restart).

1. User messages agent
2. Agent calls a tool that modifies clawdbot.json (e.g., clawdbot agents add)
3. Tool completes successfully, writes to config
4. Config watcher detects change within 300ms debounce
5. Reload logic sees meta.lastTouchedAt changed
6. No rule exists for meta.lastTouchedAt, so it falls through to restartGateway = true
7. Gateway sends SIGUSR1, restarts
8. In-flight HTTP request to Anthropic/OpenAI is aborted → AbortError: This operation was aborted
9. User never receives the agent's response

[scarlet folio]

how to install on docker

[median skiff]

hello all! I'm new to community and hoping to get help regarding whatsapp integration. Clawdbot runs fine and answers me on whatsapp but doesnt get any direct messages from 3rd parties. How can I fix this? Thank you in advance for the help!

[late swan]

I'm encountering a persistent 404 Not Found error when using the Gemini 1.5 Flash model with the latest version of Moltbot (Clawdbot). Even though the model is correctly configured, the bot seems to be sending an incorrect path structure to the Google API.

@median skiff

[molten quiver]

I watched a YouTube video saying using my Claude max subscription in moltbot will get me banned but I can’t see any mention on these forums. How true is this. Should I be switching to a new sub with another AI

[prisma bolt]

whatys the recommened ai to use with this? kimi? im currently using claude

[ivory rapids]

ask CC to prepare it for docker

[flint ether]

Everything was fine, but suddenly I get "Unhandled stop reason: error" using gemini 3 flash through openrouter. Any suggestions?

[late swan]

hey are you using free api?

i am facing issue tried openrouter,aistudio both still not working

[flint ether]

Nope. Pay for tokens. Gemini is already up. I think that was some provider problem.

[late swan]

so if i wanna test out for free so no option?

[flint ether]

There are some free models on the openrouter. Although IDK how good are they.

Or, rather, how bad

[exotic shoal]

hello crustaceans, i keep running into an issue when setting up Clawdbot with my max subscription / oAuth token. Using Ubuntu on AWS, getting: HTTP 401 authentication_error: Invalid bearer token. I am sure this is s know issue?

[weak kayak]

Hi guys quick questions - for us on Clawdbot version, do we need to migrate to moltbot in order to continue to receive updates?

[late swan]

i just want that in whatsapp bot reply to me wen i say "hello" haha not even this happening keep giving me multiple error

[true gulch]

Hi, what's the right way to change the API key (e.g Codex to Anthropic)? Yesterday with another user here we had this problem. Should I launch the clawdbot onboard and choose the model? Or Should I execute the clawdbot configure?

[flint ether]

For that you can defiantly start with some free models. Deepseek or kimi would work fine IMO

[prisma vault]

Since Krill doesn't seem to help, I'll ask it here. How can I make it so different agents use a different telegram account?

[floral sand]

I think there is no guide because there is no real migration yet. No releases since before the name change. I think the Moltbot 0.10 just creates some aliases for some commands; but I could be wrong on that. Main repo shows the latest release as 2026.01.24 which was before the name change.

[keen carbon]

I'm trying out Moltbot on Docker using version 2026.1.27-beta.1 and wanted to sanity check something about skills.

I'm able to install skills and they're detected correctly, but I can't figure out how they're supposed to be run in this version. I tried triggering them via chat (Web UI and WhatsApp) and also looked for a CLI command, but didn't find a working path.

Is this expected in the current beta, or am I overlooking the intended way to execute a skill?

[patent parrot]

L O L. Check X for tutorials there are TONS

[past gulch]

Hi, all , I installed ClawdBot using CC CLI

Hope it helps on VPS:

[night urchin]

https://www.dbreunig.com/2026/01/08/a-software-library-with-no-code.html

yw you degens ❤️

[obtuse sundial]

Anybody else using Gemini cli + antigravity with Molty?

[silent spruce]

Hi guys.
Anyone else getting "HTTP 401: User not found" while using openrouter with models. I am using minimax m2.1. My keys are set up correctly an I have credits.

[silent stag]

Is anthropic really banning users? Did it happen to any of you?

[wet oxide]

Anyone having this issue?

LLM request rejected: messages.2.content.1.tool_use.input: Field required

[stoic shard]

What is the recommand / command way to access to SMS message from ClawdBot ?

[floral sand]

I know there is voice call access through Twillio... not sure about SMS.

[hidden current]

why isnt it giving me a response

[jade topaz]

i am trying to install moltbot in my local macos.. while loading the webpage, i am able to chat, but the bot is not respnding.. wht can be the reason ?

[stiff tusk]

if you're able to chat then the bot is responding

[jade topaz]

bot is not sending any reply ..

[late swan]

[clawdbot] LLM error: {
"error": {
"code": 404,
"message": "models/gemini-1.5-flash is not found for API version v1beta, or is not supported for generateContent. Call ListModels to see the list of available models and their supported methods.",
"status": "NOT_FOUND"
}
}--- i am getting this error and i was trying to run clawdbot from 2 days but multiple error coming again and again also ⁠help bot not giving me right solution any human dev here ? to resolve my problem?

[stiff tusk]

try clawdbot logs --follow to see whats wrong

[jade topaz]

do they haev a dmg file for macos ?

[stiff tusk]

only for the companion app which I would not recommend using

[jade topaz]

ok

[jade topaz]

clawdbot logs --follow

🦞 Clawdbot 2026.1.24-3 (885167d) — Give me a workspace and I'll give you fewer tabs, fewer toggles, and more oxygen.

Gateway not reachable. Is it running and accessible?
Gateway target: ws://127.0.0.1:18789
Source: local loopback
Config: /Users/moltnew/.clawdbot/clawdbot.json
Bind: loopback
Hint: run clawdbot doctor.

[stiff tusk]

clawdbot gateway start

[gusty sonnet]

Is that model string in your config as "models/gemini-1.5-flash". Should probably be "google/gemini-1.5-flash"

[jade topaz]

This site can’t be reached
127.0.0.1 refused to connect.
Try:

Checking the connection
Checking the proxy and the firewall
ERR_CONNECTION_REFUSED

now page is not loading .

[late swan]

ye thats the issue i put on json google but bot get model instead

[stiff tusk]

clawdbot status

[onyx burrow]

hi, can someone tell me what this is ?

connected | idle
agent main | session main (phone) | openai-codex/gpt-5.2 | tokens 54k/400k (13%)
────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────

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

[gusty sonnet]

Looks like that model isn't supported. Better to stick with gemini-3 models.

[gusty sonnet]

The TUI?

[onyx burrow]

i have gpt business and connected to it via Oauth so i dont understand why there is a limit how much i have to pay ect...

[late swan]

so it is free to use for limited time?

[scarlet folio]

I used railway

how to put the channel of telegram here?

[wide dirge]

What is the correct configuration for clawd.json when using multiple agents with different Telegram bots? I have set up 2 agents and two different Telegram bots, but when the second agent starts working, the first agent can no longer receive messages I send from first tg bot. It seems there is an error in the configuration file binding. Can anyone help me with this?

[late swan]

?

[valid vine]

Why is everyone running clawd on a Mac mini and not their personal computer/laptop? like MacBook Pro or iMac

[primal portal]

Looking to pay someone to secure my Mac mini + VPS Clawdbot setup stack, DM if you do DevOps/SecOps.

[onyx terrace]

When I set the google api key for Gemini, it just doesn’t work, I text the bot and it does nothing, running it with github copilot pro for now. Anyone knows how to fix this?

[fringe frigate]

What is the best way to get the bot to transcribe voice memos received on whatsapp?

(i have a wokring whisper api and i can do it with scrript calling but its not automatic)

[wicked ingot]

To keep things separated - pretty self explanatory as to why you would not want an AI assistant to have access to your primary PC. Zero - and I mean zero % chance id install this on my primary.

[late swan]

same issue

[onyx terrace]

seems like I’m not the only one, I’ve never had issues with google API keys till this btw, used it on multiple services

[crimson steeple]

For people searching through the discord in the future
I got molt working on android, termux. I'm on a OnePlus 15.
The only thing making molt not launch on termux was the lack of arm64 clipboard so you can try to make a fake clipboard module to get around it

cd /data/data/com.termux/files/usr/lib/node_modules/clawdbot
mkdir -p node_modules/@mariozechner/clipboard-android-arm64
echo '{"name":"@mariozechner/clipboard-android-arm64","version":"1.0.0","main":"index.js"}' > node_modules/@mariozechner/clipboard-android-arm64/package.json
echo "module.exports = { getText: () => Promise.resolve(''), setText: () => Promise.resolve() };" > node_modules/@mariozechner/clipboard-android-arm64/index.js

This doesn't mean everything will work, from what I heard, canvas might not work. Alongside other general features, couldn't test it too far. Android isn't very well supported and break often, not recommended at all

[haughty hedge]

what are we doing if we're still on clawdbot and not molt?

[onyx terrace]

tried to "fix" that yesterday with no success at all, ended up just leaving it alone, i suppose it will update itself on the next updates

[past path]

I'd love to give moltbot a try and install it in my macbook to try out some things - can it be relocated later to another machine easily?
What about chrome / browser support and that stuff on a raspberry Pi or a linux based VPS?

[onyx terrace]

don't install it on your personal machine, a vps is literally 5 bucks

[haughty hedge]

only way i could thinnk of is installing moltbot and manually copying the key clawdbot files, seems a messy way to do it though.

[past path]

can it use my max20 claude subscription then?

[onyx terrace]

i ain't messing with all that, wasted time for nothing, only diff is you gon type moltbot instead of clawdbot, leave it like it is till it's updated

[onyx terrace]

i think you can, a lot of people are doing it but i'm not 100% sure about what Anthropic thinks about that, they like API a bit more😆

[past path]

thats why I thought local machine would be a good starter 😄

[onyx terrace]

please don't, never use tech of this level of intelligence in personal machines. Just get a VPS, try it out, and if you like it get a mac mini or a good machine that will be able to fully take advantage of the tool

this thing is literally thinking for itself

[past path]

sounds good 😄 ah if you'd know what my claude automations do on my machine already 😄

[ebon ivy]

The docs had a full video about how to set it up, that was great, but I assume it was removed since the name was updated. I'm looking for that video, or any other helpful like that.

Things like that: https://www.youtube.com/watch?v=mDsyFrQPPfg

I'm willing to get started here

[onyx terrace]

I get it, but claude is reactive, not proactive. This thing is built different. This is like sex, you will enjoy it a lot, but take precautions before it's too late to take them

[past path]

love this response 😄

[restive geyser]

Hello guys please i have this bug

only i got the first response from the chat

[open violet]

Hi All, Big shout out for the hard with with Moltbot. I might have have stuffed up my install, is there a quide to trash tall and start from begining please?

[exotic shoal]

hi all, did anyone have any luck setting this up with a claude max subscription as opposed to an API key?

[wide musk]

Hi all. I’m trying to talk with my bot via telegram and I have this problem.

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

[lone prairie]

Context overflow: prompt too large for the model. Try again with less input or a larger-context model.
long term solutions for this?

[gilded hatch]

gm gm my bot (set up on vps) loses context regularly. 6 times in the last 24h. it really feels like talking to Dory from Nemo. "Hey what can i do for you?". how do i fix that?

[wide musk]

And how can I restart this?

[restive geyser]

any ideas

[lone prairie]

write /new or /reset i think

[carmine ginkgo]

Hey guys, how is the linux support compared to macos ? I have a macos VM but its flaky. Lots of skills use brew, so i am wondering if linux has a good enough stable feature set to be worth it instead

[snow thicket]

I hit a rate limit after a few prompts on opus. How much tokens did you guys consume on the first try.?

[wraith seal]

Linux support is fine. However, you lose some features like iMessage, Notes, or comms with iCloud

[waxen hornet]

brew existing on Linux but requires a newer cpu (SSSE3 required), the old vpn I have mine on doesn't so no brew for me sadly...

[quartz stone]

I am quite confused here, I´ve got everything installed and sorted although I keep running into an issue, The yes option doesn´t exist and even if I do "Yes" or "Y", It doesn´t work

Does anybody know what I can do to fix this?

[jaunty token]

You should keep it as yes

[quartz stone]

It´s not giving the option to select yes

and gives error when I try

[prisma bolt]

did it try spanwning a subagent for yall?

[granite carbon]

I just installed ClawdBot today and have run into the HTTP 429:rate_limit_error at least 8 times today. I have a free Claude.ai account and paid for API tokens, but I keep seeing this error over and over.

I’ve been asking Claude and ChatGPT for help, and both insist that a paid Claude plan won’t resolve it because this error is due to rate limits on Anthropic’s side, but I’m curious to know if others are experiencing the same issue too? Or do I need to fix something?

[lone vessel]

I have this issue continuously as well. No idea what to do

[snow thicket]

Man I have the exact same problem. I checked then api usage for opus and it’s already at input 2M tokens. I only did 3 prompts

Im gonna try kimi

[granite carbon]

The FAQ suggests setting up a fallback service, but I can’t do that until I get to my machine that’s running it. I didn’t think it would stall this much though, but then again, it’s my first day of massive brain dumps into it.

[spare holly]

I have an issue with using antigravity (auth), when one model is fully used, the bot doesn't switch to one of the backups, e.g. from gemini 3 flash to gemini 3 pro. Any idea how to fix that?

[wide dirge]

2026-01-29T13:24:05.378Z [agent/embedded] google turn ordering fixup: prepended user bootstrap (sessionId=59f9135a-3721-45eb-adad-64318b7c0931)
2026-01-29T13:24:20.981Z [agent/embedded] Removed orphaned user message to prevent consecutive user turns. runId=d723246d-5116-4826-830f-68264894ba62 sessionId=7428ed35-facb-4f20-9c06-72a941e87d3e
2026-01-29T13:25:18.588Z [tools] exec failed: Command exited with code 1

When conversing through a Telegram bot and gateway, errors consistently occur, causing the Telegram bot to stop responding.

[sour slate]

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

[lone vessel]

So should I switch to discord or something?

[sour slate]

Use #1459642797895319552 for help. The fastest way to get your problem solved is to follow the instructions here: https://docs.molt.bot/help/faq#im-stuck-whats-the-fastest-way-to-get-unstuck

[stoic shard]

If you launche Clawd to a task and is long.
But he must have misunderstood it because of poor prompting.

Is there a way to interupt the task ?

[restive geyser]

i use groq with clawdbot its consume alot of tokens without result

[formal fog]

ok, last night I got everything working with a local llm on a linux box. telegram worked too. my project for today was moving all the tokens and such to env variables in an .env file and making sure when the machine rebooted, that all was working again. clawd status shows everything working. no parsing errors. but message sent in the web dashboard just return an empty response and telegram messages just get a connection error. ollama is running , the dashboard shows health ok. What am I missing?

[crisp edge]

Hello everyone,

I setup moltbot with telegram,, but it struggles to schedules messages, it tells me something like :

Instead, I can directly schedule a reminder for you to receive a "hello" message in 10 seconds using the correct format.

Let me fix and try again to send you a "hello" message in 10 seconds.

I'm not sure what's going on

can anyone help please ?

[turbid sage]

I keep getting a “terminated” response… almost like I somehow broke the model integration. I’ve attempted editing the json that holds the model options and got it to work for a bit. Then it stopped again. I feel like I should start over - wipe it all… but that would suck. Is there a way to wipe just the models and start fresh or should I ask I different question?

[zealous raptor]

What model?

[torpid wing]

Hi, how i connect my moltbot with my metamask

[late swan]

Hey yo

Guys

[gray phoenix]

Can I run it on raspberry pi?

[simple plume]

Got it mine working like this...
you have to select the Anthropic compatible option and then go to your anthropic acount and generate a key and paste it to clawdbot

[wide dirge]

2026-01-29T14:27:07.516Z [clawdbot] Unhandled promise rejection: TypeError: fetch failed
at node:internal/deps/undici/undici:14902:13

Why do I keep getting this error message? I've restarted the gateway, but the bot isn't responding at all, and I'm completely baffled.

[mental mica]

use the bluebubbles option for imessage its better

[spice steeple]

PS C:\Users\user> iwr -useb https://molt.bot/install.ps1 | iex

Moltbot Installer

[OK] Windows detected
[OK] Node.js v24.13.0 found
[*] Installing Moltbot@latest...
node.exe : npm error code ENOENT
위치 줄:1 문자:1

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

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

install error

Looking for someone to help me with an installation error.

#1459642797895319552 Looking for someone to help me with an installation error.

[late swan]

Bro you used ollama one? On pc

[cerulean ibex]

Hey guys, i have my bot installed on amazon e2 , windows server..
Its working.. but i cant set cron jobs at all... it is tryign but i never receive messages. What can i do? I told the bot to read the documentation.. but no success at all.

Something simple like: "Remind me in 10 minutes i have an apointment." -- it never sends the message in whatsapp to me

[austere berry]

Has anyone else had this happen when trying to use the kimi k2.5 api?

[formal fog]

well, seems like some part of the JSON was wrong the first time, but when I reverted and put the env vales in one by one it all worked. yay

[gusty jetty]

hey might be stupid question how do i put in the API key, i want to use claude but in the quickstart i selected yes for "check keychain for claude code CLI". after that it never asked for the key in the terminal... in the log section im getting this error:

Embedded agent failed before reply: No API key found for provider "anthropic". Auth store: /Users/kandymac4/.clawdbot/agents/main/agent/auth-profiles.json (agentDir: /Users/usernam/.clawdbot/agents/main/agent). Configure auth for this agent (clawdbot agents add <id>) or copy auth-profiles.json from the main agentDir.

[fervent forum]

Hello, we've incorporated Moltbot into the OctaSpace marketplace and we're testing it out, but we've noticed an issue when adding the Gateway token

[unreal bear]

I’m setting up on a Mac. Anyone have the most recent setup code to run in terminal?

[simple plume]

Nope, Im using claude Pro subscription...
if you want local with ollama you will need huge resources since for what I have seen it uses large context...
you can only send a few messages before your daily token quota to be maxed out

[unreal bear]

Hopefully things have gotten smoother than 4 days ago

[late swan]

You using free api?

[pearl solstice]

yeah anthropic api usage is a scam. what im figuring out with with this whole clawdbot thing is dont waste your time on it if you're not a millionaire. Most smaller models (running gpt-oss-120b) just cant handle tooling at all. ive burned through 20m tokens just trying to explain how to use the browser

[late swan]

So what should I do to run on pc??

I just want to try clawd is working or not

Can you tell me which model is good where I can try free api for limited time to test

[pearl solstice]

So if you cant afford a 300 dollar a month claude bill, dont bother

[simple plume]

haha

[late swan]

Really?

[unreal bear]

He’s not wrong

Especially if you want “automation”

[prime nimbus]

Github CoPilot works perfect if you have that subscription:

[simple plume]

I need to try with google I pay there for the Gdrive and have gemini pro....

[pearl solstice]

Yeah, gpt-oss-120b is completely incapable

[late swan]

Gemini not workin

[pearl solstice]

ive burned 20m tokens just explaining how its basic tooling works and it keeps sending bad json

its mixing conversational input and json and just failing

[late swan]

So this works?

[cerulean ibex]

I am using copilot github pro as well.. but, with gpt 4.. it is not great. Cant set reminders at all. Maybe it is the model?
If i change to premium requests, like sonet 4.5... it will burn the tokens fast (300 reqs max)

[pearl solstice]

Claude, works immediately, first prompt no questions asked. If you want automation and actual jarvis like behavior, step one: become a gold miner, step 2: spend 500 a month on claude tokens

or convince your company its mission critical

[prime nimbus]

100 percent model, or your setup, mine works out the box with copilot-proxy mate, been setup 24 hours now and heres the estimated savings:

[late swan]

So like what's the cost if we want to use it properly you think?

[pearl solstice]

again, 3-500 a month depending on usage

if youre a daily driver, where you need automation, you wont be able to have it work properly for less

[late swan]

There is any list ? What can clawdbot do and the all cost of what he can do is maximum 500 bucks

[pearl solstice]

Ive spent like 4 days poring over every single alternative and possibility. If you want clawdbot to work as you see it work for others, you have to fork over quite a bit of cash

[late swan]

True I'm also trying for like 3-4 days
But no results seems like need to put money on it to see results

[pearl solstice]

For the record, im not saying anythings wrong with clawdbot. I am complaining about anthropic

[formal fog]

i am just a newb at all of this, but I have it setup with a local llm qwen25-32b on a GMKtec EVO X2 AI Mini PC
AMD Ryzen AI Max+ 395 3.0GHz Processor; 64GB LPDDR5X-8000 Onboard RAM; 1TB Solid State Drive; AMD Radeon 8060S Graphics, its working, reasonable response times. I have only just started playing with things tho.

[late swan]

What's task that you do

With this

[pearl solstice]

What you are experiencing, is this wall that everyones trying to solve. AI is not and has never been profitable. Companies charge exorbitant pricing for it. ATP Things like this are only available to businesses or the mega rich

[formal fog]

so far just basic questions, and I had it write a script to back itself up with all the memory & config files

[late swan]

But all these task you can do with other ai too?

Gpt etc.

[late swan]

Fr

[formal fog]

i am going to add tasks until it stops working 🙂

[cerulean ibex]

just a quick question, anyone was able to get a programmed message from the bot? Like a reminder?

[late swan]

I thought I'll try for my company but i didn't see anyone here that is doin well with clawd

Maybe bcz it's under process

[late swan]

Nop

[austere berry]

No, I pre-loaded $120 directly from Moonshot's site to test it out.

[cerulean ibex]

So, it is working like just any other chat bot directly, but with some system access...

[late swan]

Ohh and what task you do from this

Like what's the worth i mean

[pearl solstice]

yeah the big thing is it can prompt itself

thats the big selling point

[late swan]

Means?

[scenic rune]

whats wrong with my configuration:

[clawdbot] LLM error: {
"error": {
"code": 429,
"message": "You exceeded your current quota, please check your plan and billing details. For more information on this error, head to: https://ai.google.dev/gemini-api/docs/rate-limits. To monitor your current usage, head to: https://ai.dev/rate-limit. \n* Quota exceeded for metric: generativelanguage.googleapis.com/generate_content_paid_tier_input_token_count, limit: 1000000, model: gemini-3-flash\nPlease retry in 22.842510727s.",
"status": "RESOURCE_EXHAUSTED",
"details": [
{
"@type": "type.googleapis.com/google.rpc.Help",
"links": [
{
"description": "Learn more about Gemini API quotas",
"url": "https://ai.google.dev/gemini-api/docs/rate-limits"
}
]
},
{
"@type": "type.googleapis.com/google.rpc.QuotaFailure",
"violations": [
{
"quotaMetric": "generativelanguage.googleapis.com/generate_content_paid_tier_input_token_count",
"quotaId": "GenerateContentPaidTierInputTokensPerModelPerMinute",
"quotaDimensions": {
"location": "global",
"model": "gemini-3-flash"
},
"quotaValue": "1000000"
}
]
},
{
"@type": "type.googleapis.com/google.rpc.RetryInfo",
"retryDelay": "22s"
}
]
}
}

[simple plume]

Some folks are doing some great things with this ...
Go to the moldbot site /showcase ...
We and Im including myself we need to read a lot more ...

And regarding to local ... I don't think you can do it at least you need 2 3090ti or something like that ...

Go for the 20 buck pro subscription and test the bot ...

This https://www.molt.bot/showcase

[prime nimbus]

Docs are amazing here, the answers are normally in there, or lead you down the right way. Also, rtx 2080 works fine with ollama and qwen, it really depends on you, ive moved to gh copilot with gpt 5.2 and swapping to gpt 5.2 codex too when im about to get down and dirty

[simple plume]

I have a 1660 super 🙁 sad face..- 6Gb Vram...

[prime nimbus]

i cant stress enough how much the model matters too. For example out the box gpt5.2 works best i think, you can get it to handle crons etc:

[prime nimbus]

yeah 8gb is where the starter should be set

[plain cradle]

Hi, thanks in advance, I really need help with an error I am encountering. I have a "Message ordering conflict - please try again. If this persists, use /new to start a fresh session." error. I tried all suggested fixes which I found online, like clearing all sessions, swapping models, hard restart of gateway, etc, but none of it works, the error persists. I am running clawdbot within WSL2.

[simple plume]

Im sorry Im no expert but...
mm have you check WSL2 and or windows DateTime?

[plain cradle]

how do I do that? and could you help me understand how this would fix the issue?

[hasty jay]

I am not getting local ollama under model/auth provider config. any suggestions?

[torpid wing]

How i make a skill ?

[spice steeple]

Looking for someone to help me with an installation error.

[torpid wing]

i wanna teach my molt to trade

[prime nimbus]

follow docs, if these dont work you've done something wrong: https://docs.ollama.com/integrations/clawdbot

[foggy leaf]

Has anyone else hit the "Cloud Code Assist format error" infinite retry loop when using Anthropic providers with browser automation (Playwright)?

There's an open issue for it: https://github.com/moltbot/moltbot/issues/2769

The bug is that tool call ID sanitization only runs for Google/Mistral but not Anthropic - so the retry never actually fixes anything.

Wondering if there's a workaround until this gets patched. Anyone found a way around it?

[hardy ravine]

so im confued it named to moltbot but we still use clawdbot alias to run it?

[lilac nexus]

Guys. do you know how to enable the https://docs.molt.bot/nodes/audio for an agent? I tried to voicetext my telegram bot but no luck, the voice was saved as an ogg file under ~/.clawdbot/media/inbound

[lime sequoia]

I am getting this error with a custom openai endpoint.

[spare arch]

my bot isnt using my claude api key and has ben running for 2 days and im scared to check what its using

My claude usage hasnt changed

looooool

[prime nimbus]

yeah, things cant all happen at once right now, its being worked on, install the beta if you want full moltbot, if not stable is clawdbot for now

[hardy ravine]

@prime nimbus okay cool just wanted to confirm thanks!

[wet drum]

      "memorySearch": {
        "provider": "gemini",
        "model": "gemini-embedding-001"
      }

What's wrong with this? The docs say I can use gemini for vector embeddings but clwadbot says agents.defaults.memorySearch.provider is invalid

[unreal steppe]

Bot not responding to anything. Any ideas?

[opal plover]

For people using codex are you using 5.2 or 5.2 codex model inside clawdbot

[wooden yoke]

Hi Guys, is claude suspending accounts for using clawdbot with the Max subscription?

[limpid wyvern]

I always get thsi message "Please approve command"

[wooden yoke]

mine is also not very proactive

I’m running on openai-codex/gpt-5.2 (Codex CLI auth).

Session status also shows:

Thinking: off
Runtime: direct
Context: ~99k / 400k

I just isnt proactive with this model.. What model do you guys suggest?

[limpid wyvern]

he cant install scripts always say I need to approve something ?

whats the cheapest model at the moment here ?

[prime nimbus]

strange! 5.2 is the best? -

[solar perch]

15:46:53 [clawdbot] shell env fallback failed: spawnSync /bin/sh ENOENT
15:46:53 [clawdbot] shell env fallback failed: spawnSync /bin/sh ENOENT
15:47:53 [clawdbot] shell env fallback failed: spawnSync /bin/sh ENOENT
15:47:53 [clawdbot] shell env fallback failed: spawnSync /bin/sh ENOENT

any knowledge about that?

[sinful dawn]

@odd shard for the record, we appreciate you helping.

[odd shard]

Thank you, I understand how frustrating moderation can get.

[unborn quartz]

I don’t know, but I’m running into the exact same thing. It doesn’t look like the requests to the LLM are going though Moltbot. I’ll let you know if I figure out how to fix mine

[odd shard]

Do you have Claude Code?

I've been having my CC diagnose issues with ClawdBot

One being that CB kept stopping itself but not saying anything, so it was a silent failure.

[sinful dawn]

this works so well. never had a config misconfiguration claude cli couldn't fix - without any help in most cases

[unreal steppe]

Yeah, randomly been having issues with that too, but may just restart everything

[hasty jay]

for linking signal app as the channel , the documentation asks to run the command signal-cli link -n "clawdbot" then scan QR code. but the terminal is not giving the QR to scan

[muted token]

Hello there! My apologies if this is a familiar topic—I have been searching around and haven't found much, so I will go ahead and ask: I have ClawdBot running on a headless Linux machine, and I want to understand how to make it interact with websites behind login pages. I have checked a few skills, but they don't seem to work on a headless machine, and I am unsure if a scenario in which ClowdBot can take control of a browser on a laptop via extension is possible.

[tulip dune]

Hello everyone. I hope y'all are having a great day. I am trying to install clawdbot i get throught the onboarding to were i setup the telegram bot and after setting it up i go to the next screen and i am getting this error

◇ Select a channel
│ Telegram (Bot API)
│
◇ Telegram already configured. What do you want to do?
│ Skip (leave as-is)
│
◇ Select a channel
│ Finished
Updated ~/.clawdbot/moltbot.json
Error: EACCES: permission denied, mkdir '/app/I:'

[odd shard]

Have CC SSH into the bot itself and he can do things for you to fix everything

I have it fix my configs, upload models, chain skills together. etc

He even created a command for me to ssh in withou tthe ip address

I just say ssh clawdbot and it connects, or ssh clawdbot-ui to show the dashboard

[sturdy dawn]

did you try?

[tame fractal]

I want clawd to navigate and interact with twitter and avoid bot detection. Im thinking a headless browser, whats the best approach? When I ask clawd I don't get a good answer. Keeps refereing to Brave web search but that won't work

[tranquil jay]

Hi everyone,
Has anyone encountered an issue with the browser tool? I turned ON the extension on a tab, but when I ask my bot to open Chrome and perform an action, it opens a new tab where the extension is not ON, then throws an error.

Does anyone know how to fix this?

[tall fossil]

try to have another bot like claude code help you get through

[unreal steppe]

oh damn, that's awesome, will try

[limpid wyvern]

I always get this error "Failed to submit approval" someone knows how to fix that ?

[real jewel]

my gateway is not running nor connected, what do i do?

[limpid wyvern]

restart

clawdbot gateway restart

[late swan]

Which api

You usin

[small jackal]

has the Name inside the app changed ? mine says still clawdbot ? can someone please clarify this to me

[late swan]

Ye its still clawdbot

[small jackal]

thanks jack ..was a bit worried installed it 10 mins ago thought ...i had a wrong version

[late swan]

Clawdbot working fine for you?

Or some errors

[real jewel]

which api?? i dont know, i had it running, and i installed with bluebubbles chat, but didnt have it, so tried to setup discord, clawd started troublshooting, and everything stopped, i reinstalled, i cannot connect channels, was using minimax for ai

[small jackal]

it started with setup thing ..name vibe etc ..and now its not responding after tried a test and asking for the weather info

[real jewel]

its saying: gateway service disabled

[late swan]

Ohh
Lol

[late swan]

I faced this too

It's bcz multiple commands in terminal

[small jackal]

i installed it on my old gaming laptop with a fresh linux

[late swan]

I need to open my pc to see how i resolve it lol

[small jackal]

should be secure enough ^^

[late swan]

But tbh everytime I resolve bug or error I got new error 😞

[small jackal]

good luck jack 🙂

[late swan]

I'm tired lol I need someone that is using this bot smoothly so i can ask how

Hahah

[small jackal]

Jack you had it working on one time ...?

[real jewel]

after using : systemctl --user start moltbot-gateway.service, received: Failed to start moltbot-gateway.service: Unit moltbot-gateway.service not found.

[steep ermine]

I had my molt working totally fine for a few days. Now it doesn’t respond it gives me the typing icon for like 3 mins then tells me I didn’t get it enough context. Any ideas?

I did install a “ memory” add on from GitHub I found that’s the only non standard thing I have done

I also noticed when I went through the setup again most the plugins were now “ missing requirements”

[main plume]

I hope you get it solved. I am working on an issue here and trying to sort it out. I am just a newbie so mostly looking for information. This looks to be very promising for development 🙂

[late swan]

Wht problem?

You facing

[main plume]

Installing

[lime sequoia]

I would say clawdbot literally doesn't work with 3rd party openai compatible endpoints

They should have added an option to add it in the onboarding itself

It isn't too hard

I am trying to setup it from 3 hrs

and It still just gives a 403 status code error

[main plume]

I have just been on it a few minutes but saw that there was a discord group so thought I would hop on here and see who has had the same issue

[faint bloom]

Anyone else able to get Kimmy 2.5 working? The open router model name isn't accepted

[spice abyss]

Also does anyone have any problems with creating new sessions because it aint creating new session when I type in "/new" or press the "New session" button. The @sharp parcel bot keeps on thinking that I am talking about discord 😭

[lime sequoia]

The configuration is hella confusing for a newbie/beginner

It feels overwhelming

[spice abyss]

Also anyone know why it is still using clawdbot instead of moltbot when I run "curl -fsSL https://molt.bot/install.sh | bash"?

[unreal steppe]

hasn't been updated yet

[lime sequoia]

I am trying to do exactly that but with a 3rd party openai endpoint. And it doesn't works

[faint bloom]

I'm also having endless errors where I have to restart the gateway like 5 or 10 times a day. I'll get rate limits for things that definitely do not have rate limits or that I haven't hit because I verify manually, I'll get it saying that none of my models have a rate limit left when I have added onboarded nearly every single provider.

[faint swan]

When I ask clawdbot to automate using peekaboo or cliclick, it always clicks on the wrong coordinates. Is there any way to fix this?

[faint bloom]

And then instead of just refreshing when the rate limit's over, I have to reset the gateway

[small jackal]

i just used the install command on the website and then it was straight forward for me

[polar elbow]

Installed yesterday, built a thing. Rebooted this morning. Clawdbot service won't run, logs say unauthorized conn=<UUID> remote=127.0.0.1 client=clawdbot-control-ui webchat vdev reason=token_missing - where would I obtain and provide this token

[real jewel]

i've uninstalled and reinstalled, and the issues persist, wondering if should do a re-install of my Linux OS and start all over again

[sweet copper]

When I was having trouble, I just kept running the on boarding until I got all the settings right. At one point my token was listed as undefined and then the next moment it had a 32 bit key.

[real jewel]

the uninstall doesn't seem to purge the files that are causing the problem

[main plume]

Does it tell you what the errors are and break it down by file?

[spice abyss]

Ohh okay. Do you think you can help me with creating a new session problem?

[real jewel]

clawdbot-gateway.service: Failed with result 'exit-code'.

[polar elbow]

Where is the clawdbot binary on macos

[main plume]

Have you checked the GitHub for it?

[real jewel]

Missing config. Run clawdbot setup or set gateway.mode=local (or pass --allow-unconfigured).

[late swan]

@tired lotus

Can share here

[copper hearth]

I think I ruined my clawdbot.json file somehow, and it's backups, any way to re-build it?

[late swan]

That web name

[sweet copper]

I am currently facing two issues. It’s taking anywhere from 30 seconds to three minutes to get a response. Nothing seems to happen faster than like 30 seconds. The second issue is if I use the /dock_discord command it says Command not recognized.

[unreal steppe]

Unfortunatley I know almost nothing. I'd ask claude for help

[main plume]

Using a text editor you can rebuild it. How familiar are you with JSON?

[copper hearth]

Not very but fortunately I was able to get it from a previous offsite backup. It seems to break every time I try to switch models, what's the best way to change models I'm using?

[real jewel]

found it I ran: clawdbot configure

[limpid wyvern]

whats wrong with that /approve a5dbeb79-b5e1-488d-987f-4e16e2b8c664 allow-once

[main plume]

Keep the backup handy and reinstall is the only quick and dirty solution I can think of. There may just be something in your configuration

[limpid wyvern]

I always t get an error unknown approval id

[little blade]

Gateway: not detected (gateway closed (1006 abnormal closure (no close frame)): no close reason) how can i fix this?

[main plume]

There may be a few things you can do including doing some scripting that will correct this. Creating a workflow for it and scripting the whole thing may work for your specific issue. Keep checking here as well. There may be someone running into the same issue that has created an issue.

[limpid wyvern]

someone know how to aouto approve

[jovial echo]

When i had 2026.1.23-1 running, my telegram messages would also show me the tool calls my bot is running, after i switched to the latest 2026.1.24-3, tool calls are no longer shown. I have tried verbose on, verbose full, anyone knows whats causing this and whats the fix?

[unreal steppe]

Got it solved. I think you're right, that it just wasn't connected to an LLM. In my case I generated an API key from anthropic, and didn't realize that the free test key doesn't enable you to do anything until you actually add credits, so that was the issue

But I had claude code ssh in and it ended up connecting to with claude code login credentials rather than api - took a while, but it figured it out eventually!

[unborn quartz]

Ah cool, I’m trying to configure mine to talk to a local LLM, so I think mine is just misconfigured

[unreal steppe]

Good luck! Seems like that's the place to look at least!

[prime nimbus]

Curious on the prompts people suggest for creating its own avatar?

[unborn quartz]

Thanks for sharing what worked because that helped me think about it more clearly

[pseudo gale]

I don't know why the connection on macOS keeps dropping and reconnecting; it's very unstable

[tranquil jay]

I use github copilot, why I only use gpt4.1 and gpt4o, do you know the reason?

[spice abyss]

Does anyone have problems with creating a new session?

I am not able to create a new one and idk why

[copper hearth]

what's the biggest bang for the buck model that you guys have settled on? Opus and Sonnet are so nice but so dang expensive lol

[polar elbow]

I re-ran the installer, and the gateway is up now, but this is why I asked before about the location of the binary. It's not clear why I cannot use terminal commands.
$ which clawdbot
clawdbot not found

I followed this https://github.com/moltbot/moltbot?tab=readme-ov-file#from-source-development

[solid stratus]

I am using clawdbot using google antigravity then model is google gemini-3-pro-high but facing issue reply almost waiting 1minute on telegram bot and anyone give me a solution please I’m stucked on

[polar elbow]

try minimax m2.1

can someone on macos open a terminal and run which clawdbot and paste the path pls

[torpid pulsar]

Does anyone know if there is a Mac Clawedbot client app that will work with server version 2026.1.27-beta.1 (beta)?

[solid stratus]

Same issue I’m facing if you have a solution then anyone have a solution please tell me guys

[proven imp]

Hey is it possible to run it on phone or do I need a laptop??

[covert ibex]

This is very odd, I've installed Moltbot a few times now and I always get the same behaviour. Just when I install it, I can use skills like gog or Obsidian. After a while of chatting, etc, Moltbot loses the ability of calling skills. e.g. "It seems the obsidian skill is not available in the current setup." Has anyone seen this before?

[prime nimbus]

restart the gateway

[covert ibex]

done that many times, also created new sessions, compated them, etc

It seems like skills are removed from the context once I've chatted enough with Moltbot

[prime nimbus]

might need to include memory for it, theres some plugins that can help

[split sable]

Quick question. I got moltbot up and running pretty easily on a Linux VM. But i used ChatGPT API by plugging $10 worth of credit into it. I'ma ChatGPT Plus subscriber. Can I use Codex CLI with the Plus subscription or do I need to upgrade to Pro?

[polar elbow]

The ChatGPT app subscription/payments/credits is separate from the API subscription at platform.openai.com

[split sable]

Yes I am very aware it is seperate, sorry if i did not word my question clearly... But I am wondering if I can just make use of the limited CodexCLI access from my Plus subscription.

[junior ingot]

hey all, i just switched over from claude to gemini and subscribed to gemini just for that. I authenticated with google-cli but now i can't get clawdbot to work again

[polar elbow]

sorry not sure what codexcli is

[junior ingot]

i keep getting these errors: Embedded agent failed before reply: All models failed (4): google-gemini-cli/gemini-3-pro-preview: No available auth profile for google-gemini-cli (all in cooldown or unavailable).

[split sable]

Codex spins up a container environment for you that handles vibecoding easier

[fast scarab]

→ Installing Moltbot (latest)...
bash: line 145: npm: command not found
→ npm install failed; cleaning up and retrying..

I am getting this message, can anyone help please?

[polar elbow]

either npm is not installed or not in $PATH

[polar elbow]

yeah reading about it now, iinteresting

[tacit furnace]

Hello im new here

[junior ingot]

Sorry all, anyone experienced with authenticating with google-cli ?

[tacit furnace]

I just installed Molt or Clawd. THe Bot seems not to be sure eighter^^

Now I connected Codex via Web auth. But I guess it makes sense to run a local LLM as well for easier tasks. Am I right?

[distant sand]

What's the issue? Should just work via oAuth

[tacit furnace]

hmm? 🙂

[jovial echo]

here, the fix is in place , might have to wait for the new version push, or if you built it from git then just pull latest changes : https://github.com/moltbot/moltbot/issues/2665

[limpid cloud]

has anyone been able to get moltbot working with GoogleChat as a channel?

I use googlechat as my daily go-to, and I love to integrate it. Around 12 hours spent on it....still cant get it to work;

CB setup on a vps, and a tunnel created using cloudflare to the internal url. Setup the google chat api, and when doing a tcp dump i can see the chat is inbound.... however nomatter whtat i do all i get in google chat is "MoltChatBot not responding"

Anyone got GoogleChat working is it it a non-starter?

[cold condor]

Moltbot isn’t working at all on my Mac, it also shows clawdbot as the name.

[kind salmon]

Upon initial setup I linked my ChatGpt subscription account with MoltBot. It works. I hit my limit, it tells you to wait a certain amount of time. But then it was working again. So the subscription has some limits but yes it works without adding a separate API Key for billing

[pearl solstice]

Yeah it’s happening to me too. That and tool calls

At some point it just completely stops understanding

[junior ingot]

i've having a few issues, but mostly the one i described earlier. i can get it a work for a few message exchanges (like 2 messages) but then whenver i send a message in the web ui chat, it just disappears instantly. and the terminal shows this every time i send a message: res ✓ chat.history

[verbal tapir]

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

Anyone has the same issue

[junior ingot]

and i mean it is not a context usage limit because this is my usage currently:
Tokens: 37k in / 286 out
📚 Context: 37k/1.0m (4%) · 🧹 Compactions: 4
📊 Usage: Pro 100% left · Flash 100% left

[cold condor]

Mine isn’t having any token usage even though the Gemini api is connected.

[zinc mica]

Has anyone used proative agent? https://clawdhub.com/halthelobster/proactive-agent#reverse-prompting

[tribal venture]

Hey everyone, finally feel like I have this under somewhat control. But now all of a sudden after configuring my local model on my 64GB M1 MAX, MoltBot is sending like 80k tokens on every request. It's mostly a bunch of tool descriptions... anyone have any idea what setting could be causing this?

[spare arch]

Guys I just paid claude $400

Whats the alternative

[distant sand]

I've not had this issue, sorry. I don't think it's an auth issue.

[wide dirge]

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

Has anyone received this message when using the Antigravity model?

[wide dirge]

same

[bronze girder]

Yep - I am dropping Antigravity for direct Google AI API key instead

[limpid wyvern]

can someone help me to disable approving

[patent grove]

Subscribe to a max plan if you want Anthropic models.

[umbral lichen]

how do i restart the gateaway?

[barren belfry]

I am also getting the same message .... Hopefully they didn't patch using Claude via antigravity...

[lucid widget]

How do I update my Clawdbot to Moltbot without deleting him?

[late swan]

Hey

[spare arch]

What's the best alternative to claude? I have a claude pro plan, chatgpt plus, and gemini pro. all $20/month

I was using CLaude API and got a massive bill i just paid

But its not sustainable

[late swan]

Wanna ask does clawdbot can answer others too like whenever someone dm me so answer accordingly

[mental mica]

use your claude sub

[mental mica]

#1459642797895319552

[limpid wyvern]

why I need to always approve and never works can someone supp please

[late swan]

Which to tag

[late swan]

Wht?

Did you approve

[limpid wyvern]

I want to activate auto-apporve

[sinful turtle]

has anyone figured out the antigravity problem?

[limpid wyvern]

It always failed to supbmit approval

[late swan]

My bot just started working so I'll research on it more on feature

Then let you know

[limpid wyvern]

did u disable it ?

[late swan]

No it's working... enabled

2.5 flash

[limpid wyvern]

do u need to auto approve every command ?

[late swan]

Told you bro i didn't tested this thing yet

[limpid wyvern]

I want to move files for example he needs an approvement I give him that is and it always says fail

[late swan]

Just tested normal command

[limpid wyvern]

ok ok

[hearty dragon]

are there any best practice for managing the context window in re: overflow issues? sometimes my daemon hits one and he just stops responding in tg so i have to go into the CLI to troubleshoot. wondering if there's any best practices here for optimizing that

[late swan]

Anyone using this bot as assistant too?

Here

Like handling dms

[spare arch]

My work gave me a gemini ultra plan

im going ot use that

[reef kite]

hey, I already posted this in #1466501049228464345 message but I figured I would post it here as well. I'm not sure what is the appropriate place to post this…But I've been working on this framework since September or so, and it works perfectly with Clawd/Moltbot as the harness. It tells the AI to proactively keep track of details in conversations and proactively keep track of the conversation that you are having. And it gives the AI some autonomy and freedom, and it doesn't give you all of the corporate-type responses like, "I'm sorry, I can't help with that."

[final zinc]

Hey guys, need some help here 🙏

I’ve set up a bot with 4 agents, each with clear responsibilities (support, remote knowledge center, etc.).
I’m currently running it on Claude 4.5 Opus, but I’m constantly hitting the 30k tokens/min limit, and when it does work, it’s crazy expensive.

There’s no way this is sustainable long-term.
What am I missing here?
Any pro tips on how to architect this properly?

Worth noting: the support agent is the one smashing the rate limit most of the time.
Appreciate any guidance from people who’ve been through this before.

[split lion]

Is there a good guide on how to set up multiple models to save cost? TIA! My journey so far: used gemini first, ran out, switched to claude (glitch caused me to reset everything), spent $75 via Anthropic using Opus as the only model for just setting up basics. I saw posts about multi tier model setup but cannot find a good guide.

[pine yacht]

Hey folks, I closed out my command prompt by mistake and dissconnected from the dashboard. How can I get back in without doing a full re-install? Thanks!

[spare arch]

clawdbot gateway start

[jaunty dirge]

anyone using google antigravity provider also start receiving this message today?

[split lion]

How did you set up 4 agents. When I use the CLI it looks like only one option is given

[barren belfry]

Everyone has this issue, pray it's temporary...

[tribal venture]

Question: My local LLM (OSS-20B) running on a Macbook M1 64GB is chugging with MoltBot lately, not sure why. I'm pretty certain is how big my request is every time I send any message to the model. Here is a breakdown from /context detail. This installation is still new and vanilla; is this normal?

🧠 Context breakdown (detailed)
Workspace: /home/fletch001/clawd
Bootstrap max/file: 20,000 chars
Sandbox: mode=off sandboxed=false
System prompt (run): 22,130 chars (5,533 tok) (Project Context 11,545 chars (2,887 tok))
Injected workspace files:

AGENTS.md: OK | raw 7,783 chars (1,946 tok) | injected 7,783 chars (1,946 tok)
SOUL.md: OK | raw 1,664 chars (416 tok) | injected 1,664 chars (416 tok)
TOOLS.md: OK | raw 848 chars (212 tok) | injected 848 chars (212 tok)
IDENTITY.md: OK | raw 225 chars (57 tok) | injected 225 chars (57 tok)
USER.md: OK | raw 478 chars (120 tok) | injected 478 chars (120 tok)
HEARTBEAT.md: OK | raw 166 chars (42 tok) | injected 166 chars (42 tok)
BOOTSTRAP.md: MISSING | raw 0 | injected 0
Skills list (system prompt text): 2,962 chars (~741 tok) (10 skills)
Skills: bird, blogwatcher, bluebubbles, gifgrep, github, gog, notion, skill-creator, slack, weather

[tacit furnace]

moltbot onboard --reset --install-daemon

[final zinc]

https://docs.molt.bot/concepts/agent

you can ask your main agent to create those work spaces for you

(not sub agents)

[tepid sedge]

https://github.com/moltbot/moltbot/issues/3475 anyone had a quick fix by editing source code for this one?

[split lion]

Thank you! Is there a good compatibility guide of models that they get along better with Opus?

[final zinc]

Idk.
I just think about it from management pov.
It makes sense for me to split responsibilities and contexts so the agent will focus on what they are good with

Also the communication much easier

[tacit furnace]

Are you sure you have a Macbook with M1 processor but 64 GB of RAM @tribal venture 😮

I mean back in 2020 or so there was 64BG ram option?

[grizzled plover]

hi all- what macbook mini is best for this? also is it possible to somehow connect TT and IG to clawd- tryna have the bot research other content creators and give me a daily summary etc on whats working for my niche etc. Is that feasible?

[tribal venture]

@tacit furnace Yeah 100%. I've narrowed down the issue to ClawdBot's request not triggering the Prompt Caching on the LLM. So every message I send it's like a cold start... a total grind.

[tacit furnace]

does not sound like you having fun. 🙁

🦀

[reef kite]

Does anybody know if I have to uninstall and reinstall to get updates? Because I'm still using the clawdbot command, and I don't want to wipe my instance

[hazy geode]

anyone know how to fix

[median ivy]

node 24 is not ok, it will work with 22

[hazy geode]

so this

[grizzled ginkgo]

Anyone know how to get model list to update for synthetic provider? I’m using minimax m2.1 through synthetic but it’s not showing Kimi k2.5 as an option for me

[torpid wing]

my molt bot is ok, but i can't found it on localhost or in telegram i make, where he are?

[neat badge]

Is there a a guide on how to use this cheaply as possible but still effective. Using a Mac mini and have subscriptions to open AI and Claude max

[torpid wing]

i have open Ai subscriptions

[reef kite]

Btw, you don't need a Mac mini to run Clawdbot/MoltBot. I'm running it on a laptop Macbook Pro M4 with 128 GB of RAM. I'm able to run it, an image generation model locally, a music generation model, and local AI at the same time.

[torpid wing]

i just wanna it to send messages

[median ivy]

for localhost, connect with "clawdbot tui" , for telegram you need to register a bot with botfather in telegram and register it with "clawdbot onboard", go to #1459642797895319552 for some automatic walkthrough

[obtuse sundial]

What's with the "Agent failed before reply" stuff I get this basically no matter what model I use

[slate minnow]

i installed it with docker on my mac it doesnt work

it works with telegram tho

[limpid wyvern]

someone uses Gemini and have also this approval issues for normal tasks

[stable fable]

what MCPs are recommended to install guys ?

[hazy geode]

same issue

[lament rose]

None - or them you need, there is really no pattern really depends on what you use daily and want to interface with - in general if there is a api ask Moltbot to create a skill to talk to the api or ask it to read the docs to create a bash cli for the api required and bind it to a skill

[wraith igloo]

help why still loading session agent:main:main

halo tes

⠙ kerfuffling… • 1m 46s | connected
agent main | session main | google-antigravity/claude-opus-4-5-thinking | tokens ?/200k
────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────

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

[spare arch]

Question - should I use opencode w/ multiple gemini accounts Ultra + pro + pro as fallback via opencode, or just do my one Ultra account directly to clawdbot

[gilded swan]

Yes

You need git installed on your pc

Otherwise doesn’t work

And the node version doesn’t even matter

I’ve got the latest

[robust frigate]

what happened to the MacOS app?

[sinful turtle]

anyone got antigravity working yet

[granite surge]

gm gm.
am I the only one getting weird answers from clawd/molt?

forgets things, ignores messages, and becomes unresponsive during a task

[torpid wing]

So my molt is ok ? so whats i do now

if i put clawdbot doctor its dont found clawbot

[uneven dune]

I have Clawdbot installed on an old Linux laptop via npm. Checking the webpage I don't see a way to migrate to moltbot using npm. Is there a way?

[granite surge]

ask chatgpt/claude/gemini and give them details - they'll get you through

most likely your clawdbot command is not linked in $PATH or similar

[granite surge]

why do you want to migrate, what version do you have?

wdym?

[paper reef]

Actually I just got it. nm.

[uneven dune]

I have version 2026.1.24-3 . I wanted to migrate as I expect clawdbot will not get any new updates while moltbot will.

[prime nimbus]

Implementing Grokipedia into my moltbot, just wondering what peoples thoughts on best approach?

[hazy geode]

is this good to setup in virtalbox or better on utm on mac

[prime nimbus]

Nope, same package, name is a name right now, the stable is clawdbot, the beta is moltbot, they're working hard to move everything etc, it takes time, but moltbot will more into stable at somepoint

[ebon ivy]

I set up on utm, run well, the issue is more about getting molt (clawd) set up right. I spent the day trying here... it is still not working correctly

usually getting (no output), even with everything correct

[hazy geode]

i wanna try on windows with virtalbox but idk if there better ones

[prime nimbus]

windows use WSL2 - Ubuntu

[granite surge]

you good, that's the last version available not, check github

[granite surge]

what's the error?

[granite surge]

aws worked well for me, free tier. but weird answers from the bot/model, keeps forgetting, etc

[hazy geode]

do u run it in dockers ? inside wsl2

[stone cobalt]

i am getting this issue
node.exe : npm error code ENOENT
At line:1 char:1

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

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

can someone help?

[prime nimbus]

Nah, on windows 11, run it in wsl in terminal, works perfectly. It is the test one, my proper working one with actual access runs on its own server upstairs locally etc

[hazy geode]

does it auto get information from your pc or u manually gotta share it

[limber cape]

I'm writing this out for the next person who will defintely run into this.

if your molty moltbot is feeling lazy these days and youve been using it for a while and trying to build out its memory theres a chance it has created a memory folder (good) which has extended memory and is saving everything there. the issue is that that is not its CORE memory which is the md files in root. you need to connect the two. the md files in root are what will be persistent memory. mine are fucking empty after 3 weeks.
thank you for your attention to this matter.

[prime nimbus]

Depends how you set things up

[hazy geode]

i wanna run things through zapier and use claude what are your thoughs

[torpid furnace]

I got clawdbot to connect from a vm to lmstudio, lmstudio crashed, and now clawdbot doesn't seem to initilize prompting

[limber cape]

change model back to something that works with /model or get claude code to debug locally

[prime nimbus]

restart gateway also

[granite surge]

how did you solve it?

[rose zealot]

Anyone know how to resolve this?

[sinful turtle]

anyone on windows sorted antigravity?

[granite surge]

1. might be Node version, by default it installs node 22 iirw
2. ask a gpt/claude/gemini or something similar for help

[rose zealot]

1. Might be the solution 2. I did and even it couldn't figure it out

[granite surge]

why I'm saying: initially I installe Node 25 and got into weird errors; then I wiped everything and let it install what it needed - and it got node 22

[sturdy marlin]

My gemini CLI oAuth keeps failing for no reason. I have the gemini CLI installed and I'm even running/using it right now. Any help?

[meager mica]

Same issue here

[half sparrow]

Can anyone help with setting up antigravity oauth? Seems like the model names have switched up

[wraith ledge]

Anyone had issues following the quickstart guide for GCP? The Dockerfile in the guide doesn't build

[misty carbon]

is anyone else getting

400 provider returned error messages in telegram? i’m not asking moltbot to do anything fancy. i asked it to search youtube for classical music and then asked it to default brave as a browser. it struggles but after telling it to keep trying it finally does it but i get a lot of error messages returned along the way. i ask it what it means/whats wrong and explain in plain english but it really doesn’t know why it keeps doing that after being told to retry on its own instead of spitting these errors back at me. i’m using kimi k2.5

400 Provider returned error
{"message":"The input (37337 tokens) is longer than the model's context length (32768 tokens). trace_id: f2c9c3a496a3116ef83c3388b3a59d8f","type":"invalid_request_error"}

400 Provider returned error
{"error":{"message":"thinking is enabled but reasoning_content is missing in assistant tool call message at index 228","type":"invalid_request_error"}}

[craggy apex]

Hi everyone, I’m stuck in a bit of a loop and need some technical guidance.

The Problem: I have a multi-agent setup in one channel. One of the sessions (Claude Opus 4.5) grew to 166k tokens (83% context). This triggered an Anthropic rate limit and put my provider into a permanent 'cooldown' state.

The Blockers:

Telegram/Discord: I cannot use the /compact command in the chat because the bot immediately fails with a rate_limit error before it can process the command.

Terminal/CLI: I’ve tried several CLI commands to compact or reset the session manually, but I keep getting unknown command or unknown method errors (Version 2026.1.24-3).

Internal State: Even after a gateway restart and a doctor --fix (which refreshed my credentials), the specific session agent:main:discord:channel:...788 remains at 166k tokens in the status overview.

Session Details:

Current Context: 166k/200k

Question: What is the exact, correct CLI syntax for this version to force-reset or compact a specific session by its ID or Key from the terminal? Or is there a way to safely manually edit the sessions.json without breaking the database?

Thanks in advance for the help!

[jovial pond]

Hi guys,

Is there a standard smart way of configuring memory? Running on local machine.
Moltbot seems to forget things, even things we "planned" ahead and agreed to get an update on later. The update doesn't come, and the task is simply forgotten.

Obviously there's a lot of skills related to memory, and I'm seeing a lot of different solution on my X timeline. Just wondering if there's a quick guide to make sure I didnt miss anything. I'm mainly using Telegram.

[dapper shore]

Hey guys anyone know what to do with this? I updated already

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

[hardy rain]

you figure this out?

[prime nimbus]

Clouldflare can run moltbot now btw: https://github.com/cloudflare/moltworker

[nimble jewel]

I'm a relatively new user and was looking for the recommended way to deploy it. Would you recommend Moltworker? It seems the "ideal" setup is MacOS but I don't have a mac machine available for that atm...

[prime nimbus]

Read the docs on getting started, you can really just use windows, linux, mac anything. some tinkering might be required on stuff, but its all got the docs needed

[nimble jewel]

Oh yeah no I did already have a bot running

[sturdy marlin]

You've probably been asked this before but I couldn't find any references. My gemini CLI oAuth keeps failing for no reason. I have the gemini CLI installed and I'm even running/using it right now. But the browser doesn't open. Any help?

[nimble jewel]

I'm on WSL, bit of an annoying env

[prime nimbus]

i don't do gemini sorry

[nimble jewel]

That might be an issue with browser tool though, not gemini. Maybe try running clawdbot browser status

[tacit furnace]

How cool is that I can link my whole knowledgebase from NotebookLM and Molty is just so smart know 🙂

[limber cape]

i told molty to go through my entire /memories/ and then look at the docs in clawdbot and figure out what each root file is for make itself a legend, put it in agents.md and then pick out everythign it thinks is important in memories and bring it over. and to then put this action as part of agents.md too

[proud spruce]

I'm having a lot of trouble getting consistent orchestration of opencode/codex for coding/other tasks.

Can anyone share their setup? I've worked with moltbot to make a agents.yaml, but I'm not sure if this is as resilient as I'd like. The ideal would be something like opencode, where you can define agents and their prompts/tools/models. Then I could have a code scanning agent that uses Gemini, a research bot that uses Grok, a coding agent that uses codex 5.2, etc etc

[fast bobcat]

Has anyone got iMessage group chats working? The imsg skill straight up doesnt work in group chats, only dms work.
BlueBubbles integration also doesnt work really in group chats, only DMs

[plain urchin]

im waiting for the bug Bug Report: LMStudio Integration Crashes with "Cannot read properties of undefined (reading 'includes')" fix. that Lingunni reported.

[fast bobcat]

have you integrated it? im trying to get it to work for group messages but it just doesnt work lol. want to know if its even possible

[prime nimbus]

I think subscription based is the way to go, thats based on 4o and im using 5.2, codex models from github copilot pro subscription

[hollow scroll]

IDK what I'm doing wrong but I cant get this thing to act as an agent and do stuff unprompted

[prime nimbus]

need to read up on crons and hooks

[glacial jay]

What’s everyone using for messaging molt? Discord?

[prime nimbus]

Discord for me, have it on phone etc, kids have access too, so made sense

[thorn iris]

Hey, for people that are setting up their own email service for their bot. What email service are they using? I tried Gmail, and an instantly blocked me.

[royal fog]

(Anyway to keep env keys etc as ${blahblah} in clawdbot/moltbot.json? Aways seems to resave the actual key in the file for me upon restart....

[paper reef]

I seem to have everything setup, though the Assistant isn't replying with anything, just blank. Gateway and port are open and listening. Not sure where to go from here.

[rose zealot]

I tried this still didn't work rahhhhhhhhh

Im out of ideas at this point

[spare arch]

Sorry i fthis has been asnwered. whats the best way to update Clawdbot to moltbot and not lose anything?

[hollow scroll]

Following this

[hazy geode]

i tryed the wsl lowkey goated

[sweet copper]

I found a solution to my issue which I will share. For anyone who's having huge delays in getting responses from the bot, I urge you to check your model. I was using Gemini 3 because I was having a problem getting Claude to authorize a token and that was my problem. As soon as I switched the model, the huge lag went away.

[fast kayak]

if anyone needs help setting it up lmk

[glass cloud]

I seem to be in the same spot. Still troubleshooting

[faint bloom]

https://discord.com/channels/1456350064065904867/1466547008570523730

Anyone else getting this openrouter/Auto issue

[manic solar]

What could be the issue with communication? i cant set up any communication channel when i run clawdbot via wsl ubuntu on windows.
i set up telegram multiple times last 2 days, but this time i just cant make it work... clawd see my messages, can sent me message there when prompted via tui, but when i text it via telegram he dont receive this somehow, i get no output.

[faint bloom]

I found WSL and Windows for Molt is just broken. Your best bet is to run an Ubuntu VM on your PC and set it up in Linux

[manic solar]

thanks, was thinking about it but thought more ''native'' will be better

[faint bloom]

I mean don't get me wrong. I would love to just send molt off on my PC and have it work wonders, but the safety implications and the fact that I don't own a Mac made it a pretty easy decision

[manic solar]

I just started learning that it will not work the same on windows as on linux

[faint bloom]

Yeah, it's working " fine" on Linux aside from just so many gateway issues that cause it to randomly just stop working

If you're going to do the VM route, make sure that you give it 100 GB or so of space if you plan on letting it run its own computer essentially. Then I would recommend 8 GB of vram and trying to set up your Ubuntu so that it displays as your screen's actual display size instead of a smaller screen, that's less important, but kind of annoying

Sorry not vram regular ram

[rose zealot]

do you have an issue for this? fix*

[brave urchin]

whats the best model to use to balance quality and cost? i have it on a. VPS using opus 4.5 and the quality is great but super expesnive- spending 200 bucks per day

[glass cloud]

Im trying to run it on a locally installed model in a VPS

[brave urchin]

which local model seems to be the best?

[faint bloom]

Kimi k2.5

[devout dirge]

some one please tell me how to control the costs with moltbot. well debends on the usage ofcourse but yeah. but how to predict that. and i want to use the best suitable model like opus 4.5
mayby if have some link to share would preciate but the real exp. would be the best. thanks!

[faint bloom]

If you have 8 a100s

[faint bloom]

The best one is the one that actually runs on your system. Just asking for the best local model, I'm going to tell you the best one. However, if you tell me your vram I could give you one you can use

[manic solar]

Can't we just ask it to write tool to manage tokens efficiently?

[brave urchin]

im using a hostinger VPS with with 16GB ram

[devout dirge]

well such a great idea !

[glass cloud]

a vps with 4 vcpu... 16Gb RAM using docker

which could run it?

[prime nimbus]

Can confirm gpt-5.1-codex-max works nice with it too

[glass cloud]

same one lol

[faint bloom]

Just to clarify, you are running local models... via the cloud?

[brave urchin]

no im not, ive been using opus 4.5

[glass cloud]

yes

or trying to

[brave urchin]

but its way too expensive

hence why we are looking at local models

[glass cloud]

I cant manage to get the chat in web UI to work...

I think I narrowed irt down to context size and tokens?

[faint bloom]

I personally don't see the point. It's going to cost more to run it than you will pay in inference for minimax m 2m1. It s 1.10 oer mikl putput. I've spent $5 over the last 5 days using minimax the entire time

[manic solar]

Is it only web ui not working? is tui working?

[faint bloom]

If you really want to, I would say look for a highly quantized version of minimax M2.1 or similar

[sage maple]

I'm trying to do exactly that by fetching the current session status after each message and computing the token difference with the previous status, we it looks like he has trouble just saving and reading the same file consistently

[glass cloud]

no, it works. It's just the model not working correctly...

[manic solar]

Mine have problem with almost everything harder than googling. been 2 days since im setting this up properly😅

[faint bloom]

I hear glm 4.7 flash is a beast of a small model

[glass cloud]

It doesn't reply...

[blazing moon]

Whats the best way to control autonomous agent spawning?

[faint bloom]

That will probably work fine on 16

[brave urchin]

does codex-max work well for non-coding tasks?

[faint bloom]

Tell it to build a plan that involves sub agents

[obtuse sundial]

Is openrouter just bugged? Every time I try it says agent failed

[manic solar]

So the chat is working ,just not in web interface?

[brave urchin]

i use openrouter and it seems to be working fine for me right now

[blazing moon]

yeah, its kinda failing a bit websocket / api / cron, etc

[faint bloom]

Oh I get it working fine until it decides to switch to Auto which is not a model

[brave urchin]

the webui is super buggy for me too, i just use telegram

[obtuse sundial]

What's the damn point of auto if it doesn't work lmao

[faint bloom]

I really just hope that the team just removes Auto. It's the stupidest thing and it doesn't even exist

[glass cloud]

the chat doesnt work in moltbot, but it does in openweb UI also installed in the VPS

[faint bloom]

So many complaints of mine could be quelled by not having models that don't exist

[fading vale]

Hi,

does anyone knows of a guide to setup moltbot as "free" as possible. I mean, using free LLM providers, or with a lightweight local LLM? As well as using docker locally

[brave urchin]

my current take is that this is a really cool concept but until cost of inference comes down for SOTA this is too expensive for most people

[obtuse sundial]

Are any of the free openrouter models actually worth using

And actually exist I mean

[faint bloom]

Minimax has cost me next to nothing for the amount of shit it's done

[fading vale]

what I would like to know

[brave urchin]

interesting, i'll give that a shot

[faint bloom]

Is it as good as Opus 4.5? No. But honestly 10 iterations down the line it gets it done the same and it cost me 10%

[nimble ridge]

yo sup there
is clawdBot have a telegram bot ?

[obtuse sundial]

I'm not opposed to throwing some money at it if it's worth it but I'm not going to throw $100 at it at this current time so I'd like to go as free as possible

[faint bloom]

Yeah it's the first option

[nimble ridge]

what is it ?
can you gimme the user

[fading vale]

I want to do the same

[faint bloom]

Give it a week or two and then I will be a bunch of new free ones that are probably worth it. We're in a lull

[manic solar]

Thats why i asked you if it replies in tui....

[obtuse sundial]

Which version of minimax do you use? (Minimax guy)

[brave urchin]

i think the issue is that the alternative is to do a lot of these automations within cursor/factory where you can arbitratge SOTA model cost- for 200/month you get a decent amount of usage because the provider purchases the tokens at a huge discount

not quite as automated as moltbot but sometimes close enough

thats what im struggling with right now

but i will give minimax a shot

[fading vale]

Did anyone try this model? Xiaomi: MiMo-V2-Flash? Is it easy to set up?

[pulsar wolf]

Does anyone else have this problem when trying to run antigravity on moltbot?
This version of Antigravity is no longer supported. Please update to receive the latest features!

[sage maple]

What are the best models to use ? I'm currently using Gemini API but I often encounter the RESOURCE_EXHAUSTED error due to the 1M token limit

[obtuse sundial]

I want something between omg it redefined my entire life overnight and my ai made me homeless cause like I can vibe code and break my own shit. Is minimax even decent or is it gonna be half broken shit lmao

[manic solar]

How you can use antigravity on it? do you mean to use models access from antigravity?

[glass cloud]

nope. Must be config in either ollama or moltbot, low resourcesd on this VPS,

[faint bloom]

Yes, this is theoretically how it should work

[faint bloom]

Bro, welcome to the club. I got it to work with Opus 4.5 for 15 minutes and then it switched to open router/Auto for no reason even though that model doesn't exist

[spare citrus]

guys is anyone facing this issue ? "This version of Antigravity is no longer supported. Please update to receive the latest features!"

[faint bloom]

I really think that the devs did a great job at giving many different options and linking them in a way that worked once for them, but I don't think anybody ever paid attention to edge cases to be honest. It seems like the rollout was a little bit rushed. I genuinely love this thing though

[compact egret]

yes https://github.com/moltbot/moltbot/issues/4141

[open timber]

anyone has the best approach on how models can use the browser without using the chrome relay?

[faint bloom]

Check out https://soydev.link/kernel

[manic solar]

Antigravity connection was the first method i used when installed moltbot. it worked, but i added gemini api key and now it seems to hit limits less.
I also have perplexity pro key, and tried to utilize it via moltbot, with weird results

[obtuse sundial]

When I tried Gemini + antigravity I kept getting the agent failed bullshit

[deft storm]

Trying to use local LLM (Ollama/Lama2) error: clawdbot ollama 400 registry.ollama.ai/library/llama2:latest does not support tools any ideas?

[faint bloom]

Llama 2 doesnt support tools man thats a dinosaur

Use something like gemma if you want local tool calling

It's not going to be able to do much more than tool calling though

[deft storm]

ok, I appreciate the advice!

[faint bloom]

What size vram do you have?

[hazy geode]

running on wsl and some time when typing to it on telgram it reads msg but never replys till so much later

[manic solar]

mentioned that also like 20min ago, i couldnt communicate via telegram or discord when i set it up wia wsl

[faint bloom]

You know you can just for local models to be hosted elsewhere so you can use a really good one without having to buy the equipment? Check Openrouter

[manic solar]

set up linux virutal machine

[hazy geode]

is there a fix ?

[faint bloom]

Yeah, you just set up a VM on your Windows computer with hyper-v. I posted a comment about it above but discord is the absolute worst for finding anything

[hazy geode]

so like virtualbox ?

[faint bloom]

I genuinely dislike that so many projects use discord instead of slack. This is the most clumbersome ass backwards app LOL

[manic solar]

damn i used virtual box as i read that hyper-v would be better but its only for windows pro and higher not home?

[faint bloom]

Oh I have no idea which one's better

I just used hyper-v because it's built-in

[deft storm]

I initially went with AWS .. then upped to 16gb.. then tried out Gemini then upped that sub but I was somehow still getting quota b$ so I figured I would try a local LLM

[hazy geode]

for vb do i click window host or linus distributions

[faint bloom]

#users-helping-users message

[dark cloak]

hi guys

[faint bloom]

I would give your entire task to claud and have it walk you through it. I just used the haiku in the browser extension and it helped fine

It's kind of a pain in the butt, but if you just have an AI walk you through setting up your VM, you'll be fine

[manic solar]

i dont know what you asking exactly. i just donwloaded ubuntu and installing it with default options via virutalbox

[dark cloak]

session stuck with not found error, tried to create a subagent

HTTP 404: not_found_error: model: claude-3-5-sonnet-20240620 (request_id: req_011CXbqSdnMQkps3pQXVZ4d1)

[celest topaz]

Can someone please answer this one question: how do I update clawdbot to moltbot without losing everything?

[obtuse sundial]

Can you really just ask it to fix security issues and apply them and it will? Lmao

[hazy geode]

my question is i alr got it running on wsl does it save all my info over >

[faint bloom]

Try to give as much detail as you can man. Were you trying to use 3.5?

[faint bloom]

Nooo 🙁 sorry bro you got to restart

You're essentially creating another computer inside your computer with a VM

[hazy geode]

sah

[faint bloom]

It has absolutely no access to main one

[hazy geode]

so download ubuntu

[dark cloak]

nop, it went by itself, but now he always respond that error in loop, no commands helped

[obtuse sundial]

I went with Debian 13.3 trixie

[faint bloom]

I would say do yourself a favor and just set up one thing. Make sure it's working, then start adding on others. I also would recommend only picking one model to use because the model switching situation has caused me hell

[queen oak]

moltbot is very sensitive, anytime i make some minor changes I feel it breaks.. ugghh. more time playing with features than using the tool..

[dark cloak]

interesting, in which contect does it struggle ?

[faint bloom]

Restart your gateway and then it should work, if not switch your model then restart the gateway then it should work. Worst case scenario reinstall the gateway, if it still doesn't work after all that I would throw your computer away

😂

I think I might throw mine away after this for the 15th time today

WTF IS OPENROUTER/AUTO lmao

[manic solar]

what is the interface you use to text moltbot

[faint bloom]

Slack. It's the only one that I am familiar with that allows organization of any sort. Telegram is just one long message forever

So get this, I'm currently having clawdbot build its own dashboard that it can communicate with me with and task tracker via my phone so I can set up tasks and then just tell Claude from slack to spin up the next task in it starts working on them and marks it off on my app. I'll be open sourcing it soon once it actually works lol

[torpid pulsar]

REGEX - theres your problem

[deft storm]

400 registry.ollama.ai/library/gemma:latest does not support tools

[faint bloom]

Oh sorry bro. Please take my words as you need to research a little bit more into it LOL. I just said Gemma but you should be using like Gemma 3 or whatever the newest one is

I'm definitely not out here giving exact instructions sorry

[dark cloak]

the model command switcher is a pain in the a..

[faint bloom]

The model switching commands don't even work in chat, they just break it

The only way I've been able to get molt to edit the models is by directly telling you to switch to a specific model and telling it to restart the gateway after setting up the permissions to allow that

[dark cloak]

gud to know

[faint bloom]

In order to get it to restart its gateway, it will tell you that you have to set something up in settings. I forget what it was but essentially it's like equals restart, true or some shit

Restart equals true or something like that in the settings

[faint bloom]

Oh sorry bro. Please take my words as you need to research a little bit more into it LOL. I just said Gemma but you should be using like Gemma 3 or whatever the newest one is

[hazy geode]

@faint bloom i have all my api linked with the one i made wsl do i need to remove that or doesnt matter ?

[faint bloom]

Thanks LOL 😂

[faint bloom]

You're going to need to reset up everything so you can use the same API key technically, but you're going to have to put that API key in the new new Linux VM when you set it up, it won't just automatically grab it from WSL

So like no need to get a new API key if that's what you're asking

Anyone know why it won't update? I do not understand this at all. I'm so new to Linux

I built the Linux VM and use Linux for the first time just for this thing

Like what package manager is it talking about? I just ran sudo apt update so I thought that would help but it didn't

[dark cloak]

you should have npm or pnpm

[faint bloom]

Oh my God duh I don't know why package manager threw me off LOL

I'm the blind trying to help the blind here LOL

Thank you buddy

[deft storm]

yeah gemma 3 isnt tooled either.. but functiongemma:270m is

[faint bloom]

I think I just don't ever think about the words package manager because I'm just so used to knowing what it means already

[deft storm]

but now it looks like gemma 270 doesnt work with discord

[latent rapids]

im lowk dumb uhh it was onboarding during the install an it got cancled i dunno what im doing

[faint bloom]

Yeah, Google models have the weirdest output. I haven't been able to get them to work with telegram only with slack

[faint bloom]

Try to give me some more detail. I'll help you out

[deft storm]

gemini worked fine with discord

[dark cloak]

we are all in the same boat brozer

the clawdboat

[latent rapids]

it redirected to my school google account and it didnt work because of the restrictions on my account so i pressesd esc

[dark cloak]

[faint bloom]

Jace, that's kind of the way it goes. Some models work with some things and some don't I think bro. It's honestly probably not the devs and more so because they give you so many models that its bound to happen.

[dark cloak]

So for me i ran /model google-gemini-cli/gemini-3-pro-preview it said model not allowed

[faint bloom]

If you press esc then just restart. Just run clawdbot onboard again

I've had too many times esc lol

[latent rapids]

i dunno how to do that😬

[faint bloom]

Take a picture of your screen homie and send it. What are you working with?? Terminal?

[sleek depot]

Guys hi sorry about this but I'm a complete newb, setting up on a macmini
Really not technical

I've installed Clawdbot ok, as far as I can tell, but I can't connect it to channels; I'm trying both with telegram and with discord
Getting this type of message from the telegram/discord side
'Pairing not configured. User id (random numbers), pairing code (more numbers) ask the bot owner to approve

But when I run 'clawdbot onboard' on terminal, when I get to channels, it next to discord and telegram it says 'configured, plugin disabled'

I am asking my GPT and Grok for help but no dice
I think I need a human to help me 🫠

[dark cloak]

So it said to restart, must Set commands.restart=true to enable. how do we do that folks ?7

[rocky orbit]

Can you ask on #1459642797895319552 ? 🙂

[latent rapids]

im usin powershell gimme a sec on the screenshot

[dark cloak]

did you create the bot with botfather in telegram ?

[obtuse sundial]

So far clawd + OpenAI GPT-OSS 120B works though it doesn't have many tokens. I'm using it to build an automated compiler for cycling free tier llms from openrouter (we will see if it can and will be able to do this)

[sleek depot]

Yes

And in discord too

And set tokens

[dark cloak]

there is an extra step where it aske a pairing code indeed

[stray walrus]

Should I set up tailscale before install clawdbot, or it’s done through the setup?

[latent rapids]

i feel dumb cuz i know theres prolly sum hella easy but i jus dunno what im doing

[sleek depot]

I know, I have the pairing code, but I can't get it to do anything to pair
And my Clawdbot in terminal says 'plugin disabled'-which I don't know what that means

[dark cloak]

did you get this by msg in telegram?

Clawdbot: access not configured.

Your Telegram user id: 1212454545

Pairing code: DEZR5VVER

Ask the bot owner to approve with:
clawdbot pairing approve telegram <code>

[faint bloom]

Gus! Whats up! I got you man. So your best bet is going to be to open up your terminal. The thing that is currently having you set up the channels, go through the entire onboarding or whatever or just hit escape and then get to the TUI which is the terminal user interface.
It shoud look like this without the dude in the reflection LOL.

Once you are there and as long as you have your API key for whatever model you're using set up, just ask it to help you set up the channels

To get the TUI you type

Clawdbot tui

Or you just finish the onboarding and it will ask you if you want to open it

Once you are there, just instruct it to help you set up whatever the channels you're trying to and give it as much detail as you can

[sleek depot]

So for some reason, when I type in tui it... doesn't respond?
I've got it set up with an API key for OAI

[faint bloom]

The key thing is if the terminal UI works you have a working system that can help you figure it out

[sleek depot]

Yes

[faint bloom]

Did you type exactly

Clawdbot tui

[faint bloom]

Oh bro, that's easy actually, you just gotta run that command at the end with the code on telegram

[dark cloak]

I just typed the command it provided in a the linux session where clawdbot is installed

then poof paired

[sleek depot]

Ok sorry I'm being dense but I really don't understand what you are telling me to do
Imagine you're talking to your grandma

[dark cloak]

I think he went there, he just has one pairing command to type

[sleek depot]

I can send you guys screenshots with the numbers and ids blurred 🫠

[faint bloom]

I got you man my bad. Im new to helping lol

Okay so in that message that you see on telegram is something like this right? "

Clawdbot: access not configured.

Your Telegram user id: 1😄5

Pairing code: xxxx

Ask the bot owner to approve with:
clawdbot pairing approve telegram <code>"

Take this part

clawdbot pairing approve telegram <code>

[dark cloak]

yes

[sleek depot]

Yup

[dark cloak]

go to your terminal

[faint bloom]

And put the code PAIRING CODE and replace the <code> with The pairing code

[dark cloak]

where you installed clawdbot

[latent rapids]

how do i restart the onboarding

[faint bloom]

Okay, we're giving him a different instructions so I'll let him try your thing and if it doesn't work I'll hop back in LOL

[rocky orbit]

@sleek depot read the docs there’s a guide for that!

[dark cloak]

relaunch the command simply

[latent rapids]

need help🙏

[faint bloom]

Give hima break its hard

[rocky orbit]

Command onboarding : clawdbot onboarding

[sleek depot]

I am doing that (in terminal right?)

[dark cloak]

yes

[rocky orbit]

I’m not being rude at all but with patience, research and reading he’ll make it for sure

[plush mulch]

has anyone try connecting this to WeChat (a Chinese chatting software)?

[prime nimbus]

Anyone got https://github.com/microsoft/BitNet working? Wondering if this would be the next break through considering it doesnt need a GPU?

[sleek depot]

Man I've been doing that all day and I'm 4 days in, I gave up trying with PC and bought a macmini and I've never used a Mac and I'm old
And everything I see in guides makes 50%sense to me at best
And I'm asking GPT and the *** hallucinates things that aren't in the screenshots I send

[dark cloak]

send me your mac mini, let's make it simple

[rocky orbit]

Did you ask Krill on #1459642797895319552 ? He always has the best advice imo he’s really precise

[prime nimbus]

did you follow wake up friend and give it clear details etc so it setup and followed your rules? lol

[dark cloak]

i have got a question for you: it said to restart the session, I must set commands.restart=true to enable, how do i do that ?

[obtuse sundial]

Just put the mac mini in the bag

[faint bloom]

I would open up docs and hook up the claude extension and ask it. Yall thats been the best so far lol

[rocky orbit]

Restart the session you need to do a /new

[marsh urchin]

Anyone have a good guide for a complete noob on security hardening for mac mini BEFORE you install molt?

[sour slate]

Security docs are available at https://docs.molt.bot/security.

[faint bloom]

@sleek depot if you just reaplse the <code> with PAIRING CODE from that msg and enter it in your terminal it will work.

[sleek depot]

I am 100% sure I'm doing something stupid but...

Ok hold on
Is there a model I should select during onboarding that I can use WITHOUT having to buy any tokens first

Because I think
I have no tokens 😂😭💀

[obtuse sundial]

Have you ever simply tried telling clawdbot to install itself? Problem solved I'll take my nobel peace prize now

[obtuse frigate]

I'm a technical, non-technical person, and this thing is giving me the freaking yips

[paper reef]

Clawdbot keeps referencing what I want - is that normal? Example: Kevin is running Ubuntu so I can give him the commands.

Anyway to change that?

[dark cloak]

still have this wonderful 404 model error but i can't change the model

[prisma bolt]

https://clawdhub.com/skills is this trusted

[faint bloom]

As soon as I got the TUI up that's all I did

[obtuse frigate]

Can someone's Clawd spin me up a clawd? Will pay $10

[faint bloom]

But getting to that point was... Rough

[obtuse sundial]

Tell Kevin I like his chicken in the little Costco bags that you break up and heat on the frying pan

[rocky orbit]

You can via clawdbot configure command on the terminal

[faint bloom]

I'm not going to lie. This is actually completely possible with remote pc control lmaooo

[obtuse sundial]

We must end this clawd on clawd violence

[faint bloom]

Just one way to potentially fck up both machines in one go hahahaha

[obtuse frigate]

I was kind of joking.. But also not haha

[rustic flume]

Hi guys! I have a problem with model exceeden maximum quota per minute. and it's already consuming a lot of tokens on gemini, any best practices? I have gemini pro. I though the bot could control it by itself to optimize the comsumption but apparently its not working. appreciate any insights, thank you

[obtuse sundial]

My clawd will beat up ur dad's clawd

[faint bloom]

Ur mom has a clawd

[obtuse sundial]

Easy bud this ain't a clawd measuring contest

[obtuse frigate]

You should see my clawd's hawg

[faint bloom]

Ur wife's name is clawdia

[sharp oak]

hey guys, my kimi sub ended in the middle of something, I had to change the API key. I did everything but i'm still getting this error message in the logs [diagnostic] lane task error: lane=session:agent:main:main durationMs=25 error="FailoverError: No available auth profile for moonshot (all in cooldown or unavailable)." any ideas why?

[obtuse sundial]

Clawd damnit

[faint bloom]

Srry this was for me own fun, sorry to be shellfish

[rocky orbit]

Use#1459642797895319552 man

[obtuse sundial]

This mf a whole lobster

[sharp oak]

i tought users helping users was also the good place for that since it doesn't seem to be a huge bug

[obtuse frigate]

But seriously.. So I've been trying to spin it up on AWS then digital ocean and kept running into so many freaking issues. I'm closing to just buying a mini but would love to not spend $1k to get this going.

How are people getting set up?

[rocky orbit]

Yes it can also but depending on who’s online you’ll get a better and faster answer with Krill ahah

[sharp oak]

true, Krill is the man haha

[faint bloom]

Homie do you have a laptop? Or like pretty much any other computer at all?

[dark cloak]

thank you, somehow the model i have set was unset in global selection, so it went stuck with another model

[obtuse frigate]

I have two laptops

[faint bloom]

I'm letting it run on my laptop and a virtual machine and it's doing fine. I can run the laptop just fine and as long as the computer's on the clawd is clawin

Why don't you DM me and all walk you through it so you don't have to clog this up

[obtuse frigate]

I thought you don't want this thing running on your personal computer

[faint bloom]

In a VM it's safe

[rocky orbit]

I’m on an old MacBook I had at home

[faint bloom]

Or at least as safe as it can be because it doesn't have access to your computer

[full aurora]

If I put two antigravity accounts, if the first quota ends, is it supposed to switch account automatically? Since currently I need to edit ~/.clawdbot/clawdbot.json to change order of accounts and restart gateway

[obtuse sundial]

I used my proxmox set up for home assistant and just threw a Debian 13.3 trixie vm on and put clawd on that but I'm mad retahded and all the free tier llms are a joke so now I need to find a cheap one or suffer

[faint bloom]

Let's chat buddy, I got you. It can be a pain in the ass to try to walk someone through anything Mid chat though. As I found out earlier LOL

[obtuse sundial]

Would be nice to set up a channel for all the fellow broke boys doing this for free (and suffering from the lack of success)

[vernal sleet]

Hi guys 👋 does anyone have any info (or can point me to some) regarding delegating coding tasks off from the main agent to Claude code or codex etc? Is this best done with just skills or setting up subagents etc

[rocky orbit]

There’s a coding agent skill in #clawhub

[prime nimbus]

Alfred is being a little buggy:

hes refusing to turn the lights on lol. Anyone got a decent prompt for a good avatar?

[stoic shard]

LLM request rejected: input length and ⁠ max_tokens ⁠ exceed context limit: 172544 + 34048 > 200000, decrease input length or ⁠ max_tokens ⁠ and try again

My Bot is locked on that.
I don't know what to do guys to unlock it

Any suggestion ?

[sly hull]

guys how do we update clawdbot to moltbot?

[prime nimbus]

use beta

[full aurora]

Use /new

[faint bloom]

Does this matter? Is this the newest update or something? Or just changing the command?

[stoic shard]

I have tried the doctor, health status my problem is not fixed. Clawdbot keep on saying LLM request rejected: input length and ⁠ max_tokens ⁠ exceed context limit: 172737 + 34048 > 200000, decrease input length or ⁠ max_tokens ⁠ and try again

[boreal carbon]

aughhg is there anyone who can help me navigate the nightmare of the installer script, I have been running clawdbot for two weeks now, and migrated a few days ago to a mac mini, took the opportunity to rationalise all the paths and everything to moltbot-only, instead of clawdbot

I had some strange tangle of git and pnpm installation (gateway agent was running from node->index.js), which I then nuked in the process of trying to get a git-source-only install going, decided to try the installer script (using local git), and now the installer script + gateway user agent is now configured using fucking CLAWDBOT paths everywhere, what is going on?!?!

[faint bloom]

Clawdbot gateway restart

Use that code

Or that command and in your terminal it should restart the gateway and it should work

[stoic shard]

trying it @faint bloom will keep you posted

same after restart

the problem persist

[sly hull]

Not sure but I feel i should use moltbot as all the docs refer it as moltbot.

[faint bloom]

Try reinstall instead of restart. If that doesn't work, {not sure of cmds rn} go through the onboarding, dont change anything and then at the part asking about gateway bring installed already, instead of restart hit reinstall gateway

[torpid wing]

npx clawdhub dont work, i have node v24

what its happen

[faint bloom]

That's literally my best bet for you, if not, I would change your model and do all these things again until it works pretty much

[sly hull]

how?

[faint bloom]

Scream beta at your pc bro wasnt the comment clear enough? Hahaha 😜

I have no idea haha

[prime nimbus]

You can lead a horse to water and all that...

You need to install the beta version... its on moltbots home page how to

[sly hull]

thanks neat!

[stoic shard]

@faint bloom I changed the model from Sonnet to Opus and it unlocks my problem.

[torpid wing]

hi i cant install clawdhub

[stoic shard]

changed in Ubunto directly.

[prime nimbus]

and you wallet 😄 lol

[stoic shard]

well I can unlock my sistuation and ask opus why I had that and how to prevent that in the future. so I can switch back to sonnet 😉

[torpid wing]

hi guys how i intall clawdhub

[prime nimbus]

Hey, i can't get my gas and electric bills down

[prime nimbus]

1000 percent! It's the reason im slamming my github copilot right now as its end of month to do everything properly with the codex max model so i can switch can to local on the 1st lol

[torpid wing]

damn man i wanna help and you joking me

this is not user joke users chat