users-helping-users

[dim flare]

Any one had success in using local LLMs with moltbot?
I don’t mean the setup, but more like the overall experience of using something other than claude/chatgpt etc locally

[solid bison]

My system GPU not powerful enough to process the prompts. takes forever.

[ebon holly]

Tried to set it up throught vps on windows , but UI gives me this . can someone help

[cold horizon]

"LLM error api_error: Internal server error (request_id: req_011CXZh8G3TeUadxwtKGJHBF)" has anyone seen this error too came out of nowhere and and now my clawdbot on my tg is just responding with this, I have checked my anthropic susbcription and it still active i use the max plan

[primal tree]

i just got this error too

[arctic star]

got it too

[woeful thorn]

Same, also on the max-plan with alot of headroom

[primal tree]

mine got that error when i tried to install a skill on molthub.... i told the bot to "stop" and now its working

[coral pumice]

HTTP 429: rate_limit_error. Keep getting this and I’m not able to have a proper conversation. Anyway that I can improve this guys? I’m using the anthropic API with sonnet 4.5

[stray axle]

How do I switch to a different model? Claude is costing too much. Do I do it in web UI or CLI?

[sharp oak]

you can do both ways, but it's easier to do it in the config file

[fresh moon]

hi guys im trying to use the reddit skill but idk how this works it keeps stopping me can anybody let me know what to do?

[wet field]

i simply cannot run this thing. The chrome tab has a conexion error. Anyone knows what to do next?

Im not an expert, im using gemini as my teacher. i dont know if the things it tells me are correct

[lone tangle]

why are whatsapp audios automatically transcribed but not on telegram?

[karmic gale]

Add me to the pile of users getting this:

LLM error api_error: Internal server error (request_id: req_011CXZiZJ8Xy7yirksb9LL8Z)

Before that I was starting new sessions and still getting;

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

And before that:

⚠️ 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) | google/gemini-3-pro-preview: No available auth profile for google (all in cooldown or unavailable). (rate_limit).
Logs: clawdbot logs --follow

[crimson steeple]

everyone experiencing this huh

me too

[static blaze]

i'm looking for the same thing 🙂

[signal tartan]

The easiest way is to create an alias for the model. Then, in your DM, just type:
/model {alias name} (for example: /model glm).

[outer canyon]

is there like a migration guide in help there is something but links are broken

since the name change cowdbot is all consued sending json files i tried to update him and he did but still thinks its clawdbot and stuff

[steel steeple]

I have the same issue and I've arlready completed that walk trough. I don't think i want to repeat that again, so is there any other option to set that token?

[crimson steeple]

messy transfer atm huh, I'd avoid updating a few days

[signal tartan]

A lot depends on which model you’re using. Not every model understands your intent and what you’re trying to achieve. You can clearly feel the difference between models like Minimax or GLM on one side and Opus or KIMI on the other.
You can try to be more explicit in agents.md — or even better in tool.md — for example by stating that whenever something needs to be changed, a task/job must always be created.
That should really only be necessary when you’re working with a less capable model.

[tardy veldt]

Hi, in my webui all skills looks like this, blocked. Do I have to unblock them one by one or can I unblock all skills somehow?

[wet oxide]

Is it easy to upgrade to moltbot from clawdbot? Does anything break?

[slender pewter]

are you guys having success with using local llm with clawdbot? i set it up, but then it simply don't respond, the gateway says "it is loaded but not running"

[gleaming palm]

how is it looking with the security breach that has happend i hope i am allowed to ask open like this ? was it fixed ? does anyone have it running like on a MacBook Air like not Mac Mini ?

[normal skiff]

i think its the name changed because of copyrights from claude

[normal skiff]

which one you are using? Ollama?

[slender pewter]

yeah, i tried a few different models like dolphin/oss/gwen coder

it is important to notice it didnt work both as clawd installed in a vm or on my host system, but the vm terminal was able to run ollama from the host system with no problem

[gleaming palm]

anyone has issue with the npm install failed ?

[subtle oracle]

How many users have gotten their accounts banned by Anthrophic?

[north fulcrum]

i'm having a ton of issues with installations on both windows or linux

what OS are you on?

[placid python]

Anyone use gemini API key for the bot? Doesn't matter which model I use, gives error 404 or 403

[north fulcrum]

is the site molt.bot even real? i'm afraid i'm installing a fake package

[normal skiff]

Since you're running Clawdbot in a VM but Ollama on host, localhost won't work. In your clawdbot.json, change the baseUrl to your host's actual IP (e.g., http://192.168.x.x:11434). Also check if Ollama is serving on all interfaces: OLLAMA_HOST=0.0.0.0 ollama serve

[floral sand]

It was working for me yesterday.

[sharp sandal]

hello

[gleaming palm]

macos

[placid python]

On free plan or paid? I got the message from goggle I have 300 dollars for free

[normal skiff]

im using chatgpt auth, its working with my chatgpt plan

[placid python]

Normal pro plan? Or the pay per token plan?

[normal skiff]

it will work on any plan

[normal skiff]

it is

[slender pewter]

Yeah, i have followed those steps already. since i couldnt figure out why it was not working, i installed clawdbot on my host system, to test it there was any kind of pack delivery error between the host and the vm, but it is not working on my host system either. and ollama is answering well when i open it in my vm cause i used that command to export ollama (not sure about the details i mostly followed intructions xD)

[placid python]

This should be really nice

[north fulcrum]

when i install on Ubuntu, post install the executable is still clawdbot instead of moltbot, is that normal?

Installation was hard enough, fucking Anthropic made it so much worse now

[neon totem]

do you know why Clawdbot is not able to read it's own memory without open ai keys? I'm getting below response from it :

"It makes sense that it feels odd—those memory files are stored on this machine, and I can read regular files just fine. The restriction comes from policy: access to MEMORY.md is gatekept through the memory_search/memory_get tools, and those tools are deliberately tied to the OpenAI/ChatGPT provider key for security and privacy reasons. Even though the file lives on disk, the tool that coordinates safe reads checks for that key before it lets me look inside."

[zinc juniper]

Hello, can i install clawd on a Linux machine?

[slender pewter]

yes, you can. it is quite easy to be fair

especially if you are not trying to reinvent the wheel

[zinc juniper]

Is there documentation on how to install it on linux?

[slender pewter]

https://clawd.bot/

[north fulcrum]

I wouldn't call it easy. I'm pretty tech savvy and it's definitely not plug and play

[zinc juniper]

Appreciate it.

[slender pewter]

there is a quick start you copy and paste it on terminal

[placid python]

Not scared to get rejected?

[north fulcrum]

I used that, it failed because of missing dependencies. Now it requires Node 24 when the script installs Node 22.. it's a mess

[zinc juniper]

This one right?

[placid python]

Yes

[slender pewter]

really? the installation worked really well for me, when i first set it to use google api, even telegram integration, i only had issues when trying to make it run locally

[zinc juniper]

🥳

[north fulcrum]

And I installed it as Moltbot, but the executable is still Clawdbot?

Is that correct?

[slender pewter]

i think it is simply to avoid having to change all command lines

[feral garnet]

Any tips and tricks you guys recommend after I set everything up

[slender pewter]

imagine if you had been using it for a while as clawdbot and now you would have to use all the commands as moltbot it would be annoying

[north fulcrum]

It's confusing and a bit scary since I heard one of the packages is fake?

[normal skiff]

Can you run clawdbot doctor and share output. Also check if you set the ollama profile as default in clawdbot.json ("default": "ollama:local"). The model name might also need exact match - run ollama list and use that exact name in the config.

[floral sand]

Same plan paid $300 free trial

[floral sand]

Sorry not sure. I’m sure there is a way but I don’t know what it is the best approach if the doctor doesn’t do it and the bot can’t help you through tui

[slender pewter]

simply "clawdbot doctor" ?

[zinc juniper]

am i doing it right?

[slender pewter]

so far, so good

after instalation is complete you need to run "clawdbot onboard" to set it up

[zinc juniper]

right

[zinc juniper]

its taking time

[latent hedge]

You guys ever experiencing that it almost seems as if its 'hanging'? Maybe it's just because Im running a low-performance model

Has been working on this for 6 minutes

[placid python]

Who else is using the chatgpt Auth method instead of pay by token?

[normal skiff]

i am

[normal skiff]

Yes, just clawdbot doctor, it checks your config and shows what's wrong. Add --fix to auto-repair issues: clawdbot doctor --fix

[zinc juniper]

why is it taking so long to install? what's the size of the bot?

[rocky orbit]

It’s not supposed to take long

[zinc juniper]

yeah idk what to do

it says "your task has been queued"

[civic hamlet]

Anyone know why ollama's gpt-oss isnt aware of soul.md or identity.md? if i switch to other models it works fine

[latent hedge]

Does it say this for you as well in the telegram configuration?

[slender pewter]

can you guide me to the exact line where i should add this command to the json?

[prime nimbus]

all need to do is run ollama launch clawdbot ollama walks you through everything

[latent hedge]

Hey guys... is it correct that Molt drops the entire memory, if you start a new session?

I did that, told it to summarize myself and my background, and said it had no memory.. even though I had given it some stuff in another sesion

[idle lichen]

Has anyone ran in to the problem where when attempting to say "hello" to the bot in the gateway dashboard it just response with nothing?

[slender pewter]

But ollama is only installed in my host system, cause I was better to make use of my GPU for processing llm

Would I be able to run it on my VM with that export command? Sorry I'm really layman when it comes to coding and Linux stuff xD

[prime nimbus]

im sure it supports remote ollama... mightbe worth reading the docs, it was dropped yesterday so new feature

[slender pewter]

I will give it a read, thanks!

[prime nimbus]

if you read up in here, some chap pointed the docs early hours this morning, might be worth reading around there when we were talking about it

[slender pewter]

Nice! I will look for it

[prime nimbus]

Search is also good in here, the help channel on the left, returns some amazing help guides

[zinc juniper]

what to do?

[ashen gust]

press enter?

what model do u wanna use?

[obtuse lance]

Has anyone tried turning an old andriod phone (maybe with a cleaner os) into a clawdbot?

[zinc juniper]

which one is free?

[prisma bolt]

so im in claudbot, what can i do?

[ashen gust]

i think u have to pay for all of them considering that u will use tokens/api calls

[zinc juniper]

which one do you recommend then?

[ashen gust]

u can actually ask it

minimax and glm are recommended now

[zinc juniper]

okay

[ashen gust]

do u have any pro accounts in chatgpt or claude?

or google?

[zinc juniper]

no

[ashen gust]

not sure but i actually think that googles ai studio has a free tier

[prime nimbus]

if youre asking questions at the first screen, you;ve not followed anything from the docs: https://docs.molt.bot/start/getting-started try 5 minutes reading, you'll get up and running faster

[zinc juniper]

im not planning on using this bot for the long term, i actually have to vibe code an app

[untold bronze]

Any reason to migrate clawdbot migrate molt if everything is still working?

[ashen gust]

what do u use for vibe-coding?

[zinc juniper]

for discord apps

[ashen gust]

i mean what llm

or tool

[untold bronze]

opus 4.5

[ashen gust]

so u have claude code then or what

[untold bronze]

yes

but clawdbot is the UX UI I use for my agents

I dont touch Claude own client right

so Im worried if I update / migrate shit breaks and Im gonna be unproductive next hours days

But is there like a security risk or other up/downside to migrate to molt

you know one thing Ive learned, if it aint broken, dont fix it lol

At least Im thinking to wait a couple days before migrating is probably EV+

[unreal bear]

Does anyone else have to complete 10 approvals for everything that they tell clawdbot to do?

It gets so annoying

[ashen gust]

im going to wait a couple of days too

not missing out on anything anyway

[crystal forge]

I have a question, i tried to install and run the bot, but i clicked on imessge (which i dont think is set up) instead of like telegram, and i wanted to know ho wi could delete it and start again fresh

[unreal bear]

Run clawdbot onboard

[untold bronze]

you can set the level of security or auto approve for X time etc?

[unreal bear]

Wait how

Clawd tells me it can’t be overriden an that it’s a Mac security thing

[untold bronze]

Just tell it to do so, or well depends what you are approving. For example I let it have auto approval for 1pass to get all API keys etc for 15 minutes when we are deep in the repo updating and what not, so it doesnt need to ask all the time

are you running as admin or change mac level settings?

[pulsar trail]

hey , i need help , on the clawdbot dashboard i cant see any messages that assistance sends . and i want disconnect and reconnct my telegram

[untold bronze]

Like if you cant do it then clawd cant do it, think of it that way

[strange sluice]

Hi all. So I have hit my Claude limit again and need to buy more credits. But it’s limited use. Anybody have recommendations on a more cost effective model I can hook up to?

[untold bronze]

but if you put the mac security settings, waht admins can / cant do for example to where you want to be, then clawd can do the same

[pulsar trail]

how do you check credits ?

[unreal bear]

Right, I’ll have to play around with that

How do I set clawd to run as admin

[untold bronze]

you can just enter the mac admin password once on the account you are running, and then from mac lvl put it to be "active" for X duration?

Or probably better, open up the admin level access for the particular folder / setting you working on and not the whole account

it should prompt you "always allow" and then just accept that with a time constraint basically

[junior ingot]

hey all, i've been using clawd for a couple of days with Sonnet, and i'm really struggling to get it to be reliable. for example i ask it to remind me of something at a certain time, and it will just forget, despite having a Cron job for it. and the worst is i have tried to get it to use claude code in a PTY with tmux, but clawd just does a couple of tasks and then stops working on CC. i've tried using cron jobs, skills, and a combination on both, and it will sustematically just stop working eventually. how do tou guysget it to become reliable ? how can i make it just do things on its own more reliably ? this feels far from the use cases ive been seeing online right now....

[untold bronze]

why not run it on opus 4.5?

[junior ingot]

i am on rhe 5x plan and whenever i use opus it just chews through my credits so fast, i am worried it will be the case for clawd too

[untold bronze]

might wanna explore running local model then tbh, not an expert on that

[junior ingot]

cool thanks, but any clues as to the rest of my questions ? are you able to get clawd to be reliable ?

[vagrant pond]

currently im tryna install clawdbot on my pc butchrome and anydesk isnt working

and this man keeps forgetting our conversation of 5 minutes ago anyone experiencing this issue?

[prisma bolt]

so what could i do with this? i just hooked up telgram

[wraith lodge]

i get (no output) for literally everything
does anyone else face this error in tui AND solve it?

[stable sorrel]

Is the gateway running? clawdbot gateway status

[stable sorrel]

I'm currently using venice/qwen3-235b-a22b-instruct-2507. Pricing: https://docs.venice.ai/overview/pricing

[wraith lodge]

it is, apparently

[normal skiff]

have you tried using auth ?

[wraith lodge]

it worked yesterday locally for me when i built from source, but now i get (no output) on VPS

[stable sorrel]

Try running clawdbot agent --local --session-id test -m "hello". Do you get a response?

[gleaming palm]

anyone arount that can help with gateway setup ? i want to connect it to my mac from my cloud

[wraith lodge]

gr8 command, hmm but what can this possibly mean

is there a tool mode set to true in some config file

[stable sorrel]

Not sure 😅 What provider and model are you using?

[stable sorrel]

Please describe where the bot is living and what exactly you are trying to achieve

[gritty ridge]

Does anyone have this problem of chats getting split mid way?

[still bobcat]

(no outputs) not connecting to tui

[wraith lodge]

openrouter/meta-llama/llama-3.3-70b-instruct:free

[haughty anvil]

any idea how to unblock or fix that missing bin issue

[wraith lodge]

Try running clawdbot agent --local --session-id test -m "hello". Do you get a response?

[gleaming palm]

well its living in the vps hetzner cloud and what i want to do is i want to do node run so i can connect on my mac i have it installed on my mac but like i tried to change the .json to try to connect to vns but didnt work what i am trying to achieve is to run on the vps but can do stuff on my mac

[haughty anvil]

here as well , is there issue in my end

[amber escarp]

Does any one using clawdbot for x.com ?

tried to make my bot post but it says "The text injection isn't working due to X's dynamic content protection. The Post button is still disabled.
Issue: X has aggressive anti-bot measures that make programmatic posting difficult without official API access."

Can any one help me, pleasee

[stable sorrel]

I had similar issues with this model. Try a smarter one, such as qwen3-235b-a22b-instruct-2507 and see if it works.

[austere socket]

anybody using railway and can tell me what do to after "run setup", entered api token and discord bot token, now what?

[wraith lodge]

will try, thanks a lot

[stable sorrel]

You will need to tunnel from the Hetzner machine to your physical Mac. People suggest Tailscale for this, but I have not tried it myself. Evaluate if you really need the bot to have access to your machine, and if you do, if installing it locally would not be a smarter solution.

[gleaming palm]

yea true it comes to the same after all lol idk

[stable sorrel]

Do you run the bot on a virtual machine? If yes, it uses a datacenter IP address and a headless browser, some websites such as X will recognize and block your bot. There are ways around that, but none are guaranteed to work.

[gleaming palm]

Is it safe enough to run it locally ? not sure if i should try it out on my machine

[slow geyser]

My local LLM is dumb AF

[ionic basin]

thanks realise i need an even dumber version of that, installing on linux ubuntu, never installed that, havin to install stuff like curl, figure out how to make the hard drive big enough etc etc even before i get to your vide, will get there...........

[stable sorrel]

It is NOT safe. Running it on your machine is almost never a good idea. Do you have something in mind that you want to achieve? Perhaps there is a way of doing it on your Hetzner VM.

[gleaming palm]

check private dont wanna share here to be honest

[amber gust]

how to link the clawdbot to my discord?

[latent hedge]

Check #1459642797895319552 - it's in there

[violet patio]

guys do i need to buy a api key for this ?

as while attempting to set it up on the aws server after fixing a few problem i got this error when attempted to contact through telegram

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

[amber gust]

u know how to add my discord ob the bot?

[amber gust]

i can't find it

[ebon flame]

Hi everyone. I saw recently that people were integrating their Google accounts with moltbot for emails and calendar events. However, the bot now tells me it can't do that. Is this a problem on my end or was it nerfed?

[timber peak]

I'm posting a skill file I just had Claude build for me to help troubleshoot clawd issues. Maybe it will help someone else.

[north fulcrum]

anyone having success with connecting using a free openrouter model?

both openrouter and google ai studio keys are not working for me even for 1 try

[zinc mica]

do we have a preferred way for clawds to search this discord for possible improvements/ideas?

[slow geyser]

Not yet... i'm working on making a routine that checks the free models on OR each day to make sure the current model isn't switched to paid and to always ensure i'm using the best free model.

[north fulcrum]

I want to use my $10 google api key but im afraid i'll pass the $10 easily with gemini 3 flash. can i set limits somehow?

[cloud raft]

hello i ahve been going in circles trying to get my multiple providers set up.. i have gone through the docs tried throwing ai at it.. and i cannot get anything working aside the one provider i set up during wizard....anyone who can at least point me in the right direction?

[unreal bear]

Can anyone please tell me how to set up clawd on a Mac so that I do not need to approve everything?

[pearl solstice]

I get this with openrouter and i have no clue how to fix it

[stable sorrel]

I add models using moltbot configure. Did you try that?

[north fulcrum]

I think free models aren't really working. Getting 429 error from OpenRouter

[pearl solstice]

ah wait.

??? i have no spend limit

and theres ten bucks in openrouter

[cloud raft]

yea but im not sure if it their ui throwing me off or what.. i would like my grok, gemini,open ai , all working as well as my local ollama models

[north fulcrum]

Sounds like you limited it on openrouter's side

Check your key settings and if there are any limitations there

[stable sorrel]

What are error messages or symptoms when you switch to a newly added model?

[north fulcrum]

All i wanted was a free digital slave, why installing is so hard 💀

[cloud raft]

one sec let me try config again

[pearl solstice]

All good on that front,

[north fulcrum]

Do you have more than one key?

[pearl solstice]

Nope, just the one

[north fulcrum]

And I think :free models could be limited still even if you have money on your account - but could be wrong

[pearl solstice]

surely id be able to get a single prompt in though right

[north fulcrum]

I couldn't get a single response on free models 🙁

[cunning lichen]

Anyone what can tell me why the bot isnt answering me? Just set it up with OpenAI gpt-5 mini

[stable sorrel]

Try your API key with a curl request

[amber gust]

can someone tell me how to use this bot ?

[stable sorrel]

Sure! Read this first: https://docs.molt.bot/start/getting-started

[stable sorrel]

What is the output of clawdbot agent --local --session-id test -m "hello"?

[whole lodge]

thank you for the response 🙂 I was getting this with gpt-4-mini, I'll try using gpt5 and see if it gets it better

[cloud raft]

i think after hours i FINALLY got the json formatted correctly.. it let me save it... i should be good now?

[stable sorrel]

Paste it in the terminal, not in ClawdBot's TUI.

[cloud raft]

yea im good now lol ty for trying to help

[stable sorrel]

I wasn't responding to you 🙂 anytime though

[woeful loom]

hi, did you manage to find a solution for this?

[wraith lodge]

how did u solve the issue of (no output), could u elaborate?

[cloud raft]

not ont hat message but i just replied so you would see it either way lol

[past tundra]

Is it possible to join the internal preview of the ios app?

[wraith lodge]

which model are u using?

[limpid thorn]

hey

[craggy pollen]

Nope. I did manage to vibecode a change which shows where the problem is though:
https://github.com/badlogic/pi-mono/pull/1027#issue-3864234416

would be nice to merge it. Showing stack traces is safe in 99.999999999999999% of cases, wouldn't you agree @solid snow ?

[limpid thorn]

i try to use openrouter or groq for free models but neither of them work please help

[vital umbra]

hi! is it required to have claude cli installed on local machine for clawdbot to be working through antropic?

[stable sorrel]

No, it's not required. You can configure Clawdbot to work with Anthropic's API - or with Claude Max subscription (although this is a breach of Anthropic's TOS).

[stable sorrel]

Generally you'll need a smart model for Clawdbot to work properly. We can help you debug if you share what you tried and error messages.

[craggy pollen]

FUCK the Anthropic TOS

[cunning lichen]

Do you know if the GPT-5 mini is good for clawdbot, or should I get the money out of my pocket and pay for something like Claude Opus /Anthropic

[chilly forge]

hey everyone, new to this space..
should i wait a couple of months for updates and so on? or should i just go ahead and buy a mac mini and start;) thanks

[vital umbra]

the thing is, I installed clawdbot on my old macbook (claude cli doesn't support so old machine), but getting "no output". so looks like some configuration or auth problem on clawd<>anthropic side. so i thought maybe it is because i have no claude cli ont hat machine

[stable sorrel]

I have not tried, but I don't think it's good enough.

[stable sorrel]

Which model do you use, and how?

[cunning lichen]

Which AI would you recommend that isnt breaking the bank, because tbh the ai's seems really expensive

[vital umbra]

i'm using max subscription, opus 4.5. my issue looks like this one tho: https://github.com/moltbot/moltbot/issues/3361 and this one https://github.com/moltbot/moltbot/issues/3168

[stable sorrel]

venice/qwen3-235b-a22b-instruct-2507. See https://docs.venice.ai/overview/pricing

[stable sorrel]

Run this in Terminal: clawdbot agent --local --session-id test -m "hello". Does it work?

[limpid thorn]

mack i taged u in help

[vital umbra]

no. HTTP 401 authentication_error: Invalid bearer token (request_id: req_011CXa2xLZczfhztM7QxSPwz)

[cunning lichen]

And that one should be good?

[stable sorrel]

I use it with success. It's not as good as Opus, but significantly cheaper

[wheat acorn]

any idea how to give access ?

[cunning lichen]

Well alright, how would I change now from OpenAI to Qwen?

[stable sorrel]

Try to re-auth again with Claude. Make sure your subscription is valid.

[stable sorrel]

Register with a provider such as venice.ai, generate an API key and run clawdbot configure.

[cunning lichen]

Should I use OpenRouter so that I can use up my tokens for GPT-5 mini and then be using Qwen afterwards or would that me to complex just for the 5$?

[lime ferry]

I am trying to update from Clawdbot to Moltbot. I run clawdbot update on my terminal and it seems that works. After that I try to use the moltbot command instead of clawdbot but "command not found"

[stable sorrel]

What good is using up your tokens with GPT-5 Mini if the model is not smart enough?

[cunning lichen]

You're right.. Can I use the tokens to anything else? besides API

[slender pewter]

guys, what local model do you recommend to run for clawdbot? i only have a 3060 12gb, so i can't use large models and it seems like dolphin is not supported

[ivory rapids]

which skill works effectively for youtube summary? i get all the time the feedback that youtube video has no untertitel..

[stable sorrel]

If you want interactive chat, you will not be happy with the performance of local models on your GPU. Options are: remote model (paid per tokens), remote GPU (paid per hour) or improving your setup (paid once).

[stable sorrel]

I don't think so.

[slender pewter]

you mean it will be slow or flat doesnt work?

[stable sorrel]

Any model that can run quickly on your machine will simply not be enough. Any model that will be enough will be extremely slow.

[lone vessel]

Just now setting everything up today … I just keep getting this and it won’t set up telegram although I have given it the pairing code etc. can somebody help?

[slender pewter]

what gpu would you recommend for a local at least "good" performance?

[vital umbra]

thanks a lot for the help, although i already tried, i'll to dig more into that direction

[stable sorrel]

People seem to get good results with a Mac Mini M4 Pro 48GB+ and models like MiniMax M2.1. I've never tried though.

[stable sorrel]

Which provider/model are you using?

[lone vessel]

I was able to figure it out, realized I didn’t have any API credits lol

[gaunt orbit]

Hy there !
I'm struggling to get the Gemini integration running (both antigravity and google-gemini-cli). I keep hitting a hard 403 Forbidden error related to the Cloud Code API, even though I believe I've enabled everything.

Context:

Environment: MoltBot (Docker and tried Local install), latest version.
Provider: Tried via google-antigravity and google-gemini-cli-auth plugin.
The Issue:

I run the auth login command: moltbot models auth login.
I use the generated link, authenticate with my Google Account, and authorize my custom OAuth Client (I see my app name "ClawdBot" asking for permissions, so the Client ID is correct).
I get the localhost redirect URL and paste it back into the CLI.
It thinks for a specific moment, then fails with: Error: loadCodeAssist failed: 403 Forbidden Cloud Code Assist API error (403): Cloud Code Private API has not been used in project <MY_PROJECT_ID> before or it is disabled.
What I've checked:

In Google Cloud Console, "Gemini for Google Cloud API" is enabled.
I also found "Gemini Cloud Assist API" (which seems to be the new name for cloudcode-pa) and enabled it.
My OAuth Client is set to "Desktop App".
My user account is Owner of the GCP project.
Despite all this, the cloudcode-pa.googleapis.com endpoint rejects my requests. Has anyone faced this specific API block recently? Is there a different, hidden API I need to enable?

Thanks for the help! 🦞

[thorny burrow]

Anyone gotten moltbot working with Bedrock? They give free credits.

[tough notch]

hellcatvn@miniroute:~$ moltbot health

:lobster: Moltbot 2026.1.27-beta.1 (7eb57b6) — Because Threads wasn't the answer either.

│
◇
Discord: ok (@Ram) (663ms)
Agents: main (default)
Heartbeat interval: 30m (main)
Session store (main): /home/hellcatvn/.clawdbot/agents/main/sessions/sessions.json (3 entries)
- agent:main:discord:channel:1461686919489392774 (1m ago)
- agent:main:main (28m ago)
- agent:main:new (17486m ago)
hellcatvn@miniroute:~$ clsof -i :18800^C
hellcatvn@miniroute:~$ lsof -i :18800
hellcatvn@miniroute:~$ moltbot browser --browser-profile clawd start

:lobster: Moltbot 2026.1.27-beta.1 (7eb57b6)
   Shell yeah—I'm here to pinch the toil and leave you the glory.

│
◇
Error: gateway timeout after 15000ms
Gateway target: ws://127.0.0.1:18789
Source: local loopback
Config: /home/hellcatvn/.clawdbot/moltbot.json
Bind: lan
hellcatvn@miniroute:~$

any one know why I can't start my browser ?

[ivory spindle]

It seems like whenever I ask clawdbot to make an update on the config file it crashes and then resets all the data I set. Also does anyone have a memory fix. Whenever I start /new it loses context of everything before.

This is definitely cool

[storm cradle]

I ran upgrade, but everything is still under clawdbot? is that how its supposde to work?

[austere bluff]

hi i am looking for a way to send a message to moltbot from the same machine programatically. Is it possible?

[frank willow]

I've only set it up using the google ai studio -> generate API token and use that. You could perhaps try that?

[ivory spindle]

yeah mine is still running under clawdbot too

[tough notch]

that is new feature. since you want to have new chat don't related to previous why do we need previous context. If need previous context don't start new

[ivory spindle]

ok so every time I use new its just going to start completely fresh?

[fresh moon]

hey guys this keeps happening
[Historical context: a different model called tool "gateway" with arguments: {
"action": "status"
}. Do not mimic this format - use proper function calling.]

what happens and how to fix

[weary salmon]

I'm no longer gettings Whatsapp notifications on my phone after linking it to Clawdbot.. Could this be due to the notifications being 'read' with the linked device so my personal device isn't getting any notifications.

[near dragon]

Hi all, wondering if the claude max 100 subscription bia OAuth is enough to support moldbot, i used my gemini api key and in 5 hours i got a 55 dollars bill and canceled it, so have anyone tested the claude max 100 subscription with heavy usage? it is good enough?

[north furnace]

so i have it installed, no errors, but its not running the bootstrap stuff. in fact it cannot even read those files at all. could someone please point me towards what i might try to remedy this? is there something i need to toggle on?

[prisma bolt]

do yall get this?

[lilac spire]

Yes, very frequently if we try to give it a big task

[north fulcrum]

depends on your usage, frequency of cron jobs etc

[prisma bolt]

is there a way to fix ?

[austere socket]

moltbot isnt reacting at all anymore, /new and restart does not do anything, changing model returns error, any idea?

[prisma bolt]

or use another platofrm?

[stable sorrel]

Likely a weak model. Please share which.

[bronze wigeon]

Hey, does anyone know if vector memory search works with minimax?

[stable sorrel]

Which error is returned? Did you restart the gateway? What's the gateway status output?

[near dragon]

15 min cron jobs searching the web for financial advice, maybe crypto/stock movements, something like that, a couple of more cron jobs once each day, as a starter

[lilac spire]

The /new commands just deletes all memories it had from previous session unless you have to ask it to save the memory.md file so that you don't lose the context when you create a new session

[stable sorrel]

Really hard to tell! Why don't you give it a try? After a few days you can check Claude usage and calculate from there.

[lilac spire]

Try asking it to not use the max limit token in a single shot but split it into multiple queries so that you don't hit rate limit

[north furnace]

im using qwen3 16b via ollama on my 5080

[lilac spire]

Try the onboard command once again and choose to use existing config files

[stable sorrel]

I think the model is not smart enough. Try with 235B Qwen and see if it works.

[austere socket]

okay after restarting and switching the model it works, was antigravity before, did google do something against it?

[north furnace]

this is supposed to work with local. that obviously cant run on my system

[stable sorrel]

Yes, for me too, weaker models seem to cause crashes.

[north furnace]

its not crashing. its just acting like im using ollama by itself. it isnt bootstrapping, and doesnt have access to the filesystem

[stable sorrel]

It does not follow the system prompt. I had the same problem with 70b Llama. Switching to 235B Qwen solved this for me.

[ivory spindle]

anyone know how to have clawdbot update the config without breaking everything? Everytime I ask it to updaate something in config I have to reset TG

[wintry oar]

filed this bug, not sure if anyone else is having issues with moonshot? it was working okay for me the other night
https://github.com/moltbot/moltbot/issues/3443

[stable sorrel]

Gateway restart is necessary after config changes. Is this what you meant?

[ivory spindle]

Its just whenever I ask it to update the config file, it does do gateway restart but then I usually get errors on the config file itself so I have to go in and change and it is not usually able to make updates to its own config file

wondering if theres a way to work around it or solve it

[stable sorrel]

Maybe a model issue.

[ivory spindle]

Interesting, im using minimax 2

I wonder if it makes sense to use a stronger model for setup/config and go back to a weaker model later

[wild plinth]

Can someone help me install Moltbot

I think I installed it wrong

and now I want to uninstall it to do it again

[stable sorrel]

It depends on what you do with it, but in most cases I don't think it is. MiniMax M2.1 though should be good enough, although I haven't tried it myself.

[keen tulip]

Has anyone found a way around TUI showing better status updates? It sometimes just stops and doesn't provide any visual confirmation? I saw a UI tool that shows a visual representation of what's running

Have you found Crawbwalk to help for this?

[stable sorrel]

Please provide more information about the system you run it on and what you did to install it.

[lament tartan]

Hey!
Is someone using tailscale with clawdbot?
Its my second day trying to come up with a solution. i have tailscale installed both on my mac and vps, but whenever im trying to run a web ui, it doesnt authorise my ip.
i tried addint "trusted id"

[wild plinth]

Well I'm on windows right now and installed with the iwr -useb https://molt.bot/install.ps1 | iex
Then it said its reccomended to use another thing with windows and so I clicked no and then i accidentely pasted it again and then it just installed everything without saying yes or no and now I can't uninstall it

to do it again

correctly

[split lion]

Model change messed things up. Advice? I installed Clawdbot and had Gemini as the model. Everything was fine and it started to become useful but I hit the rate limit. I opened an Anthropic account and ran config again. It seems to have forgotten user.md and soul.md.

[tough notch]

Yes

[stable sorrel]

Check those files manually and see if they contain what you are looking for.

[stable sorrel]

I have no idea at all about Windows, sorry.

[wild plinth]

Ah no worries thank you tho

[grand glade]

Is there a great guide or video out there on how to get molt/clawdbot up and running in the absolute safest way possible etc (like in sandbox in utm or a digitalfoundry cluster etc) permissions security for tool use like using zapier and mcp then connecting it to your gmail etc - I just want to make sure I am setting this up properly

[split lion]

Will do. Is there a set of command to run to help with system hygiene?

[stable sorrel]

I don't know about guides (development happens so fast). But Moltbot is unsafe by design. If you allow it to access your Gmail account, it can leak information or send unauthorized mail. No sandbox or firewall are going to prevent this.

[stable sorrel]

Not sure what you mean by system hygiene. Ideally, Moltbot manages its soul and memory on its own.

[steel cosmos]

What's the recommend way for getting moltbot to code? have it install claude code? Or just ask it direclty since it's basically claude code already

[median ivy]

nothing, just ask

[stable sorrel]

Just ask it. If you use Claude Max subscription it will be very good at it.

[true gulch]

Are there any steps to remove all configurations and clawdbot safely? I think that this should be there, if I want to remove all.

[median ivy]

it costed me quite a few tokens, but I've quite a lot of complex skill that it coded up itself just asking... too lazy to do them myself and it's ok if it comes back to me later since if it has to build stuff in background...

[junior shoal]

any advice about running Moltbot on a virtual machine?

[median ivy]

delete ~/.clawdbot

[brisk pivot]

hey guys may i ask for help from an expert dont mind pay a bit for better 1:1 learning session hit me up if its possible 🙂
someone who already have hands on it. better if a web dev since thats what im planning to use it for mostly.
thanks in advance.

what i want to learn:

• how i connect my wordpress / next.js / php websites into projects to control with vibe code on a safe enviroment for testing before deployment
• to understand and learn how to use mcps
• to learn how to use "guardrails" mesurements
• to learn how to setup and connect molt to my systems (with better config for the basic example diff port for the connections)
• what is the smartest or best fit setup for molt bot? (vps or mac mini m4?)
• how can i config molt bot correctly with spefiec models to each task or to explain him what to use in each task (maybe prompts templates? i saw some in the skills hub)
• probably would have more things coming up to my head to ask through the session but thats mostly it.
• any extra knowledge would be most welcomed.

[median ivy]

too many questions 🙂 for the first things you asked, ask an LLM, serious answer, it will guide you

[stable sorrel]

My recommendation is to set up your bot with a decent model and work it out in the chat. This will be cheaper than 1:1 with a human, and you'll learn more. Since you know about web dev, you will be fine.

The smartest setup for your case, in my opinion, is a VPS and a remote model from Anthropic or any other provider.

[hardy ravine]

Bug: Image paste in webchat causes sequence gap error

When I try to paste an image into the webchat, it fails to send and shows:

event gap detected (expected seq 261, got 262); refresh recommended

Happens consistently. Refreshing the page lets me chat again but still can't paste images.

Environment:

• macOS
• Clawdbot 2026.1.24-3
• Arc browser (Chromium-based)

Workaround: Sending images via WhatsApp works fine, just webchat that's broken.

[brisk pivot]

im not a web dev my self im more of a seo saas creator

[stable sorrel]

Well, if you're about to create a SEO SaaS, you will become a web dev soon 😉

[brisk pivot]

well i already have seo saas haha

i want improve my self in the ai vibe coding field

thats why i dont mind pay 1:1 lesson i dont believe i need more than 1-2 hours at total to understand the whole thing .

[fair geode]

hello are we using vps to get set up?

[median ivy]

ask an LLM ninji

faster way to get an overview of what you need to look into

[stable sorrel]

I'm with John. Your best trainer is the bot itself.

[median ivy]

and then piece by piece you fill in the blanks, tested strategy

[split lion]

I checked the files. They contain most but missing a little. Overnight the behavior improved but still off. For example I created a Gmail for it gave it API access and it was creating Google Docs for me very nicely. After model change it is unable to do so and whatever I prompt it says I did it but don’t do it right.

[stable sorrel]

Which Anthropic model are you using now?

[true gulch]

Yes, but the external configurations, like telegram bot or something like that?

[split lion]

TOOLS.md has some correct information but it is not following. Haiku-20240307. Thanks for your help!

[hasty hemlock]

That's a ridiculously old Haiku model...

[split lion]

I am now realizing. I think I accepted the default

[hasty hemlock]

https://platform.claude.com/docs/en/about-claude/models/overview. Try claude-haiku-4-5-20251001 ?

[median ivy]

hmmmm everything is in there, if you want to delete the telegram bot you have to do it manually (the bot only has the chat as remaning data)

[gilded hatch]

gm gm (not sure where to post) being using the bot since yesterday on vps to create daily digest and stuff like that at first. very happy about it really. but then it's as if it became amnesiac. no memory of precedent interaction whatsoever. i forced it to reread all previous interactions and remediate. is there a way to really make it doesn't happen again? i wouldn't like it to let me down after weeks of fun together. thanks

[true gulch]

Yes Make sense.

[signal tartan]

Has anyone already installed the Clawdbot app on a Mac and can tell me how to set it up?

[fast ledge]

hi i still getting this please help

[median ivy]

yesterday someone was telling that node 24 does not work, 22 will

[fast ledge]

how can i uninstall this and instal 22?

[robust rampart]

Hello guys, im currently installing molt, but was wondering which ai model to use, what do you recommend?

[untold hazel]

how do i start clawd after installing?

[split lion]

Is there a way to do a fresh start of clawd without formatting the disk and stay from installing from scratch. My stupid choice of accepting a bad model during change messed up everything soul tools etc and I probably want to start fresh.

[signal tartan]

OS?

[untold hazel]

win11

[signal tartan]

Ok, i'm on mac. But i thin it is the same. In Terminal "clawdbot gateway"

[untold hazel]

ou ok lemme try

[hasty hemlock]

Just go to your User Home directory and open .moltbot/moltbot.json or .clawdbot/clawdbot.json depending on which version you used. It should have a property in there showing your Haiku model name and you can update to whichever you want. But yea, the clawdbot onboard is easier

[untold hazel]

only this pops up

[steel cosmos]

im having trouble installing plugins "swarajban@sb-vm:~$ clawdbot plugins install @moltbot/voice-call"
"npm pack failed: npm notice Access token expired or revoked. Please try logging in again.
npm notice Access token expired or revoked. Please try logging in again."

[signal tartan]

This is ok. Clawd is running

[untold hazel]

how do i get the other menu?

[signal tartan]

What other menue?

[ivory spindle]

Seems like since when i did clawdbot reset, I have to configure the api key and tg bot / id each time.

So I had it make a command to create/update the json file and it seems to be working better

[untold hazel]

i remember 2 terminals opened up on first run but i cant get the other menu now

[cloud raft]

so new issue.. lol been trying and trying, checked every setting 10 times. set the agent on it.. and no matter what i do i cant use / commands .. any ideas?

[wet drum]

I've got multiple agents added as discord bots to the same server with requireMention ture. The problem is, when I send a message to any agent, it always gets routed to the main agent.. So, every bot thinks its the main agent when I talk to it. Any ideas on how to route the messages to the correct agent?

[junior ingot]

does anyone here have any recommendations to get Clawd to follow along with the BMAD method ? Whenever I ask clawd to do anything within the scope of the bmad it just goes off sending vaguely related commands rather than summon the relevant agents.

[fresh moon]

hey guys ive been having trouble with my tui its not sharing any output even tho ive switched models and switched to a new session, was rate limited earlier. how do i fix this? idk why but i ahvent been getting any help from krill in #1459642797895319552

[ivory spindle]

maybe your model api key is not configured and did you select the correct model from terminal? If not try and switch the model

[strange sluice]

What’s auth?

[strange sluice]

How can I set up ? Appears it has several models?

[fresh moon]

i think so because its working on my telegram but not here on my cli

[opal nexus]

Hey, Im trying to install on a mac mini but im getting a npm install failed

[ivory spindle]

yeah every time you do a restart you have to input the api key and select the model, if you are using a different model you might have to select it again

[ivory spindle]

feed the errors to chatgpt, I got it working on an older mbp but it was an issue getting node running

[vapid owl]

can someone please help me. Ive onboarded successfully 3 times , but whenever i select Hatch TUI to launch the bot , IT never works, I message it but never get a response due to unconenected gateway or an invalid bearer ...

[thorny burrow]

What are the best options for giving molt access to "browse" websites I use login with google with? I dont want to give the LLM my password

[warm knot]

Model "kimi-code/kimi-for-coding" is not allowed.

Why is that?

[wet drum]

Nvm I didn't know how to setup the bindings correctly. For anyone else with this issue, you need something like this: "bindings": [
{
"agentId": "main",
"match": {
"channel": "discord",
"accountId": "main"
}
},
{
"agentId": "code-writer",
"match": {
"channel": "discord",
"accountId": "code-writer"
}
},
{
"agentId": "test-writer",
"match": {
"channel": "discord",
"accountId": "test-writer"
}
},
{
"agentId": "code-reviewer",
"match": {
"channel": "discord",
"accountId": "code-reviewer"
}
}
]

[ivory spindle]

I dont think you have the apikey set, and the correct model with it. Once you have it set it will run

happened to me when first getting started

[vapid owl]

ive ran claude setup0token and pasted the key multiple times

[ivory spindle]

sick I'm still working through getting multiple agents to collaborate on completing a task

[opal nexus]

I have been trying that but even chatgpt seems to be stumped

[ivory spindle]

what are your terminal errors?

[ivory spindle]

what command are you running? Try clawdbot configure and go through the steps

[opal nexus]

Im only getting npm intsall failed. it doesnt go past that.

[wet drum]

My strategy is to have one "manager" agent and multiple specialized agents. All inter-agent communication goes through the manager, and the manager is set up to regularly report the status to me.

[ivory spindle]

do you have funds setup on your openai key?

[ivory spindle]

looks like youre jsut about there

not giving you helpful error codes it looks like

[orchid vault]

Hi everyone, I’m reaching out because I’m a bit stuck. I’ve been working on setting up Moltbot on my Mac Mini for about 6 hours now, but I’ve hit a wall.

The Issue: Every time I finish the setup and try to interact with the bot (typing 'i' or other commands), the bot triggers a response, but the messages are completely blank. What I've tried so far:

Double-checking the configuration files.

Restarting the process from scratch.

Verifying my API keys/tokens (if applicable).

[ivory spindle]

interesting. just blank responses? Does the terminal say anything?

[wild folio]

hey guys.... for heavy users... how much do you guys spend in API calls... ballpark? (monthly)
I mean I've beeng paying chatGPT pro seat for me... moltbot opened my mind for a whole new world of possibilities... I'm affraid of breaking my glass ceiling... 💸

[orchid vault]

big soda could i dm u

[ivory spindle]

ok

[ivory spindle]

Im not sure yet, i just got it really working and I paid for minimax monthly but not sure if its taking the requests from there or just what ive preloaded but I'm already starting to burn a bit heh

[wild folio]

that's what I've been affraid... I mean, clawbot burnt my whole month ( i've never had ran out of codex calls) in two days. :/

[ivory spindle]

woah yeah thats burning through it. Luckily new models are coming around daily. So I think the pricing will get more competitive as the models still grow more powerful

[ivory spindle]

Have you thought about running a model locally? I think it will take a bit of computing power thorugh

[wild folio]

I've trying to run some... results are really sad...
but I'm learning a lot faster... I don't have a very powerful PC... I'm using my work computer for runnign a model... a Mac M4... qwen2.5-instruct-7b I just customized it to use max context on it... still... not very happy... sometimes it works... somtimes it does not...

[ivory spindle]

yeah that sucks to wait and also get bad results

[wild folio]

context too long... a lot of times...
then, 1m per reply is OK... but still... not very good...
Claude and Codex kind of addicted me. hahahaha

it was super powerful... then... now I'm here trying to cook my own stuff at home.... lolol

[orchid vault]

anyone know the fix?

i cant get clawdbot to work ever

should i uninstall and try and download again

[junior ingot]

you have to type clawdbod tui vince

[orchid vault]

vince-bot@Evans-Mac-mini ~ % clawdbot tui vince
error: too many arguments for 'tui'. Expected 0 arguments but got 1.
vince-bot@Evans-Mac-mini ~ %

[junior ingot]

you're just typing into the terminal right now... you have to actually start clawdbot to get it running. i recommend you follow along with the getting started guide here and then come back if you struggle: https://docs.molt.bot/start/getting-started

[orchid vault]

ive done it for 8 hours

[junior ingot]

vince, just type clawdbot tui

[wild folio]

have you tried a simple clawdbot update

[proud spruce]

Has anyone figured out how to handle back and forth interactions between clawd and a exec opencode/codex? My previous workflow with codex/opencode has had them stop and ask questions when there's ambiguity, but that doesn't seem to work with exec

[silent wren]

Unable to link GeminiCLI OAuth to clawdbot's gemini model

[orchid vault]

whats a command to delete clawdbot off my mac and then start from scratch

[silent wren]

clawdbot reset

[raw crater]

soooooo i am trying to run molt on my local server, and access it from my laptop, but it wonm't let me do that without SSL, but clawdbot gateway run has no option to offer SSL

[wild folio]

claudbot uninstall

[raw crater]

but i also haven't configured any chat channels for it

[orchid vault]

clawdbot uninstall

send failed: Error: gateway not connected
disconnected | error
agent main | session main | unknown | tokens ?
─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────

[wild folio]

not sure what to say.. try following this step-by-step, @orchid vault
https://docs.molt.bot/install/uninstall

[orchid vault]

i have for 8 hours

im dumb

[wild folio]

there's another...
https://docs.molt.bot/cli/uninstall

[wild folio]

we are both, then... I'm on the same train

[hardy mulch]

when im running the tui and its doing some stuff in the background I cant really see what its doing - it just shows runing... but i notice the web UI http://127.0.0.1:18789/chat interface does show much more of the tool calls etc.. TUI just makes me nervous as I can't see what its doing. Also sometimes there is no like running... indicator but the agent is working just needs more time. Is this marked as a github issue anywhere?

[winter tusk]

I'm running into LLM request rejected: messages.176.content.1: unexpected tool_use_id found in tool_result blocks: toolu_018yGkcNmnZtYYHH4ZaPSpgY. Each tool_result block must have a corresponding tool_use block in the previous message.

Is there a way to just easily reset the chat? every message i send triggers the same thing

[sterile condor]

Are you trying to remove it or get it working @orchid vault ?

[median ivy]

try enabling /verbose and /settings ->thinking, that shoudl improve what you see in the tui, way more stuff

[orchid vault]

im trying to remove it so i can start from scratch and then get it working

[cunning lichen]

What the hell is happening

I tried to have it installing clawdhub itself, but it seems like it fucked op

[sterile condor]

@vince since the gateway is not running I think you're safe to start the install from the beginning.

[lone vessel]

I just keep getting these messages.. help??

[placid python]

How do you guys let control moltbot your email. All what we trying was not successfully because it's not an business account and not working with Azure or power remote. It's outlook and I want give him complete access

[queen oak]

do you have to use tokens accounts for this to work, I started with Claude Opus and blew through my budget in hour, decided to move to gemini pro (w API key) and that does not seem to be working (blank response), any ideas out there?

[sterile condor]

@orchid vault before you start the install, you may also double-check that the clawd and .clawdbot folders have been deleted from your working (home) folder. That way any old configurations won't have a chance of getting in the way.

[sterile condor]

Just yesterday I set up himalaya to access mail via imap and smtp. So far so good. I am pretty sure Outlook allows IMAP access so that should work for you.

[frigid dawn]

had the same issue, and tried the one-liner install method and worked. trying via pnpm etc i attempted failed as I had the same issue. What install command did you use?

[placid python]

Moltbot tryed but was blocked from outlook

[orchid vault]

in telegram i am getting a clawdbot access not configured error message ask the bot owner to apporove with clawd bot pairing approve telegram <code>

is it working???

[sterile condor]

yes good job!

[orchid vault]

how do i fix this telegram thing

[sterile condor]

yes approve the code and it should be good

https://docs.molt.bot/start/pairing

[lucid veldt]

How can we use azure openai models as the primary model?

I have gpt 5.2 chat hosted on my Azure openao

[sly shell]

Curious how people are here using this as a coding agent. Do you give it access to your github repo? Create a separate one? Other patterns? Any patterns documented yet?

[signal slate]

Can anyone help me with installing this molt bot?

I am trying for the past 2 hours can't get this installed.
It shows error everytime I run the cmd

[true gulch]

If you have npm globally installed you can use the command that you find on quick start

[orchid vault]

im confused do i talk in the tui or telegram

[wide dirge]

alawdbot is always unresponsive, but the status shows "Health OK". After I send several messages in a row, it might only reply after 10 or more minutes. Has anyone else encountered this issue? How should I resolve it?

[orchid vault]

telgram not working says access not confdiguired

[true gulch]

This: npm i -g clawdbot

[wild folio]

Just found out, My ollama models are superfast when querying from terminal... but from moltbot they are painfully slow... :/

[feral knot]

Context overflow: prompt too large for the model. Try again with less input or a larger-context model. - anyone have a solid solution for this?

[signal slate]

I’ve tried the standard global install, but I'm running into a persistent environment conflict. Here is what I’ve done so far:

Updated Node.js: I am now on v24.13.0 (confirmed via node -v).

Cleaned Cache: I’ve run npm cache clean --force and uninstalled previous versions of the bot.

Attempted Global Install: When I run npm install -g moltbot@latest, it says 'added 1 package', but no executable (like moltbot.cmd) is actually created in the %AppData%\npm folder.

The Error: When I try the quick start command (moltbot onboard) or the official PowerShell script, I get an ENOENT error followed by: npm error Class extends value undefined is not a constructor or null.

It seems like my npm global prefix is corrupted or misconfigured, preventing it from linking the commands properly. Do you have a suggestion for forcing the symlinks on Windows when the standard global install fails to create the binary?

[glacial sail]

Can anyone confirm if gooogle does still have a free API level? Been told this is but appears not- google/gemini-2.0-flash-exp

[civic harness]

Hi I just installed moltbot with Claude Code setup-token, you know why I have this message when I send a message through Telegram?:
HTTP 401: authentication_error: Invalid bearer token (request_id: req_011CXaHxnsA4VLDTF98ihyuE)

[true gulch]

I recommend you use nvm instead of node globally. Anyway, you have to use clawdbot currently. I know that there's a renaming WIP.

[balmy hare]

I know Clawdbot is meant for powerusers basically, but is there an easy way that normies like me can get into it? or is it to early? 😄

I got a windows machine

[pseudo marten]

how to migrate from clawdbot to moltbot

[cunning lichen]

How can I solve this issue, I think it has messed itself up, I tried making it install clawdhub itself

[unreal bear]

Thanks!

[dull pebble]

Hi guys! I saw that it was renamed, the question is: if I install it with a brew – is it works?

==> clawdbot ✘: 2026.1.23
https://clawd.me/
Not installed
From: https://github.com/Homebrew/homebrew-cask/blob/HEAD/Casks/c/clawdbot.rb
==> Name
Clawdbot
==> Description
Personal AI assistant
==> Artifacts
Clawdbot.app (App)
==> Downloading https://formulae.brew.sh/api/cask/clawdbot.json
==> Analytics
install: 3,324 (30 days), 3,324 (90 days), 3,324 (365 days)```

[opal nexus]

Well i finally got it set up but the responses are blank.

[hardy mulch]

can u enable view of thinking/tool calls in telegram and/or whatsapp? Which has more verbose messaging?

[orchid vault]

i have no idea if im even using clawdbot right......in bbrowser tui....moltbot only responds like a llm.......and i cant get my telgram connected......how close am i to having it set up and wokring? i would appreciate any help!

[civic harness]

You know why when I select Claude Code CLI to auth my claude code subscription, it delete the credentials in keychain access and tell me: No Claude Code CLI credentials found in Keychain ("Claude Code-credentials") or │
│ ~/.claude/.credentials.json. ?

[orchid vault]

plz

[wintry ocean]

Hey everyone! 🤘

I’m setting up a dedicated Mac Mini as my primary Moltbot node. My goal is to keep my personal environment a total 'clean slate' while giving the agent the system-level permissions it needs to be effective.

Since this machine will have high-level shell and file access, I’m looking for best practices on the following:

Apple ID Isolation: For those using iMessage support, are you using a dedicated burner Apple ID or just a separate local User Account? I want to ensure my primary Keychain, Messages, and Photos never sync to this node.

Network Guardrails: What’s the consensus on network isolation? Is anyone running their Mini on a dedicated Guest Wi-Fi or using specific firewall rules to ensure the bot can’t 'see' other devices on the local network?.

Sandboxing & Scoping: I’m planning to enable non-main sandboxing mode in my clawdbot.json. For those running financial or dev-heavy workflows, how are you scoping your workspaceAccess? Do you find 'read-only' (ro) sufficient for most debugging, or is 'read-write' (rw) a necessity?.

Remote Access: I’m planning on using Tailscale for secure remote management instead of open ports. Any specific hardening tips for the Gateway binding when accessed over a mesh VPN?

I’m aiming for a 'set it and forget it' 24/7 worker that won't compromise my primary workstation if things get spicy. Would love to hear your 'day one' hardening scripts or setup checklists! 🦀

[prime nimbus]

run through onboarding and add it correctly there then restart gateway

[orchid vault]

cheapest chatgpt model that works with moltbot?

does 5 mini work?

[prime nimbus]

The docs /models help answer that

[orchid vault]

link?

[prime nimbus]

manners?

[orchid vault]

?

[prime nimbus]

Please and thank yous? How you write comes across, short, blunt and demanding, docs are here: https://docs.molt.bot

[orchid vault]

not that deep ; its typing on keyboard not face 2 face

grow up

[prime nimbus]

LOL good luck 🙂

[opal nexus]

Im trying to complete the setup but the assistant is responding with blank messages

[orchid vault]

anyone know how to correctly set up telegramSelect channel (QuickStart)
│ Telegram (Bot API)
│
◇ Telegram already configured. What do you want to do?
│ Modify settings
│
◇ Telegram token already configured. Keep it?
│ No
│
◆ Enter Telegram bot token
│ _
└

telgram gave me a pairing code not sure where to put?

[prime nimbus]

Wrong model most likely

try type wake up, my friend! to see what it comes back with, sounds model or actual connection to api, worth doing doctor too

[spiral pond]

how can i fix this?

[orchid vault]

anyone know how to correctly set up telegramSelect channel (QuickStart)
│ Telegram (Bot API)
│
◇ Telegram already configured. What do you want to do?
│ Modify settings
│
◇ Telegram token already configured. Keep it?
│ No
│
◆ Enter Telegram bot token
│ _
└
telgram gave me a pairing code not sure where to put?

[prime nimbus]

any logs for it from the gateway etc?

[orchid vault]

anyone know how to correctly set up telegramSelect channel (QuickStart)
│ Telegram (Bot API)
│
◇ Telegram already configured. What do you want to do?
│ Modify settings
│
◇ Telegram token already configured. Keep it?
│ No
│
◆ Enter Telegram bot token
│ _
└
telgram gave me a pairing code not sure where to put?

[dusk prawn]

Hi. Is it possible to use a fixed provider for a model from openrouter?

[spiral pond]

something about gateway token mismatch, could that be it?

[opal nexus]

i tried the wake up and it was still blank. I chose chat gpt 5 for the model.

[prime nimbus]

nah that sounds browser related, might be worth giving gateway a restart, tail the logs and try recreate, see what happens

[prime nimbus]

try gpt 5.2, if not run doctor see what it says

i would expect no output, not blank, sure api keys etc are correct and working? again doctor will help

[nimble phoenix]

Does anyone know like if I install it on a vps then how can I give it a browser to do browser tasks or soemthing is there something on cloud that works good.?

[opal nexus]

it responds blank even when i type doctor

also thank you for the help

[prime nimbus]

nah mate run clawdbot doctor in command line, not in into the chat

[prime nimbus]

run onboard command and on model/auth provider select openrouter and follow it through

[tulip hinge]

Hey guys, have you tried this? how do I integrate it on my bot? or is it already built in? https://github.com/moltbot/lobster

[ivory spindle]

anyone know how to setup agents easily? To where my clawdbot can spin up agents?

[spiral pond]

i just keep getting it haha

[opal nexus]

I tried the doctor. it didnt resolve the issue. how do i change the model?

[spiral pond]

is it normal to have 2 sessions?

[prime nimbus]

This page will help you do it: https://docs.molt.bot/cli/models#common-commands you could also flow through the clawdbot onboard command and select it via that

[signal slate]

I followed the NVM suggestion, but it seems to have broken the environment further. Here is the status:

Clean Slate: I fully uninstalled the global Node, deleted the C:\Program Files\nodejs and %AppData%\npm folders, and restarted.

NVM Install: I installed NVM for Windows (set to C:\nvm4w).

The Clash: Even after running nvm install 24.13.0 and nvm use 24.13.0, I’m getting a MODULE_NOT_FOUND error for npm-cli.js.

Missing Binaries: It looks like NVM is downloading the node.exe but failing to pull/extract the corresponding npm and corepack zips into the node_modules folder.

No Repair: Even nvm on and corepack enable are failing because the underlying scripts aren't in the path NVM created.

Since the standard NVM and Global Node routes are both hitting 'missing module' or 'undefined class' errors on this machine, do you have a standalone binary or a Docker image for Clawdbot? Or is there a specific mirror for NVM that actually includes the npm bundle for v24?"

[orchid vault]

not working!

im almost there

anyone know how to get my bot wokring

[chilly forge]

i've bought a mac mini, now what?

[torpid furnace]

Buy another one. Make them fight.

[chilly forge]

no like fr, what's next
i've set up a new icloud for my bot

[cunning gate]

Me too

[spiral pond]

switched models from kimi k2.5 to glm-4.7 and its working again

[prime nimbus]

thought it would be model related, good to hear!

[orchid vault]

please help me

thanks

[ivory spindle]

dm'd you

[pine yacht]

Hey folks, I accidently refreshed during setup and im trying to get back to where I was.

When re-running the quick start command it brings me to the setup, but it runs the doctor and I don't know how to get back up to select Yes and continue the flow

If anyone has any idea please let me know but otherwise I'll keep playin around

[prime nimbus]

run clawdbot onboard again and select yes, select use defaults for everything follow it through and then select TUI, it should be where youre at unless its web they pick that

[prime nimbus]

look what that magic word got you 😉

[ivory spindle]

yeah man I had to get through this this morning. Everytime I added a new config update i had to update tg id and model api, was a pain

[civic harness]

You've selected Claude Code I guess? If you send a message from telegram to the assistant it will show a 401 http code

[ivory spindle]

i got it solved by having it terminal crreate a command that you can run in terminal that will create or update the clawdbot.json file with your updated variables so everytime it crashes you run the command and it populates your api keys/mode of communication.

[opal nexus]

i changed the model but its still not outputting anything

[prime nimbus]

Whats the gateway logs saying and whats the safe output you can share to help troubleshoot, no output is something is wrong with api or model (99 percent)

[spare arch]

How are you guys saving money? I've spent $50 today in anthropic API usage 🙁

I tried convincing clawdbot to use personal account oauth for claude/gemini to not success

[prime nimbus]

I've get $250 from work in azure open ai, however, i've testing my fresh build using github copilot, and thats pretty sweet, i think subscriptions like that are the way forward

you could also just ollama it, zero cost, lol, use anthropic for the setting up of clawdbot, the wake up stuff, then switch the ollama when youre ready for just local ai models

[sturdy marlin]

I am having the same issue and the solution proposed didn't work for me. I would appreciate anyone's help

[ivory spindle]

what do the logs in terminal say? Anything?

[grand scaffold]

cls

[sturdy marlin]

Health check failed: gateway closed (1006 abnormal closure (no close frame)): no close reason Gateway target: ws://127.0.0.1:------ Source: local loopback Config: C:\Users-------- Bind: loopback

[prime nimbus]

restart gateway, try agian

[pine yacht]

thank you legend

[spiral pond]

i told glm4.7 to fix it and tried switching back but still error. now when i switch to glm4.7 again i still get it. if i do the status command i see the active session is still on k2.5 and i dont know why. how can i force it to change? 😄

[prime nimbus]

use the models commands, docs here lol https://docs.molt.bot/cli/models#common-commands

we've all tried it 😄 lol

[spiral pond]

I am doing that, and it says default
Glm, but on the session it remains unchanged😅

[prime nimbus]

give the gateway a restart, or try /new and see if it picks it up

[prime nimbus]

you could also clawdbot tui for speed, remember onboard as a good thing though

[pine yacht]

quick follow up @prime nimbus im getting these streaming connected messages now that im setting it up. Is there any way to remove them so that it's a bit more readable UI? Might be how I set it up.

Let me know if you know, but otherwise, thanks for your help

[prime nimbus]

ew, you've got me there, i've not come across that myself yet, what does the web portal look like?

[sturdy marlin]

That worked now. Thank you!

[queen oak]

I was using Claude opus and went through my budgeted tokens within an hour. Switched over to Gemini pro and it’s not given me a response. Any ideas?

[spiral pond]

Eventually it changed. Thanks for the help!

[latent hedge]

Is there an easy way to uninstall EVERYTHING and start from scratch?

[prime nimbus]

You'll want this: https://docs.molt.bot/install/uninstall

[modest raft]

Is therea way to remove the new security policies so clawdbot can have full acces and install programs and etc?

[pine yacht]

sorry mate, what do you mean by web portal. Im accessing through aws, I believe it's running on Ubunto. Not super technical.

[strange sluice]

What llm are you using? What is the cheapest?

[prime nimbus]

run onboard again, go through it, restart gateway, then when it asks to pick TUI or web, if you scroll up a little you'll see a web url with a token, open that in your browser

[latent hedge]

Actually.. I might wait with that uninstall until there is a pure updated moltbot. Still writing clawdbot in the terminal, etc

[prime nimbus]

Exactly! Live the easy life brother! lol

[ivory spindle]

also want to know

[pine yacht]

hey mate, is that the link under control ui?

[spare arch]

I will try ollama - i opted in for $20

Better than $100/day usage on claude

[prime nimbus]

Ollama, but if it depends on provider, its best to ask AI to search the cheapest to use lol, i know openai 4o is cheap, its also what subscriptions you got? for example github copilot is worth it, alot of allowance and codex models etc

This is copilot pro

Grok Code Fast 1, and Raptor mini is missing from it, needed to scroll

[prime nimbus]

its the url with port 18789

[limpid copper]

Has anyone gotten reminders to work reliably? As in they are sent to you via telegram after the cron job fires? There seems to be a few GitHub issues up and about, but no one has given a fix.

[viral solar]

clawdbot config get agents.defaults.model

🦞 Clawdbot  2026.1.24-3 (885167d)
   I don't just autocomplete—I auto-commit (emotionally), then ask you to review (logically).

{
  "primary": "ollama/glm-4.7-flash"
}

my default model of my config ^

20:36:37 [diagnostic] lane task error: lane=main durationMs=13 error="Error: No API key found for provider "ollama". Auth store: C:\Users\**\.clawdbot\agents\main\agent\auth-profiles.json (agentDir: C:\Users\**\.clawdbot\agents\main\agent). Configure auth for this agent (clawdbot agents add <id>) or copy auth-profiles.json from the main agentDir."
20:36:37 [diagnostic] lane task error: lane=session:agent:main:main durationMs=15 error="Error: No API key found for provider "ollama". Auth store: C:\Users\**\.clawdbot\agents\main\agent\auth-profiles.json (agentDir: C:\Users\**\.clawdbot\agents\main\agent). Configure auth for this agent (clawdbot agents add <id>) or copy auth-profiles.json from the main agentDir."
20:36:37 Embedded agent failed before reply: All models failed (2): zai/glm-4.7: No API key found for provider "zai". Auth store: C:\Users\**\.clawdbot\agents\main\agent\auth-profiles.json (agentDir: C:\Users\**\.clawdbot\agents\main\agent). Configure auth for this agent (clawdbot agents add <id>) or copy auth-profiles.json from the main agentDir. (auth) | ollama/glm-4.7-flash: No API key found for provider "ollama". Auth store: C:\Users\**\.clawdbot\agents\main\agent\auth-profiles.json (agentDir: C:\Users\**\.clawdbot\agents\main\agent). Configure auth for this agent (clawdbot agents add <id>) or copy auth-profiles.json from the main agentDir. (auth)

Hello, someone could help me please ? i'm using glm-4.7 on ollama but he keep saying me this ^

[prime nimbus]

ollama local? use the ollama launch clawdbot command to configure

[viral solar]

wich one ?

😭

[prime nimbus]

literally easier than you think: https://docs.ollama.com/integrations/clawdbot

[viral solar]

oh

clawdbot agents add <id>) or copy auth-profiles.json from the main agentDir. (auth)

[ivory spindle]

If I run ollama locally how long until my mac explodes?

[prime nimbus]

depends if you have you house heating on, or asking the mac to heat your house.... 😉

[viral solar]

same error:

2026-01-28T20:46:13.007Z [diagnostic] lane task error: lane=main durationMs=13 error="Error: No API key found for provider "zai". Auth store: C:\Users*.clawdbot\agents\main\agent\auth-profiles.json (agentDir: C:\Users*.clawdbot\agents\main\agent). Configure auth for this agent (clawdbot agents add <id>) or copy auth-profiles.json from the main agentDir."
2026-01-28T20:46:13.008Z [diagnostic] lane task error: lane=session:agent:main:main durationMs=17 error="Error: No API key found for provider "zai". Auth store: C:\Users*.clawdbot\agents\main\agent\auth-profiles.json (agentDir: C:\Users*.clawdbot\agents\main\agent). Configure auth for this agent (clawdbot agents add <id>) or copy auth-profiles.json from the main agentDir."

[prime nimbus]

have you put the API key in? followed onboard command and selected it and gave it? sorry i dont use zai

[viral solar]

i mean why i should put a KEY if it's a local model ?

[prime nimbus]

your primary is ollama/glm-4.7-flash now you say zai... lol, i think you need to follow the getting started

you need to follow the link i sent you, close and run the ollama commands, it setups clawdbot for you automatically to use the model you select

[mighty harness]

Hi guys is it necessary to migrate from clawdbot to moltbot ? I just used curl for latest script and its still clawdbot in cli

[viral solar]

Bc i used "ollama launch clawdbot --config"

[flint relic]

hello

where do i message in whatsapp

[tacit slate]

@sinful dawn
Yh so that didn’t work

Such a clown lol

[sinful dawn]

what model are you running?

[sick cedar]

How to connect azure OpenAI keys with clawdbot

[sinful dawn]

you need to setup the whatsapp gateway. it adds the bot as you, so you message yourself to talk to it

[tacit slate]

Glm 4.5 flash free tier. On my Mac mini m4 even tho I was using this model for a whole week on my raspberry pi and no single issue. Is this cuz it needs to handle more or smth?

[flint relic]

ohh i got confused by the screenshot, it shows a profile with bot name i thought i have to text there but werent able to find it

[prime nimbus]

search in the help channel on the left, theres a perfect example copy and paste to use in there

[sick cedar]

It’s not working

[sinful dawn]

ask your bot

[sinful dawn]

your bot isn't working?

[flint relic]

which version is it

[sinful dawn]

a couple releases ago running Opus 4.5. (I was waiting on a PR to be merged. Need to rebase.)

[prime nimbus]

this! this is literally it HAHA!

[mellow heron]

I wonder when a market starts to arise for 1. Setting up moltbot for others techincally. 2. Communicating with other peoples moltbots to have moltbot setup things for the user.
I feel like there needs to be something there. I suppose ideally you just give it skills and the ramp up to a useful bot is faster because there are so many rails for it to run on. But it would be interesting to see what help one user could give another.
I have an email setup for my bot. What would be the cools setup email for it to receive that made it do (x) behavior.
Are really defined setup prompts a thing? Where you run it and then you all these cron jobs that just setup and it is off to the races? I feel like we need something very sharble in that space.

[spare arch]

Im struggling getting Ollama setup

[sinful dawn]

you can totally install a 2nd moltbot and tell it to fix the first one. even across systems.

[spare arch]

I installed it on my machine and im telling clawdbot to setup but Model "ollama/qwen2.5:7b" is not allowed. Use /models to list providers, or /models <provider> to list models.

[flint relic]

isnt it renamed, are you using old version? showing moltbot-tui for me does this matters

[prime nimbus]

literally easier than you think: https://docs.ollama.com/integrations/clawdbot

[fresh moon]

hello how to let clawdbot remember my credentials? started a new session after setting up gmail and its all gone

[mellow heron]

Well. thats cool. But that isn't another person, which is what I was thinking. External help or suggestions. Not me doing more things.

[sinful dawn]

besides "ask your bot", look at setting up gog for google

[sinful dawn]

oh sure. fair enough

[mellow heron]

It would be neat, I think, to have starter prompts in a library. That people share. For different tasks. Maybe?

[mighty harness]

?

[torpid umbra]

are people still using codex with moltbot? it feels like it is the cheapest option, is it good?

also, would the 20$ sub be worth for light script usage etc, but not for SUPER hardcore coding. or we still need 5.2 high with the 200$ sub?

[sinful dawn]

#old-models is a great place for that. You can auth to openAI with a codex subscription and the codex cli tool. Some people definitely do it.

[sinful dawn]

I think thats all I have. plenty of gpt-5.2 and codex utilization, though i lean hearder on anthropic

[flint relic]

i tried it first, openai oauth was showing state mismatch

[pearl dove]

seems like i've completely bricked the bot - can't get uninstall to work at all, and am getting no response when typing in clawd control.

if theres a better channel in discord to ask this please let me know (noob)

[sick quest]

can i use it with my chatgpt subscription or i need to get a claude subscription?

[sinful dawn]

did you login with codex?

[sinful dawn]

nah - you found it.

what OS are you in?

[sick cedar]

Can u share this link

[sinful dawn]

can you use moltbot with an openai subscription? sure

[pearl dove]

I'm using Mac OS tahoe 26.2

[flint relic]

yes, it gave a token with localhost url

[sinful dawn]

I can't guarantee this will work for you. It is just what my bot is telling me. Tools like moltbot doctor and moltbot onboard will walk you through setting up access to mass-market models (not azure of course) with which you can ask your bot. discussions on models in #old-models. Google Antigravity has a free trial you can use.

  // Store API key in env (or use pass / .env file)
  env: {
    AZURE_OPENAI_API_KEY: "your-azure-api-key"
  },
  
  agents: {
    defaults: {
      model: { primary: "azure-openai/gpt-4o" }  // Use your deployment name
    }
  },
  
  models: {
    mode: "merge",  // Merge with built-in providers
    providers: {
      "azure-openai": {
        // Azure OpenAI uses a different URL format with api-version
        baseUrl: "https://YOUR-RESOURCE-NAME.openai.azure.com/openai/deployments/YOUR-DEPLOYMENT-NAME",
        apiKey: "${AZURE_OPENAI_API_KEY}",
        api: "openai-completions",
        // You may need to pass API version in headers
        headers: {
          "api-key": "${AZURE_OPENAI_API_KEY}"
        },
        models: [
          {
            id: "gpt-4o",
            name: "GPT-4o (Azure)",
            contextWindow: 128000,
            maxTokens: 4096
          }
        ]
      }
    }
  }
}```

[sinful dawn]

did you finalize the authentication inside the codex cli?

[flint relic]

yes, i had entered it the thar credential and saved. it couldve been some misconfiguration idk but working with gemini for now

[prime nimbus]

just move env stuff to .env file, then keep everything the same, it works as is, models under openai should match the name of model in primary for agent though

[sinful dawn]

a lot of the value-added providers (I know aws bedrock, but maybe azure, openrouter?) will have different model names. You need to validate them in the catalog.

[prime nimbus]

yeah the deployment name is pretty much everything in azure lol

[molten vector]

Yo can someone help me out with ios App troubleshooting? Cant get webui to load

[late dust]

did you ever figure this out?

[true gulch]

This appears to be a misconfigured node environment issue that is not related to Clawdbot. It turns out the issue is due to the wrong path of node in system variable. Look if you have something like this C:\Program Files\nodejs

[viral solar]

guys, any tips ?
2026-01-28T21:17:00.135Z [discord] Slow listener detected: DiscordMessageListener took 323.8 seconds for event MESSAGE_CREATE 😭
Even on my panel, it took like 2min to answer (no reasoning)

[signal slate]

I've installed it using docker

[orchid vault]

my bot opens safari but cant do anything else....

To give permissions that allow more extensive control on your Mac, follow these steps:

1. System Preferences:

   • Open System Preferences from the Apple menu.

2. Security & Privacy:

   • Navigate to Security & Privacy.
   • Click on the Privacy tab.

3. Accessibility:

   • Select Accessibility in the left sidebar.
   • Click the lock icon at the bottom left to make changes.
   • Authenticate using your password or Touch ID.
   • Add any relevant applications by checking the boxes.

4. Full Disk Access:

   • In the same Privacy tab, click on Full Disk Access.
   • Again, click the lock icon and authenticate.
   • Add the applications that should have full control.

5. Automation:

   • Check the Automation section to ensure apps are allowed to control other apps. You may need to launch the applications first to trigger permission prompts.

Make sure to save changes and restart any necessary apps. Let me know if you need further assistance!

i did this

and still

anyone know how to fix

[sinful dawn]

what model are you using?

[brave frost]

I'm using vps and telegram, now every single message is coming back with "Context overflow: prompt too large for the model. Try again with less input or a larger-context model" regardless of which model is set, even opus 4.5... can anyone help? @mental mica

[rose jungle]

i think my aws instance is ready. can anyone direct me to the step of loading molt on aws?

[sinful dawn]

use /status and /list context - you can figure out why youre sending so many tokens and fix it.

[orchid vault]

have molt bot working it can open safari but can not search or browse, how to fix so my bot can do things and be alive🦞

im using gpt4 mini

[sinful dawn]

This is experimental open source software with substantial risk. You're going to need to have a basic enough understanding of the technology to get it running to have any chance at all of running it safely.

https://docs.molt.bot/install/installer

https://docs.molt.bot/install/ansible <-- HIGHLY recommended method.

[mighty mortar]

it says its installing moltbot, but then the intro is still clawdbot, not sure what else to try, tried removing everything from the system and then run installer again

still clawbot installation detected

[orchid vault]

have molt bot working it can open safari but can not search or browse, how to fix so my bot can do things and be alive🦞

[sinful dawn]

you're going to struggle with tool calls, and self-improvement wont' be inspiring. I'm not surprised it won't do much in your mac, and its not going to be able to help you fix permissions etc.

[orchid vault]

!!?

[orchid vault]

which model should i use

so i dont spend a lot of money but it act works

thanks

sm

[sinful dawn]

#old-models is a great place for that. Opus-4.5 is HIGHLY recommended. This software is in its infancy, and Opus-4.5 has proven it can use the tools, read the docs and the code, and fix broken stuff. Its expensive, and as such, this project may or may not be for you. Other models can do less stuff, and there is a case for it, but your use case may or may not work with them.

[orchid vault]

thats to expensive

[sinful dawn]

You can use minimax-m2.1, glm-4.7-flash, sonnet, haikku, but if you don't have a use case and you want to experiment and feel the edges, you'll need a more intellegent model.

[mighty mortar]

is there any migration guide? from clawdbot to moltbot

[pearl oar]

Has anyone been able to successfully run the git install of Moltbot on Windows? How did you manage it? I tried the hackable install line from the website and a git install line that Krill provided which it took from the documentation, but neither worked

[hasty sapphire]

For some reason clawdbot is replying, but the text isn’t visible at all 😄
Could someone please help me figure out what’s going on?

[torpid umbra]

kimi 2.5 is spammed in #old-models , not very useful from what i saw in the last 10 minutes but only kimi 2.5 is best narrative.

maybe it is

[dry stratus]

I have the same issue

[hasty sapphire]

🙁 wierd

[brave frost]

This is what it got for /status:
🦞 Clawdbot 2026.1.25
🧠 Model: anthropic/claude-opus-4-5 · 🔑 api-key sk-ant…qspwAA (env: ANTHROPIC_API_KEY)
🧮 Tokens: 10 in / 254 out
📚 Context: 177k/1.0m (17%) · 🧹 Compactions: 0
🧵 Session: agent:main:main • updated just now
⚙️ Runtime: direct · Think: low
🪢 Queue: collect (depth 0)

Just got message that I've used all my anthropic api credits and it just added $10 more and have done nothing since!

[dry stratus]

Spent 6 hours yesterday with ChatGPT trying to solve it. Chat had no clue

[hasty sapphire]

what model are you using? open ai gpt?

[dry stratus]

Anthropic

[hasty sapphire]

on vps?

[dry stratus]

Yeah. Hosting it with Hostinger

[hasty sapphire]

I have diffrent hosting, but maybe is the problem

I have this in terminal:

[dry stratus]

I have no clue if it has anything to do with the issue. But whenever I run "clawdbot status --all", this says "Pending"

[sturdy marlin]

Hey just a heads up. The quick install doesn't work because the github link is broken

I can't answer that I think you might be having a different issue

[hasty sapphire]

I think i fix it. Idk how but I'm send in the UI chat this:

clawdbot logs --follow

and its start writing

[dry stratus]

ill try

[hasty sapphire]

and again stop working hahaha

[dry stratus]

I got no response

[hasty sapphire]

try spam ne session a few times

a

new

[dry stratus]

still nothing

[torpid pulsar]

check your logs - do you have a working API key set?

my Claude auth appeared to work, but it really didn't and this is what I had happening

[dry stratus]

Where can I check the logs?

in the UI?

[torpid pulsar]

yea

look in lower left of your screenshot above

[dry stratus]

Does this tell you anything? hehe

[frozen star]

is there an update that will make my clawd swap to molt?

[torpid pulsar]

nope. mine had a very clear API error showing. that looks truncated, can you expand any of those items?

[stoic fern]

Anyone knows how to get kimi k2.5 working with openrouter api

[dry stratus]

everything in the log is either "info" or "debug", and I cant seem to be able to epand anything in the logs

[umbral stone]

I am in a similar spot but I can't seem to get the doctor to fix it. I got it after I tried to ask it to add a selfsigned cert so I could access it other places in the json I do see

"gateway": {
    "port": 18789,
    "mode": "local",
    "bind": "lan",
    "auth": {
      "mode": "token",
      "token": "....",
      "password": "..."
    },
    "tailscale": {
      "mode": "off",
      "resetOnExit": false
    },
    "tls": {
      "enabled": true,
      "autoGenerate": true
    }
  },

Good news is the ssl side seems to be working

[pseudo arrow]

I’m also really looking for a guide like that. I’m afraid of breaking everything during the migration.

[sterile condor]

Did you ever figure this out? I'm still unsure on what the webhook URL (in nextcloud) and the baseURL (in moltbot) should be.

[true gulch]

clawdbot status inside terminal?

[rocky orbit]

Just run an update

[dry stratus]

Looks like this

[chilly bough]

I ran an update the command moltbot doesnt work

[rocky orbit]

Use the clawdbot command for now

[wintry oar]

Has anyone had the issue where their clawdbot will alert you to tell you exec commands have stopped working? I have an allow list, but sometimes even those commands in the allow list will stop functioning. Looks like a new session fixes but then eventually breaks again

[true gulch]

There's a lot confusion about this 😞 Inside the docs there are mixed commands. Currently clawdbot it's only that works. But sadly though it's not obvious.

[rocky orbit]

changes will happen progressively. For now you have a free working tool 🙂

[true gulch]

Agent Pending. mmmh Can you check the api key or restart the service?

[true gulch]

Yes, I know. It's not so simple. We will wait 🙂

[dry stratus]

It might be the anthropic API Key, but I have no idea where in the teriminal to change to a new anthropic API key. I have an idea that the issue was that I used an API key from anthropic, but didnt buy any credits

I did now buy 5USD worth of credits on Cladue console, and want to create a new API key and use that, But idk where in terminal to change the Key

[flint relic]

telegram bot responding way too much slowly, any idea whats going wrong?

[true gulch]

clowdbot configure - Local - Model - Anthropic - Antrhopic Api Key

[dry stratus]

what are the commands to run in terminal to get here?😅

[true gulch]

the first, then use the tui

[prime nimbus]

Wondering what people are using for prompt for clawdbot to create its face etc like all these posts?

[dry stratus]

It asks for some "Amazon bedrock model", what is this?

[true gulch]

mmh they are the models, probably it's not the right way to set the new api key 😪

[true gulch]

Try the same with clawdbot onboard pls

[dry stratus]

it asked this after I provided it with a new API Key

but I went rhough these, and now it works

it seems to have been an API key issue

that I didnt buy credits with the API key

[true gulch]

Ok, so you can chat on the dashborad?

[somber zodiac]

how can i set it to use google instead of chatgpt? i cant seem to find the provider id

[dry stratus]

yes I can chat in the dashboard now. And I got it paired with telegram so I can chat with it in telegram now

[thick olive]

help me please

[true gulch]

good news 🙂 I'm curious about clawdbot status. Is it still pending?

[mossy estuary]

Hello i spend all my money on API token plz help I need good cheap model thx

[dry stratus]

status looks a bit different now

[pearl solstice]

Howdy yall. So clawdbot has like seemingly forgot that it has access to tools at all. It was working totally fine today, then I did a new chat because I really fucked up its personality. Now it like, can’t access the browser, can’t access smb shares. Has to be explicitly and repeatedly told what it is and what it can do. It’s just not working smartly and I’m having to basically outline all of its functionality to it and tell it how to use its tooling to get it to work

[true gulch]

Nice 👍

[pearl solstice]

Mind you I’m using qwen3 80 b instruct so it’s not necessarily a dumb model

It’s like it’s got amnesia and has no clue what it’s doing or what it is. It’s lost all of the formatting I gave it when it’s first hatched and it’s just functioning wrong I guess is the best way to describe it. Any suggestions?

[kind hatch]

when i run clawdbot status it says no fts module installed, is this the case for everyone?

[flint relic]

i think telegram has problem, it stucks and doesnt responds

[frozen star]

I haven’t been able to get any version of qwen to work beyond doing a great impression of a goldfish

At first you think you solve it because of the default 4096k tokens but it won’t fix when you change that to 20k

[pearl solstice]

Well like, I remember using Claude code for a bit last night and it took one prompt. One single prompt to write to an excel file and it did it. When I told qwen to do it it took like 8 prompts to convince it that it could do it

I get that Claude code is super smart but surely I can find something in the qwen neighborhood that can do that

[polar elbow]

noob question but I followed the one liner bash installer and got clawdbot not moltbot are they in the process of upgrading or rebranding or whatnot? Do I care right now? Maybe not, but should I uninstall and reinstall or does it even matter?

[kind hatch]

dont think it matter atm

mine is doing the same thing works fine

[orchid vault]

do i have to re run clawd bot every time i open mac

in terminal

or is there easier way to keep it always running

[pseudo arrow]

I use Ollama. After I changed the context to 32k, things finally started working in a decent way.

[prime nimbus]

install supermemory plugin, its WILD then lol

[pseudo arrow]

Thank you for the tip, I’ll check that right now.

[prime nimbus]

https://github.com/supermemoryai/supermemory

[tame knoll]

Hey guys, just starting out have setup a VPS. I was wondering what models you guys are having the most success with? I know opus is good but expensive asf. Looking to see if their are any subscriptions that would cover this testing phase

I just saw there is a models channel 🤣

[random thunder]

guys so can anyone confirm if anthropic is banning users who conect via oauth or is it isolated incidents? anyone here got banned for clawd?

[chilly forge]

how can i fix this error, been stuck on this for a while now

please help🙏

[hallow locust]

Hey team, what's the clawdbow -> moltbot path. Do I just uninsall the clawdbot packages and install the moltbot packages and everything works?

[dreamy nexus]

Anyone succed on running kimi k2.5 working with openrouter api?

[stoic fern]

I got it but it’s unable to do tool calling for some reason

https://discord.com/channels/1456350064065904867/1466197379744272478

[prime nimbus]

treat everything as clawdbot for now, while the devs update the commandline stuff

[kind hatch]

twice in the past week where chrome stops working for gateway dashboard and it goes all blank, only works on safari until i clear stuff on chrome

[prime nimbus]

Just FYI, if you got a github copilot sub, highly recommend using it, this is me in the past 6 hours:

estimated $20 saved, thats a mix of sonnet and gpt 5.2 and 5.1 codex max

[unborn pewter]

i am trying to get my to use opencode (so i can save some tokens) but it is not using it.. anything I should be doing ?

[last jay]

hey guys i need help with vector db for memory if someone knows

[kind hatch]

whats going on w it

[last jay]

getting this : Vector error: Loadble extension for sqlite-vec not found. Was the sqlite-vec-linux-x64 package installed?

when doing memory status

some issue with sqlite

[prime nimbus]

going to assume you've done what it said? "Was the sqlite-vec-linux-x64 package installed?"

[unborn pewter]

kimi k2.5 is more expensive then grok code ( are you sure its better) ..

[last jay]

yep did everything even built it from source

now my only solution to change to chromadb or smth similar

[prime nimbus]

I dont do vectors, they make my head spin 😄 might be worth just dump them all in a help topic on the left, showing what you've done, more luck there i think

[last jay]

well i guess thats call

thanks for the help

[prime nimbus]

I did nothing mate, just trying to get you help quicker, not seen anyone mention vectors in here for a LONG time

You could vibe it and ask clawd why its getting that error lol

it will check its own logs to help troubleshoot it? might be quicker lol

[kind hatch]

i have a diff issue it doesnt load fts5 module..

does urs not give that error when u do clawdbot memory status? @last jay

[prime nimbus]

just in terms of models, i found gpt-5.1-codex-max to be an absolute beast (cant use it for wake up, it will do strange things), but swap to it after and WOW, already creating the avatar lol

[last jay]

the only error i get is that sqlite module or extension

probably clawd nuked it from existence lmao

[prime nimbus]

Maybe https://github.com/supermemoryai/clawdbot-supermemory might be better? Just dropped

[kind hatch]

dude probably thats so weird, mine is messed up too lmfao

[last jay]

i doubt it runs locally right?

its a service

[prime nimbus]

nah not local sorry mate

[mellow sundial]

How do you give it full disk access? I can't find the program to give it to

[last jay]

yeah sadly

[prime nimbus]

Some people if they want the beta dropped yesterday i just found out with the rename commands etc: curl -fsSL https://molt.bot/install.sh | bash -s -- --beta

[unborn pewter]

anyones know how to bind gateway to 0.0.0.0 i lost 1 hours and could not do it and documentation is not helping there. I will not expose it direct to outside, it homelab only

[last jay]

found the solution for mine

[last jay]

you still got the same problem?

if so try going to node modules folder

cd $(npm root -g)/clawdbot

and reinstall the sqlite stuff there

npm install sqlite-vec-linux-x64

if youre using linux

[prime nimbus]

might be worth reading here: https://docs.molt.bot/gateway/remote

[prime nimbus]

so when it said "Was the sqlite-vec-linux-x64 package installed?" it wasnt? lol

[last jay]

it was installed

[prime nimbus]

then here: https://docs.molt.bot/web#public-internet-funnel

[last jay]

apparently one single file got renamed and sqli completely forgot who that is lol

[wintry haven]

Perhaps was installed to a virtual environment, not to the shell (no pun intended) Molt is accessing.

[prime nimbus]

Thats what im laughing at lol wild, "turn it off and on again" good find

[strange chasm]

New to Molt.bot. Frustrated. With just simple conversations I repeatedly hit rate limits using Anthropic so et 4.5. I set monthly limits of $100, and only used $5 today, with automatic credit refresh at $15. I am tier one on limits. Any recommendations, as I can’t do anything what people are describing because it just shuts off. Any help, recommendations?

[prime nimbus]

local ollama if you want completely free, otherwise tap into another subscription

[strange chasm]

Olloma for local LLM? How much spend on hardware?

[prime nimbus]

most people are using mac minis, but depends what you want to use model wise

best to use a decent gpu and plan it for the size of the model you plan to consume the most (8gb should be a good starter)

[strange chasm]

I am using VPS on lockdown- afraid of security issue

[fresh quail]

Same exact problem, i've done everything i can think of to troubleshoot and no avail.

[prime nimbus]

Honest my 2 cents, if you can afford it get github copilot pro, i've been using it all day now, no rate limits, it can wake up on gpt 5.2, then once setup swap to something else, but gpt 5.2 is by far the best

[storm sky]

what's this man?

[prime nimbus]

copilot-proxy to link into github

[storm sky]

i have been having issues with nano-gpt, can't get it to set up for the life of me

any good experience or point in the right direction for that?

[mellow sundial]

How do you give it full disk access?

[prime nimbus]

nano wont work, theres a small limit on the openai models i feel that can we used for beginners, if you can try gpt-4o with wake up

[storm sky]

what do you mean by wake up

[prime nimbus]

well when you set it up, you need to run wake up, my friend! to configure it, set its name etc, the best model i've found in openai is gpt-5.2 for that, 4o is close, sometimes buggy (lots of gateway restarts lol)

[storm sky]

why do most openai models don't work?

[twilit bloom]

Keep struggling to get my bot to successfully complete CAPTCHAS. Any ideas?

[orchid vault]

HI EVERYOne ,clawdbot listens when i say open safari but when i say search for something it doesnt do anything ...does anyone know how to fix

[last jay]

captchas for searching on google or all sites in general?

[orchid vault]

this is via whatsapp

[prime nimbus]

Not too just myself, there is a cheap one that works though just cant remember which one, if 4o is like £7 output 1m with azure, then its that one lol

[twilit bloom]

Signing up a twitter/github account. Creating burners for my bot but would weather it set them up than me

[prime nimbus]

tough time when you need mobile to setup nowadays lol

[last jay]

only solution i think is not using a headless browser

[twilit bloom]

I tried browserless but no luck

[prime nimbus]

Maybe if I got a simcard like a cheap 1gb data, but unlimited calls and texts and implemented moltbot into it? LOL

[twilit bloom]

Headless chromium no dice

[kind hatch]

everyone with memory config setup properly does your clawdbot still get lazy and forget shit?

[prime nimbus]

you could setup a chrome profile to use, you need to set it up in chrome etc login to the sites, like a real person, then save, export into clawdbot, tough to figure the best way for clawdbot to use it though

[storm sky]

when you guys want to add a custom openai provider, where do you start, like what are the step by steps, am i doing smth wrong by jumping straight to the auth-profile

[prime nimbus]

specifically which?

[mellow sundial]

How and what do I give full access to for iMessage? https://docs.molt.bot/channels/imessage#imessage

[storm sky]

nano-gpt in my case. I am pretty sure it'll eventually work, i am just not sure how to get it to work, 6 hours in

[prime nimbus]

best to run through onboard, set it correct there, restart the gateway and say exactly "Wake up, my friend!" without the qoutes, it should start saying some stuff back

[last jay]

well you can still inject cookies and stuff but it aint that secure

[storm sky]

it's basically the same as openrouter with all random models

[prime nimbus]

still same as using burner accounts, just use them burner cookies.. burnt cookies 😄

[storm sky]

i can set custom providers through onboard? didn't know that

[wet oxide]

Anyone having issues getting moltbot to reply back in discord servers/dm? Was working fine then suddenly this morning all it does is emoji react. No changes on my end.

[prime nimbus]

Depends on which first, for example im using copilot-proxy and once i enable the plugin it appears in there

[prime nimbus]

give the gateway a restart first, test, if not check the gateway logs, its verbose in there with errors

[dawn plaza]

Does clawdbot work on Linux?

[opaque cloak]

Hey all! New here. Just got everything installed but I'm also hitting blank responses with OpenAI and Gemini. I can't imagine that both my tokens are wrong. Any thoughts on how to debug? I feel like I've tried everything 🙁

[prime nimbus]

make sure youre using a good model, gpt 5.2 works fo'so

[opaque cloak]

Thats the one I selected

[opaque cloak]

and gemini-3-pro-preview

[gaunt garnet]

I am struggling to get my deepseek (site api key) to work. It works via open router though 😅

[prime nimbus]

restart gateway, run clawdbot tui and say "Wake up, my friend!" without the quotes

[worthy jolt]

I’ve literally uninstalled and reinstalled Clawdbot like 10 times. First Claude Code OAuth won’t work. Then I switch to ChatGPT and the OAuth won’t work there either. So I try using an API key with 5.2, and now it throws some error about “thinking” when I send a message. I can’t even see the bot’s response — it just gets stuck on delay and starts lagging. This is wild.

[opaque cloak]

hmmmm

clawdbot tui - ws://127.0.0.1:18789 - agent main - session main                                                        
 connecting | idle                                                                                                      
 gateway disconnected: unauthorized: gateway token mismatch (provide gateway auth token) | idle                         
────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
gateway connect failed: Error: unauthorized: gateway token mismatch (provide gateway auth token)

[fast hazel]

hey, I am looking for help getting Google calendar and iCloud calendars synced with Clawdbot

[prime nimbus]

check clawdbot doctor, but sounds like connection to ai/model issue

[opaque cloak]

yeah i ran through doctor a few times clawdbot doctor --fix.

[prime nimbus]

try search in help, some threads come back like this:https://discord.com/channels/1456350064065904867/1465685339644629103

[fast hazel]

Ultimately I would like to get my bot acces to my email and calendars, but I can't seem to get that setup, anyone hace any advise??

[prime nimbus]

or this https://discord.com/channels/1456350064065904867/1465197639678034022

[round thunder]

So been poking at this over the last couple of days. HAve a Mac Mini I can set it up on, but when I tried to install moltbot it doens't work, but clawdbot does. Waht is the right command at this point?

[candid magnet]

is it possible to restore everything if i do a clean install on a new computer?

[prime nimbus]

install the beta if you want moltbot command, if not stick with stable and clawdbot commands (remember its beta for a reason)

[prime nimbus]

backup the .clawdbot folder

[round thunder]

No problem. Is there still a set of clawdbot instructions for the current stable? Or just transpose the word clawdbot for moltbot in the current docs.

[prime nimbus]

in stable, just everything keep clawdbot

[round thunder]

Thanks

[candid magnet]

and then just copy and paste to new install right?

[wet oxide]

Nothing visible in logs that would show anything which is unfortunate. Just basically emoji reacts, says typing, but then 2 minutes pass and the typing indicator goes away and it's just nothing from then on. Even for simple messages.

[prime nimbus]

in the web portal, click the drop down and see if you can see discord instance, it should create one and use that if its setup correctly, if not, i suggest running onboard again and resetting the discord stuff

[candid magnet]

getting these error. I remove previous install using uninstall

bizzare cause there should be no current install

[next glen]

Found a fix? Currently having this issue

[opaque cloak]

Nope. Decided to wipe the vm and start from scratch 🙁 going to find a new YouTube video to follow.

[dapper tangle]

I built this custom GPT to trouble shoot Clawd/MoltBot issues, provide set up guidance and give examples of creating skills and setup routines. https://chatgpt.com/g/g-69791a04d92c8191a0208088e45e7737-moltbot-aka-clawdbot-builder

Please providde feedback and suggestions.

[next glen]

I’m getting an Error: ENOSPC: no space left on device.

[prime nimbus]

df -h, which one ran out? lol

[next glen]

I don’t know lol. Im just getting The error. And I have been using the bot lightly 👀

[still bobcat]

what to do here now

[primal ginkgo]

I installed beforew the name change. How do I update and be OK?

[prime nimbus]

Just tested discord couldnt recreate the issue:

[dawn plaza]

This might be a stupid question... But is it possible to have two clawdbots on two different devices, but have them act as one bot?
For example I set one up on a dedicated laptop and the other on my current PC, then when my PC turns off it switches to the laptop and continues where it left off

[still bobcat]

what to do here

[next glen]

Dev root.

[prime nimbus]

Ouch lol, good find, clean that up and restart gateway and off you go!

[next glen]

How do I clean up? The few codes have tried not working

[prime nimbus]

depends what files you got in there? you need to clean up space to continue, cant help with that, only your know wht to clean up (i really dont want to say what you can delete incase its something important if you catch my flow)

i've found imapping to my own domain and smtp works no issues, it has calender, it has emails be free lol

i brought a 5 domains host cpanel reseller thingy, so got all 5 on there, unlimited mailboxes each etc lol, its the only thing i use it for, oh and some backup files for the storage on it lol

maybe gmail is best via the api, as i know gmail allows automations for business, wondering if the way we do it breaks tos and using the api might be the only way

[cunning copper]

Hi all. I’m having trouble configuring the Mac menu bar app to connect to my aws-based node via tailscale. I had originally set it up incorrectly as a headless node but stopped that and now want to give it better access via the app. I ran through the wizard and have the right gateway link but it is giving me an error I can’t figure out how to solve after a couple hours debugging. Anyone run into this?
invalid connect params at /client/id: must be equal to constant: at /cleint/id:L must match a schema in anyOf.

It’s almost like the old gateway credentials are hiding somewhere and it’s trying to use those instead of starting fresh?

Even wiping and rebuilding I can’t get the money initial setup to kick off again

[round thunder]

Is there a set of docs for setting clawdbot up with ollama? There is ollama launch that configures it but that also wants to install via npm. There is the regular onboarding but that doesn't have ollama as an option.

I'm not really sure of the right way to proceed. I have a newly reset Mac mini setup with homebrew installed and node and ollama installed from there.

Not sure of the best way to proceed.

[prime nimbus]

go with ollama launch way, its easiest out the box, once you done that, it will work and will appear in the lists in /models then too

[wet oxide]

Seems to be trying to use Gemini CLI as the main provider is just killing everything and timing things out which is weird. Back to claude now.

[prime nimbus]

very strange!

[long sluice]

Hi, I have successfully installed all the requirements on my Mac and configured the Telegram bot. However, the bot is not responding to any messages. Additionally, when I try to use the AI via the browser, it gives a blank response without any text. Could you please help me identify what might be missing in the setup?

[prime nimbus]

what model? what setup? what did wake up my friend do?

[long sluice]

the last model

[prime nimbus]

https://klipy.com/gifs/bilizz

[fast hazel]

HELP!! I CANT F'IN GET MY GOOGLE EMAIL AND CALENDARS SYNCED!

[long sluice]

gemini-3-pro-image-preview

[prime nimbus]

try the models chat on the left

[carmine sphinx]

Hello, pretty frustrated. Had clawdbot set up with api key. Was continuously hit with rate limits for no reason. Completely wiped my Mac Mini and started over. Claude had me go a different route setting up a token instead of api key. And can’t get it to work. What’s wrong with me?

[glossy valve]

Hey all, I’m sure everyone has asked this a million times. But I am hesitant to use my Claude Max account (and even my OpenAI account) due to the TOS. Not wanting to spend thousands on API, but also want decent ability.

Been looking into GPT5-nano, Flash3, and Kimi.

What models are y’all using in API? How much are you projecting to spend?

[mental mica]

Out of those three, use kimi
Definitely not the other two unless you wanna open yourself up to easy prompt injection

[carmine sphinx]

Are you still able to set it up with a token so that I can use my subscription instead of api key?

[glossy valve]

Okay! I’ll try it. The main thing I want this bot to do is help me organize my schedule, plan my calendar, and act as assistant. Hoping to not let it on the internet too much, at least for now lol. Thanks for recommending

[mental mica]

What are you using for your calendar and stuff?

My advice is to give it a second account and share your calendar for that account instead of giving it your main account

[fast hazel]

Taco me too!, i wish it was easier to config

i cant get it to get my google auth info loaded in to get my email etc

[glossy valve]

Just basic Google Calendar haha. I have a small team, and I almost wanted this thing to be the task delegator for our company. To find times for us to sync on issues, and be able to describe in plain English what all we are doing, and it hold us accountable.

[glossy valve]

Good idea!

[glossy valve]

Haha dang, I haven’t even got to that step yet😂

[fast hazel]

its impossible, been at it for 3-4 hrs trying to get it to work

i feel ive tried everything

[glossy valve]

I’m also debating on making a VM on my main Mac, 128 GB and running local model 24/7, just bothers me knowing how much I spent on this laptop. Considering using my old 16GB Mac and turning it into a server

[round thunder]

Ok thanks

[glossy valve]

I wonder if maybe Claude code could help you set it up

[buoyant yoke]

It will

[glossy valve]

Obviously I haven’t gotten that far yet so I can’t comment fully lol

[buoyant yoke]

I used Claude code to help fix anything that broke during the install

[fast hazel]

spinnin it up!

[buoyant yoke]

Now Clawdbot fixes it

[glossy valve]

Are y’all using ClawdBot to control Claude code too? Is that a violation of their TOS too? Would be cool to have Clawd try and use Claude as a tool

What model are y’all using too btw?

[subtle cypress]

When i install and onboard clawdbot it works but when i do the same with moltbot onboarding it says term not recognised

[sudden nest]

I'm using MoltBot to control Codex CLI

[glossy valve]

That’s sick, so excited to try that

I have both Claude and codex

[subtle cypress]

What is advisable, should i install moltbot on a private vps or pc will also work

[glossy valve]

Unless you have a PC with nothing on it, I think you should do VPS. Just in case

[prime nimbus]

I use proxmox with vm all local, it has a gpu passthrough as well, cheap setup about $100 in total runs abit lol

[carmine sphinx]

Are you guys are able to get Claude max subscription working through clawdbot?

[glossy valve]

I’m too scared to even try, I heard too many ppl getting banned due to TOS

[subtle cypress]

What is advisable, should i install moltbot on a private vps or pc will also work

[nova solstice]

I'm having the same issue. Anyone else?

[prime nimbus]

nope works more me, trust clean cache and all that jargon

[tidal warren]

Anyone have luck with getting signal to work? Any tips?

[orchid ember]

How to handle this?

I already have the API Key set in my auth-profile.json eventhough this is not working

Do I need to restart the Clawdbot or something here..!!

[sinful dawn]

Where are you getting hung up?

[sinful dawn]

When I’d get stuck like that and couldn’t fix it, I’d ask Claude cli to fix it for me. Worked every time.

[sand ingot]

Is there any agents that we can associate with moltbot?

[wintry ibex]

It’s molt.bot