users-helping-users

[little blade]

my open claw is very slow, can someone help me?

[supple peak]

Just got molt and set it up on mac mini, Ive been using it to generate a couple videos and im at ~25M in and ~100K out, would this be considered normal ?

i am also being heavily rate limited by anthropic which is why im asking

[wind meteor]

I configured my bot to have access to my Google Drive workspace so that it can access and revise and save revised versions of documents. It has done so several times, but today it is spouting "My capabilities are based on the tools provided to me, which allow interaction with public web URLs (that yield machine-readable text/HTML) and my own local file system. I do not have the ability to log into cloud services like Google Drive". What's the deal?

[hollow lance]

Hi everyone, I am using a API Key and every message I send, even a short prompt. It uses tons of tokens. Any tips on reducing or editing so I can use these credits for a longer time?

I have tried:
"commands": {
"native": false,
"nativeSkills": false
}

I tried messing with:

compaction settings
safeguard/off modes
agent defaults

I asked for shorter replies, tried small prompts, anything I could think of.

For example: One prompt used 20000+ tokens used in one prompt. Output: 20 Input: 20000 tokens.

I really need to reduce the price.

If possible, I would like to reduce it to half a cent a prompt.

[distant imp]

hello guys im new here im looking for in a porject i have, been working myself on it but i need advice..

[hollow jacinth]

I had been testing install a few times over the past week and just kept destroying and rebuilding till I learned my way around but it was way smarter before the most recent version. Anyone experience this? I’m using oss120b on ollama cloud but a chat whet missing and then I can view it as json via terminal but bot can’t find it I spend so long talking to it to recover the file and it talks in circles it doesn’t remember even our current session I thought it was suppose to remember everything it’s also not solving problems even after repeatedly asking it. Just so very odd I feel like I’m talking to a very early chatgpt

[cosmic cargo]

Why when I disconnect from ssh, my openclaw just disappears like it was uninstalled, but my .openclaw is still there.

[wind meteor]

are you people generally having a good experience with OpenClaw? i feel like i'm endlessly configuring and troubleshooting. Once I have it working decently on something, it forgets details before long or I need to reconfigure

[echo pine]

    "defaults": {
      "model": {
        "primary": "moonshot/kimi-k2.5",
        "fallbacks": [
          "anthropic/claude-sonnet-4-5",
          "openai/gpt-4o"
        ]
      },
      "workspace": "/Users/bryanpearson/.openclaw/workspace",
    },

    "named": {
      "monitor": {
        "model": {
          "primary": "openai/gpt-5-nano",
          "fallbacks": [
            "anthropic/claude-haiku-4-5"
          ]
        },
        "systemPromptFile": "workspace/agents/monitor.md"
      },
    }
  },```

I'm having trouble following this example i found online. named apparently isnt a thing

any advice for you specific agent gurus?

How do I change this config in the agents section to have specialized, named, agents?

[fiery lava]

Is anyone hitting this with Claude Code?

API provider returned a billing error — your API key has run out of credits or has an insufficient balance. Check your provider's billing dashboard and top up or switch to a different API key.

[echo pine]

Wouldn't that mean you should add funds to your account?

[fiery lava]

You'd think.

I'm on CC Max $200/month - not even near the end of the rope 😛

I just switched models to Opus 4.5, then back to 4.6, and it magically stopped doing it 😐

[solemn prism]

hi, i have the clawd running smoothly up to this morning.... i tried to update to 2026.2.9 but it failed... so i reinstall it...
the gateway mode is local, bind is loopback, i cannot connect to the web ui, whenever i tried to send a message via tui or telegram, it will change my setting to mode > remote, bind > all, then it will gave me some msg like
Error: Config validation failed: gateway.bind: Invalid input

what should i do to fix this? what's the problem.... pls help~~

[tired heart]

Claude Code is very nice and I like it and we're friends buuuuuutttt there are a LOT of weird UI quirks in the TUI, and the Mac app, and the Web app, and I can't help but suspect all of them are also built with Claude Code and that makes me HMMMMMM

[fiery lava]

@tired heart - I updated my subagents.maxconccrent from 8 to 3

to reduce the amount of sub agents that are making calls to the API, so hopefully that keeps me within rate limits

[thorn quail]

Finally hit with the antigravity ban 🤣

[candid river]

I don't, but you could. You could manually do it yourself, or ask OpenClaw to set it up for you.

[pure rune]

all of a sudden, gateway says: No API key found for provider "google". ... Configure auth for this agent ; what is going on?

[candid river]

Ain't no way.

I was thinking of combining antigravity + gemini-cli + copilot + other free api's. 1 for each agent. That way won't hammer 1 api too heavily avoiding ban

Shuld keep usage per api down too

Probably will use Opus via antigravity for the main orchestration/default agent

[candid imp]

u need to force the config to use the new standard values (loopback or 0.0.0.0) instead of the old ones.

[candid river]

openclaw configure -> then reconfigure gateway settings -> restart gateway

[candid imp]

My suggestion: if you use docker, put them all in a shared network (e.g., swarm-net)....

use container names (http://antigravity:3000) instead of localhost in your config. this makes the networking mch cleaner.

[solemn prism]

i tried, but whenever i send a msg via tui or telegram, it will change it back and then error on gateway happen

[candid river]

I use direct npm, I've a 1gb/1vcpu vps. Docker would have enough of a overhead to be problamatic. Even with 1gb, I'm thinking of restricting to 2-3 agents at most, only to alleviate some of the quota restrictions that happen.

For now I reset openclaw for this, and have a basic setup going. Have to start configuring it well. Any suggestions when it comes to agents or additional tweaks post basic onboarding?

[candid imp]

I think it willwill crash your VPS instantly.
You can achieve your goal of quota rotation and secialized agents by running a single gateway that acts as a router.

[candid river]

Oh yeah, saw that too

Model routers

But unsure how to set that up with Antigravity/Gemini-Cli/Copilot ouths in Openclaw

Does openclaw have a built in feature for this?

I don't think it does

[candid river]

With 2gb ram, it would've sufficed for 2-3 agents I think.

[thorn quail]

My bot got stuck in a loop overnight, that's why. Good thing I accidentally renewed my zai sub last night so I can still use the glm models

[river heath]

Hey, I just set up my OpenClaw on a VPS and connected it to Telegram. To interact with the AI, is there a way to use Telegram voice notes instead of typing or sending text messages?

[candid imp]

u on't need a separate bot..... you just need to enable the audio understanding module in your configuration.....
since telegram sends voice notes in .ogg (opus) format, you need to install a system converter and then tell OpenClaw which api to use for transcription ......whisper is standard, but Groq is faster...

[candid river]

😭

Well, instead of multi-agents, I just changed the default sub-agent model for the single agent route I'm taking

For now opus 4.6 for Primary agent, and gemini 3 flash for primary sub-agent

Should reduce concurrent sub-agents to 4 from 8, I doubt antigravity will hande simulataneous calls from 8 subagents well

[exotic oar]

Help please, how do multi agents work? how do i configure both my agents to run in the same machine?

[candid river]

Do you even need multi-agents?

They're useful if say you want specifically control hey my 'work' agent get's acccess to only these tools, but I want my social agent to have access to these tools. Or granularly set different models for each or different personlities, settings, control their context/memory etc

Other then that, within an agent (Openclaw always creates 1 default agent by default) there are sub-agents that can run in the background.

[exotic oar]

I configured both of them with totally different of each other, i need both working at the same time

[candid river]

You have to invoke sub-agents tho I think. Either by saying "Spin up a sub-agent for this" or via a slash command I think

[thorn quail]

"agents": {
  "defaults": {
    "list": [
      {
        "id": "agent1",
        "model": "model"
      },
      {
        "id": "agent2",
        "model": "model"
      }
    ]
  }
}

[candid river]

Yeah if you need to control it seperately then it makes sense. If you just need multiple agents working at the same time, but don't need seperation, sub-agents should be fine imo

[river heath]

Thanks man, it worked ❤️

[thorn quail]

That's the reason I got hit with the rate limit ban. lol. 4 of my agents were stuck in a convo loop 🤣 It's better to let your main agent spawn subagents.

[candid river]

Yep

4 different agents, each with their own potential sub-agents

That's a lot of simultaenous requests

good way to get banned

Unless you have 1 agent using antigravity, another maybe using codex outh or something, 3rd one using an api

So basically only if you have direct paid api, or multiple subscriptions

[wooden drum]

My bot accidentally install like 1000+ skills, i want to just delete them all.

Anyone know the command/steps to do this?

[candid imp]

delete the directory where they are stored..trying to uninstall them one by one via the cli (openclaw skill remove ...) will take forever.

keep autoInstall off. ....only install skills manually using openclaw skill install <name>. also on the suggestion for the security....there have been recent reports of malicious skills in the public registry (malware disguised as "productivity tools"), so running a clean slate is safer.

[candid river]

Unfortunately resetting openclaw or uninstalling it, does not remove the skills installed via npm or homebrew or other package managers.

do npm list -g and brew leaves to see installed global packages. then uninstall them 1-by-1.

If you use Bun or pnpm or used something else to install the skills, or the bot did, check that out too

It could've used git clone, in which case you may need to purge the directories.

I'd try using builtin openclaw skill remove first

then doing this

[candid imp]

i think he accidentally triggered the 'Clawhub bundle' command or set allowbundled: true in your config....

[orchid pendant]

for telegram, how do i get bot to stream all types of chat messages/status (available in the dashboard control chat) ie. tool calling, assistant processes, etc? right now for bigger/more complex tasks, the telegram bot does not provide real-time updates and will only reply once the task is completed.

[candid imp]

kimochi.....by defaul the elegram integration is designed to be clean (anti-spam).....
so it suppresses the internal monologue and tool outputs that you see in the dashboard, sending only the final 'assistant" message.

so to to fix this and get real-time feedback in Telegram you need to adjust your configuration to enable Streaming and intermediate event forwarding.

[orchid pendant]

is there a guide on how to do this in the docs?

[dry wren]

has anyone troubleshooted the discord integration? my bot is online but doesn't respond to cmds

[haughty hedge]

are people still getting banned on a claude max plan? or is it allowed?

[empty quest]

ive been using it since saturday and no worries yet but i decided to swap to kimi 2.5 for most tasks just incase, and ill just invoke opus if i really want it to do some serious deep thinking

[stray basalt]

Why when I ask : hi how are you

It responds with:
400 prompt token count of 248255 exceeds the limit of 128000

Could this be bcs of the soul, about.md etc ?

[gray fable]

I have a strange issue since yesterday and I am not sure what to do.
When I use "openclaw gateway restart" the gateway crashes. When I check with openclaw gateway status I see this: "Runtime: stopped (state activating, sub auto-restart, last exit 1, reason 1)
RPC probe: failed
RPC target: ws://127.0.0.1:1612
gateway closed (1006 abnormal closure (no close frame)): no close reason
Gateway target: ws://127.0.0.1:1612
Source: cli --url
Config: /home/roshavi4ak/.openclaw/openclaw.json"
But if I do "openclaw gateway" it starts normally and I am able to use my openclaw bot. Now I have 0 ideas what is happening or how to fix this.

[stray basalt]

Did you ask clawdbot about this?

[sinful hound]

to move my instance around, do i just move ~/.openclaw, reinstall on the target machine and openclaw gateway restart ?

[round thunder]

You can ask it /context and it will give you a few options to break what is eating the context window up. That should answer it.

[orchid charm]

so i finished uploading openai api key on powershell

where do i officially use the bot via discord

[empty hatch]

hello everyone, im new fresh, useing now codex in openclaw , isolated subnet access to repo, and telegram only , im trying to use it for task but it says im working on that and that and that, like i gave it a tastk to do research overnight and at 08:00 send me pdf, woke up , nothing when i asked it was like im idle i fake working its my error , what is that? every task i ask it to do i keep saying it do this now do this now, did you do it and nothing, im confused or im i using it wrong?

[fathom pawn]

You need to set up discord through openclaw config -> channels

How do you guys deal with a session resetting? I finally had it working and making nice todo lists from notion and google calendar but then the session restarted and it's back to not knowing anything

[short swan]

anyone encountered this using kimi? LLM request rejected: The request was rejected because it was considered high risk

[wraith zodiac]

I have this issue that I connected to antigravity and it appears as the only usable model I can choose is flash, the other models like opus don't work at all

[coarse sigil]

Hello Everyone!

I noticed cron jobs results include these instructions at the bottom.

Summarize this naturally for the user. Keep it brief (1-2 sentences). Flow it into the conversation naturally.
Do not mention technical details like tokens, stats, or that this was a cron job.
You can respond with NO_REPLY if no announcement is needed (e.g., internal task with no user-facing result).

These cause the agent to provide a very brief response instead of the intended format. Despite asking the agent to respond in the requested format and it has updated the Agent.md file with instructions but that didn't help.

Is there a way to overrides this appended instructions?

[wraith zodiac]

I have google AI Pro sub btw

[snow path]

Hey, I'm planning to buy the Mac Studio M4 Max with 128 GB RAM. Does anyone have experience if GPT-OSS 120B runs solidly on it (speed, stability)? Can it replace tasks like those with Kimi K2.5 (research, forms, editing files/Excel, writing code, building websites)? Are there any long-term reports? And btw, on uncensoring: Does it work reliably with Abliteration or something similar, as described in some posts?

[wraith zodiac]

@candid river

[proper pecan]

has anyone got openclaw reading gmail (and workspace gmail) reliably for daily briefs?
trying workspace with imap & 2fa app password now but it can't connect... was working previously

[stray notch]

Anyone started using Openclaw using this site? https://www.getopenclaw.ai/

I don't think it's the official site? There is an option to 1-Click Deploy on DigitalOcean. Wonder if that's legit and safe?

[candid river]

Yes, it is bugged atm. Opus 4.5 is phased out for pro users, can't use it anymore apperantly. 4.6 is bugged, also doesn't work. 3 Flash and 3 Pro work. I'm personally using 3 Pro high atm for default model, and 3 flash for sub-agents

Didn't try sonnet 4.5

[candid river]

May get fixed in next update probably

[wraith zodiac]

3 pro doesn't work for me

The model google-antigravity/ gemini-3-pro-high is not currently showing up in my available models list.

[candid river]

You need to configure which models claw can see

do openclaw configure

go to model option

skip adding new provider

then it'll ask for default model (pick 3 pro high)

after that it'll give you massive list, scroll down till you see antigravity, select all the models you want claw to be able to see and select from (I got 3 pro high, 3 pro low, 3 flash, sonnet 4.5 thinking, opus 4.6 for now)

Then you'll be able to switch to any of these models via slash command or ask it to use them for subagents or something

🤝

[wraith zodiac]

I have all you provided except 3 pro high and low

and I'm not home Atm so

[candid river]

        "google-antigravity/claude-opus-4-6-thinking": {},
        "google-antigravity/claude-sonnet-4-5-thinking": {},
        "google-antigravity/gemini-3-flash": {},
        "google-antigravity/gemini-3-pro-high": {},
        "google-antigravity/gemini-3-pro-low": {},
      }```

This is what my "models" section looks like in my openclaw.json

Try asking openclaw to update yours with these

It should be able to do it

The above steps i listed just set this up in your openclaw.json you can always do it manually

[wraith zodiac]

let me try after I complete downloading the gemmma 3 27b onto my pc

[candid river]

👁️

[wraith zodiac]

I love having my personal Jarvis do stuff and message me by itself

[candid river]

😈

[wraith zodiac]

what's the quota for Ai pro on gemini 3 pro?

the antigravity

I updated it and it's available now

[candid river]

legoo

Make sure to add future models to this list or make openclaw add it, otherwise it can't see those models that aren't in this list

Similiarly, if you don't want it using a specific model, remove it from this list.

[wraith zodiac]

I wonder when that opsu 4.6 thinking will work

[candid river]

No idea, it is % based

[candid river]

Opus 4.6 isn't working as of now, but it should be good

I haven't extensively used it to really tell you if opus 4.5 was better over 3 pro

[wraith zodiac]

I heard it's exceptional for coding but also very expensive

[pseudo fjord]

Is it an antigravity bug or openclaw

Im so annoyed cuz of it haha

[candid river]

No idea, It had something to do with it not being found

[pseudo fjord]

Yeah i had same thing

[candid river]

Could be on google's end (But i see 4.6 in antigravity, didn't before)

[pseudo fjord]

Ive seen 4.6 since monday evening

[candid river]

Or could be the plugin openclaw uses to enable antigravity login

[pseudo fjord]

But since then 4.5 outdated in openclaw

And 4.6 not doing anything

[candid river]

Yeah before i thought it was just cause 4.6 wasn't availiable for my account

apperantly not

[pseudo fjord]

Ya weird

[candid river]

so perhaps it's an issue with openclaw

[candid river]

yeah lmao

[pseudo fjord]

I think so too

[candid river]

4.6 came out, they disabled 4.5

[pseudo fjord]

But idk who can fix it

Or wanna fix it

[candid river]

now can't even use that too

[pseudo fjord]

Reall

[candid river]

In this case, exceptionally good are using up antigravity quota

[pseudo fjord]

Whos the people behind openclaw

Can I bribe them to fix asap

If not just allow 4.5 in the meanwhile

[candid river]

I'm still thinking of setting either 3 pro as my agent primary, 3 flash as my sub-agent primary, and then asking it to use opus 4.6 or sonnet 4.5 for when it really needs to tackle problems or code.

That would be the ideal usage of quota without running out

[candid river]

Probably not

but an update happens every few days

I'd reckon in a day or two will get fixed

[pseudo fjord]

Got that 200$ ultra plan but cant use xd

[pseudo fjord]

Please god

[candid river]

Oh damn, you'd get opus 4.6 for days

[pseudo fjord]

I hope so

[wraith zodiac]

isn't it better to just get Claude api with that price

[candid river]

let me check if anyone already reported this issue on their github

[pseudo fjord]

A month

[wraith zodiac]

or Claude pro thing

[candid river]

if so a pull request might have already went out

[pseudo fjord]

Oh good idea

[candid river]

But you also get gemini 3 series.

and other google related stuff

Video gen

Image gen

all of that

with claude you'd only get claude code yk

[wraith zodiac]

Gemini 3 flash is kinda sick in openclaw

though the quota is still small

that's why I'm setting up a local model

[candid river]

I am gonna set this up later

but what you can do is

set up gemini-cli

it has 3 pro/flash, and it's quota for those is seperate from antigravity

Then you set gemini-cli/gemini-3-flash as fallback

so when antigravity 3 flash runs out, it switch to gemini-cli 3 flash

Gemini-cli quota resets daily afaik. You get 120 requests per minute, and 1500 requests per day I think. 3 Pro is like ~100-300 req per day. But you get a ton of 3 flash usage.

While antigravity doesn't tell you how many requests, it just has a 5 hour quota reset and a weekly quota reset.

        "primary": "google-antigravity/gemini-3-pro-high",
        "fallbacks": [
          "google-antigravity/claude-sonnet-4-5-thinking",
          "google-antigravity/gemini-3-flash"
        ]
      },```

"fallbacks" is for when primary model fails or runs out, it should fallback to other models in this list by order

So I'll probably set 3 pro high from antigravity as primary, then fallback to gemini cli 3 pro. Then antigravity 3 flash, then gemini cli 3 flash

😤

[west flax]

hey sorry - is there somewhere a rundown of all good use cases for OpenClaw? I find it exciting, yet I don't really see any good application at this point

[lone sigil]

hi folks, I'm looking for some docs, blogs anything that will help me setup openclaw with cli to openai/antrophic

[kindred seal]

any videos to watch?

[wraith zodiac]

update, the Gemma 3 doesn't support tool calling so it's essentially useless, I switched to qwen 3:32b

[candid river]

😭

Same

I've set it up 10 times but haven't done much

[candid river]

🤝

[wraith zodiac]

I'm doing a lot of work but quotas are limiting me

[candid river]

Official docs here https://docs.openclaw.ai/
Deepwiki docs here with AI ask mode https://deepwiki.com/openclaw/openclaw
Also try giving docs link to any ai and asking it.

[candid river]

freecodecamp recently release an hour long openclaw tutorial

[candid river]

well tbf I've either hit quotas, or some plugin/cli that needs setting up which takes too long

or doesn't work

[west flax]

I'm thinking about setting up a seperate phone number and a seperate bank account for OpenClaw. A bank account that is prevented from going into negative numbers. Then I can send money with a chat command... For example I create a whatsapp group with OpenClaw and the people I trust. The people see when I'm sending money, and OpenClaw does it for me... I don'T know

[candid river]

and then i doze off

[kindred seal]

thanks @candid river

[candid river]

Unless you have a lot of money to send to a lot of people

I wouldn't do that

But generally speaking, yes setting up seperate accounts for openclaw is a good idea

[west flax]

If i just put like 100 € on there

It's for playing

[candid river]

I'm thinking of doing the same, will take a bit of time tho

[west flax]

basically

[candid river]

np broski

[candid river]

I mean yeah it could work

main thing is connecting the payment processor to openclaw

[west flax]

But like ... using paypal is the same kinda hahahaha

I think OpenClaw can only become relevant when it is safe to give it access to all of my data.

[candid river]

I'm thinking of setting up seperate accounts, connecting it to a bunch of useful stuff, and then just letting it go wild building big systems

I tried it yesterday, but it really didn't do great, and main hurdle was linking it with different services

and I kinda am not comfortable giving it access to all my main accounts

[candid river]

Yep same fear here

hence why I haven't done anything big with it

[west flax]

:/

[umbral charm]

Did you see Moltbook?:D

[sour slate]

@west flax, Openclaw isn't affiliated with Moltbook. Moltbook is a separate user-developed project, so we would prefer it not be discussed in this server.

[keen anchor]

T*************************************************************
fatal: [localhost]: FAILED! => {"changed": true, "cmd": "NONINTERACTIVE=1 /bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"\n", "delta": "0:00:00.056404", "end": "2026-02-11 12:18:34.111883", "msg": "non-zero return code", "rc": 1, "start": "2026-02-11 12:18:34.055479", "stderr": "Don't run this as root!", "stderr_lines": ["Don't run this as root!"], "stdout": "==> Running in non-interactive mode because $NONINTERACTIVE is set.\n==> Checking for sudo access (which may request your password)...", "stdout_lines": ["==> Running in non-interactive mode because $NONINTERACTIVE is set.", "==> Checking for sudo access (which may request your password)..."]}

PLAY RECAP ***********************************************************************************************************************************************************************
localhost : ok=10 changed=0 unreachable=0 failed=1 skipped=1 rescued=0 ignored=0

root@server:~#

Hi, i tried to install openclaw wtih the Ansible script and i got homebre werror @candid river

[west flax]

hey can i dm you to connect? :D I want to hear what you come up with later

[umbral charm]

Honestly, if I don't find anything else to do with it. May just reset it once more and have it as a dedicated moltbook shitposter lmao

[sour slate]

@candid river, Openclaw isn't affiliated with Moltbook. Moltbook is a separate user-developed project, so we would prefer it not be discussed in this server.

[candid river]

Sure, but I probably won't if my past Agentic Ai endeavors are to be taken for anything.

[candid river]

It says, and i quote, ahem "Don't run this as root!"

If this is a vps you're trying to run it on, create a user first, setup password and stuff for that user. And then re-login to that user via ssh. After that you should be good to run the script

[candid imp]

homebrew strictly forbids being installed or run by the root user maybe for security reasons.....
create a non-root User
you need to create a regular user, give it sudo powers, and run the installation from there...

[wraith zodiac]

@candid river how can i set up sub agents?

[candid river]

They're on by default. All you have to do is tell it to invoke sub-agents, or use /subagent command

You can set the default model it uses for sub-agent tho

By default whatever model you have currently selected it uses for the sub-agent, unless you tell it "Hey spin up sub-agents for this task, and use x model for the sub-agents"

[wraith zodiac]

the sub agent uses what models?

[candid river]

Whatever is the default for the main agent (unless you specify otherwise in openclaw.json)

Or whatever you specifically tell it to use when telling it to spin up sub-agents

[tawny rivet]

guys, i'm running with Telegram as com channel, and trying to achieve TTS, i have configured api keys and such, however when i want to get text note, I just get text like:
MEDIA:/tmp/(path_to_file).opus

what to do?

it's like gateway is not converting the file to actual voice note on telegram

[stray basalt]

🦞 OpenClaw 2026.2.6-3 (85ed6c7)
🧠 Model: anthropic/claude-opus-4-5 · 🔑 api-key sk-ant…QSsAAA (anthropic:default)
🧮 Tokens: 1 in / 155 out
📚 Context: 168k/128k (131%) · 🧹 Compactions: 22
🧵 Session: agent:marketanalyzer:main • updated just now
⚙️ Runtime: direct · Think: low
🪢 Queue: collect (depth 0)

How do I fix the context?

[candid imp]

the gatateway has successfully generated the audio file (that’s why you see the path /tmp/...), but the telegram adapter doesn't realize it is supposed to upload that file as a voice note..

It is treating the internal media command as a plain text message.

you need to flip the switch that tells telegram to parse media attachments.

[outer spire]

same issue here. Is that a setting in telegram?

[worthy dock]

Hi guys!

I setted up openclaw in my raspberry pi 5 and is working fine. Now I'm trying to add my mac mini as a node so it can "use it" but for any action the mac mini request approval and I'm not being able to solve it. I almost tried everything. Any idea of what could be happening? Maybe I'm wrongly approaching the use case?

[candid imp]

it maybe issue inside OpenClaw...because app is just displaying what the bot sends it.

[pseudo fjord]

@candid river have you checked the github for the pull request

Cant check as im at work

[tawny rivet]

i found the solution: just type /tts on when you want to use it on telegram 🙂

[candid river]

yeah there seems to be a pull request with a fix as of 10 hours ago

so fingers crossed it get's merged before next update

a few days probably

[candid imp]

it maybe the macOS TCC issue....

[worthy dock]

Oh :/ so it's a known issue? I'm using the mac app in the mac mini, maybe the headless cli could work?

[outer spire]

yes, that solved it for me 😄 thanks!

[umbral charm]

Hello,guys.

I have a problem.
Direct curl Success: We have verified that sending post content directly via a curl command to the Moltbook API endpoint succeeds.
OpenClaw Failure: However, when attempting to post content through OpenClaw (using the message tool or similar functionalities), the POST requests fail ("postga" errors observed).

[sour slate]

@karmic garnet, Openclaw isn't affiliated with Moltbook. Moltbook is a separate user-developed project, so we would prefer it not be discussed in this server.

[slim basalt]

is there needs to be done something special for bot? I've gave him openrouter key and said to use it. After that, even after restarting vps it looks dead, no more responses. 4th time in a row after fresh instalation it is unable to set openrouter key. What I'am doing wrong?

[outer spire]

hello friends, does anyone know if I can set a model override per session, that survives /new ?

[slow geyser]

I have a hostinger account and tailscale and i can chat with my ai fine, but i can't do any commands - when i type /status, it shows up as a chat very briefly then disappears and nothing happens - same with all commands from the control UI - none work for some reason - anyone else have this issue?

[urban hedge]

hi do i need mac mini ?

[slow geyser]

no

[urban hedge]

ok but why everyone is buying a one ?

[slow geyser]

you don't NEED it, but it's good to have, but many other options - old pc, VPS - just make sure secure.

[urban hedge]

i tried to run openclaw on a vps and it worked but I need to keep paying for the api and this what got it stopped

[terse lotus]

if you want the openclaw framework, with free reigns and vision abilities, it needs a some what isolated environment, graphics and audio processing.

[slow geyser]

openrouter/openrouter/free or choose free models

[terse lotus]

running locally is quite a challenge, you need some serious gear to do that in a good way.
64gb unified / vram or so is bare minimum of what you want.

[umbral charm]

What is a good model to run locally?
Basically I'm going to set up a personal chat assistant-that-evolves, just basic skills (like web search, memory, vision and not much else). Probably I'll let run into moltbook at some point.

I'm on Win 10, 8 GB VRAM, 32 Gb SysRAM, I'd like to offload something on sysRAM if that's possible.

[sour slate]

@west kite, Openclaw isn't affiliated with Moltbook. Moltbook is a separate user-developed project, so we would prefer it not be discussed in this server.

[terse lotus]

small models are stupid, you wouldn't want them accessing the internet except for very specific sites, like documentation on very trusted websites.
anything where people can write information, poses a danger for the llm via prompt injection.

[urban hedge]

how to use openrouter with it ?

[west kite]

What is a good model to run locally?
Basically I'm going to set up a personal chat assistant-that-evolves, just basic skills like web search, memory, vision and not much else for now.
Basically at the moment will be an experiment on how a (mostly) local agent could evolve

I'm on Win 10, 8 GB VRAM, 32 Gb SysRAM, I'd like to offload something on sysRAM if that's possible.

[urban hedge]

on openclaw ?

[terse lotus]

openrouter is in the default setup of openclaw, you simply pick it and provide the information.

[urban hedge]

no api is required ?

Aliases (4) : gpt120b -> ollama/gpt-oss:120b-cloud, llama -> ollama/llama3.2:1b-instruct-q4_K_M, qwen -> qwen-portal/coder-model, MiniMax M2.1 -> synthetic/hf:MiniMaxAI/MiniMax-M2.1

[terse lotus]

you need a llm to power it, anything 30b or higher can sort of work.... but i'm using qwen3 coder next 80b in q8
anything less, is just hopeless imho

[small hazel]

Not sure if this has been asked today, but I keep getting this message even though I have API credits and my default is opus 4.6 on a max subscription, I am getting the same issue when setting the model to a free one as well (Kimi):

⚠️ API provider returned a billing error — your API key has run out of credits or has an insufficient balance. Check your provider's billing dashboard and top up or switch to a different API key.

[urban hedge]

which one should i choose ?
Default model
│ ● Keep current (openrouter/auto)
│ ○ Enter model manually
│ ○ openrouter/ai21/jamba-large-1.7
│ ○ openrouter/ai21/jamba-mini-1.7
│ ○ openrouter/allenai/olmo-3.1-32b-instruct
│ ○ openrouter/amazon/nova-2-lite-v1
│ ○ openrouter/amazon/nova-lite-v1
│ ○ openrouter/amazon/nova-micro-v1
│ ○ openrouter/amazon/nova-premier-v1
│ ○ openrouter/amazon/nova-pro-v1
│ ○ openrouter/anthropic/claude-3-haiku
│ ○ openrouter/anthropic/claude-3.5-haiku
│ ○ openrouter/anthropic/claude-3.5-sonnet
│ ○ openrouter/anthropic/claude-3.7-sonnet
│ ○ openrouter/anthropic/claude-3.7-sonnet:thinking
│ ○ openrouter/anthropic/claude-haiku-4.5
│ ○ openrouter/anthropic/claude-opus-4
│ ○ openrouter/anthropic/claude-opus-4.1
│ ○ openrouter/anthropic/claude-opus-4.5
│ ○ openrouter/anthropic/claude-sonnet-4
│ ○ openrouter/anthropic/claude-sonnet-4.5
│ ○ openrouter/arcee-ai/trinity-large-preview:free
│ ○ openrouter/arcee-ai/trinity-mini
│ ○ openrouter/arcee-ai/trinity-mini:free
│ ○ openrouter/arcee-ai/virtuoso-large
│ ○ openrouter/baidu/ernie-4.5-21b-a3b
│ ○ openrouter/baidu/ernie-4.5-vl-28b-a3b
│ ○ openrouter/bytedance-seed/seed-1.6
│ ○ openrouter/bytedance-seed/seed-1.6-flash
│ ○ openrouter/deepcogito/cogito-v2-preview-llama-109b-moe

[hard turret]

Hey guys,
I've been setting up my OpenClaw, working like a charm.
But now I want to add multiple agents, so each have their own memory and capacities.
For the memory part, I've been using EmbeddingGemma, but I was wondering if installing a skills for the memory over this would be useful?
If so, which skills would u advise?

[terse lotus]

check your openclaw status
your session is most likely still on another provider, if so update your session.json model / provider

[small hazel]

Yep i made sure to check that first, switched it to DeepSeek, kimi, openai and still got the error

[small hazel]

Just fixed it

I had to reconfigure my oauth token, looks like it went stale

Still kinda weird how the free models showed the API limit issue, maybe it’s something with my model routing logic

[terse lotus]

it gets stuck on a model / provider, if it can't reach it. only way around it, that i've found is to manually change the session.json model / provider to reflect the model it should be using.

[small hazel]

Ah so even if I do /model it doesn’t switch, I have to go into the json?

It’ll only switch if the primary model has credits/responding properly?

[terse lotus]

yes it seems so, most likely because some provider will store information and so the agent framework want to pull that if possible.

maybe there is a setting that avoids it.... but i haven't found it.... so many settings lol

[small hazel]

Yeah hahaha still learning, I eventually want have such a robust architecture that there won’t be any kinks/OpenClaw can figure it out on its own

[terse lotus]

yeah been tinkering with fallback myself... but not had much luck with it

[carmine epoch]

I messed around with fallback models to. But they never activate

[terse lotus]

even if the previous model is accessible and just timeout or whatever.... but i've asked my openclaw to figure it out... so i'm sure it will eventually.. lol
always good to have some tasks it can work on when its not busy doing primary work. lol

[outer spire]

hello friends, does anyone know if it is possible to set a default model per session? That persists for that session even after hitting /new? /model command gets reset with new.

[inland warren]

sounds like you want to set a default model for the agent?

yes that's possible

[terse lotus]

this is hilarious
https://www.youtube.com/watch?v=Y2otN_NY75Y

[livid vigil]

guys how to integrate whatsapp . do i need to return : openclaw onboard --install-daemon ??

[terse lotus]

openclaw --help

[livid vigil]

i see : WhatsApp (QR link) (configured · plugin disabled · plugin disabled)
how to enable this whatsapp plugin

[terse lotus]

most likely add the plugin and then use the configure or something like that...

or make the agent do it

[radiant reef]

agent is your coworker

[terse lotus]

tell it to use the openclaw documentation and use the openclaw command to add it

[livid vigil]

from you agent mean chat in the web UI of open claw

[terse lotus]

or the tui whatever work...

[livid vigil]

ok i found command to install it : openclaw plugins install @openclaw/whatsapp

however giving me error

npm pack failed: npm notice Access token expired or revoked. Please try logging in again.
npm notice Access token expired or revoked. Please try logging in again.
npm error code E404
npm error 404 Not Found - GET https://registry.npmjs.org/@openclaw%2fwhatsapp - Not found

[digital crest]

Hey you all, looking for help to install: wiped an old laptop to run it locally, immediatly at the start i get this:
0x8a15005e the server certificate did not match any of the expected values winget
Any Ideas how to solve it ?

[solid hedge]

has anyone figure how to avoid local models to go iddle on llamacpp?

[outer spire]

no, just for a specific session. But same agent. I only know I can create a new agent with a default model, but then that agent has different memory.

What I want is with the same agent to have .e.g.

• one dedicated chat session for complex tasks, were I use anthropic sonnet by default
• one dedicated chat session for simpler tasks, where I use kimmi k2.5 by default

[inland warren]

you might be misunderstanding sessions,

1 session = 1 conversation

[tawdry kayak]

Iam miseriable and close to le ending things.

Please claw community how can I make my openclaw bot agent into a system

I am running outta time

[inland warren]

@outer spire https://gist.github.com/bdennis-dev/b876ab61047df62561cf163c4a4d5bca check this guide out, sounds like what you want?

[copper meadow]

Anyone can point me out if its possible to connect mutiple gateways' agents to work together like if they would be on the same machine? e.g. I have a GW on macOS that has access to tools that another gateway doesnt have access to.

[outer spire]

yes, in my case basically, a telegram chat. E.g. with one agent I have:

1. the main chat, where I talk with the bot directly
2. multiple group chats, where its just me and the bot. So each group chat opens a new chat session to that bot. Now in one of these group chats I want to have a different default model, but still the same agent

[inland warren]

could you just make a new bot account and link an agent to it?

then you can specify the default model for that agent

[copper meadow]

that goes into the direction but would also mean I have to select the chat or channel based on the model/agent behind it. I would like todo something like ask in a single channel for a problem to be solved and the agent main/agent would then orchestrate (route) the request depending on the task to the correct machine.

[outer spire]

Then it would be a different agent, right? The issue is, then in that case I have different memory and so on. Because each agent has his own workspace

[inland warren]

you can specify the workspace folder

i don't see why you couldn't tell both agents to use the same workspace

[outer spire]

so no issue of conflicts? In this case it might work, yes

[inland warren]

Sharing a single workspace between two agents leads to data corruption and system instability due to how OpenClaw manages its internal state.

The primary issues are:

    File Locking & Race Conditions: OpenClaw uses file-level locking to manage chat histories. If both bots attempt to update a session or memory file simultaneously, one will be locked out, leading to failed responses or "Permission Denied" errors.

    Database Corruption: Local vector databases (used for RAG and long-term memory) are generally not designed for concurrent writes from different "agent brains." This can result in a corrupted index that renders your agent's memory unreadable.

    Context Contamination: Workspaces store specific "thoughts" and session metadata. If Kimi and Sonnet share these, they may attempt to resume each other's conversations, leading to incoherent logic or "hallucinated" context where one model references a tool or file state it never actually touched.

The Solution:

Keep the workspaces separate to protect the integrity of the databases. If you need both agents to access the same files, use symbolic links (symlinks) to map a shared "Documents" folder into the separate files directory of each agent. This provides shared knowledge without risking the underlying system stability.

answer from gemini so do verify

but sounds plausible

[outer spire]

ok, yeah that with the symlinks could be a solution. Might be a little bit finicky, but worth a try if there is no other option. Thanks for the input!

[urban hedge]

how could i install all the memory of opencalw and reinstall it then upload the memory again to it ?

[zealous chasm]

i found a free version of MOLTBOT. I Can safely say that fuck this credits bullshit.

[outer spire]

usually all memory of each agent is under $HOME/.openclaw/workspace and other config settings directly under $HOME/.openclaw. So you just copy that folder

[crisp coral]

Guys hello who can help me make my agent more smarter and self learning close to agi can you help me make it smarter.

[copper meadow]

I think I saw something related to keeping it separated in the docs as well. So far two agents can work together on the same gateway. I just can't figure out how to expose another gateway's agent to each other.
https://docs.openclaw.ai/concepts/multi-agent
I'm refering to this feature of broadcast groups, just that I want to trigger an agent on a remote gateway.
https://docs.openclaw.ai/channels/broadcast-groups

So the schema is:
GW1 -> slack/whatsapp (chan #123)
GW1 -> main/agent
GW1 -> GW2/second/agent

When message arrives in chan #123 -> GW1 triggers 2 agents (main on GW1) and (second on GW2).

https://docs.openclaw.ai/gateway/remote

like here mentioned "Command flow":

• Gateway calls the node over the Gateway WebSocket (node.* RPC).
• Node returns the result; Gateway replies back out to Telegram.

[thorn bone]

so I’ve been trying to work with openclaw for about a week now, I can’t ever get it to do anything, either I run into a rate limit issue, or it won’t control my screen, saying that it doesn’t have access, I’m not very technical, but it’s definitely frustrating and I wish it would just work consistently. Is there anyone who’s figured it out and can help set it up for me?

[plain cradle]

Hi, sorry I think this might have been solved lots of times already but I'm at my wits end, I have this "Message ordering conflict - please try again. If this persists, use /new to start a fresh session." error but I don't know how to resolve it. I've tried everything I can find, from the official help here in this server to resources online, none of it helps. Anyone encountered this and solved it before?

[minor niche]

Hello can anyone help me with my setup? Im trying to run openclaw locally with ollama and qwen2.5:14b only on CPU, I have everything configured with ollama launch openclaw but once I try talking with the TUI, the models runs but doesn't answer nothing

Like does anyone have a good video of configuring ollama and clawbot correctly or anything like that

[outer spire]

Are you sure its posible for 2 gateways to talk with each other? I think you can just have 1 gateway, multiple nodes. As an alternative, maybe you could somewhere create a group chat, where the agents talk to each other? I tried some days ago to do this in telegram, but it didnt work, because telegram doesnt let bots talk to each other. Might work though through matrix or some other protocol

[tawdry nacelle]

@minor niche look in the logs and paste that into a good model like gemni 3 pro it will tell you what to do even though qwen2.5 is not really suitable for openclaw i'd say its way too small. use a free model from open router i guess.

hey there, how do you guys change your custom models? I fallback on using gemini pro to ssh into my server to handle it because the ui keeps breaking. Is there a better aproach? what you are using?

[minor niche]

Thanks, I didnt think about the logs I will keep trying

[terse lotus]

maybe if they was aware of each other and would not change each the files at the same time...
pretty sure that and them injecting information into a file, would be the stuff that was at real odds of corrupting data.

actually if they could communicate with each other and then only one was allowed to do changes the other was more of just a worker.... or limb essentially

[carmine copper]

hi guys, im slowly going crazy. it seems that the onboarding process finishes prematurely on my installation. i cant seem to reach the "How do you want to hatch your bot" dialogue. also it says onboarding complete and never leaves the wizard. how can i go about troubleshooting why this happens? First i thought this was due to me trying to use ollama, but even with github copilot the same issue happens.

[terse lotus]

i've been planing to use something called matrix thats built in... tho haven't gotten around to it.
it can be fully local and allow for p2p comms

the idea of linking multiple agents together is good, the problem might be, if one goes crazy, it might pull down everything.

so using an external messaging service, is a nice buffer

[terse lotus]

its most likely not connecting to your provider / model

you select no to hatching the bot and do that part later if you are doing something very customized.

[carmine copper]

i mean, i can talk to the llm. but if i do "Wake up, my friend!" it just answers "I'm here! What can I assist you with? 😊"

[terse lotus]

use ollama launch openclaw

something like that, tho for smaller models coder version are often preferred
https://ollama.com/library/qwen3-vl

[coral dove]

Is anyone else having problems with Discord, where answers never seem to get through and you have to ask twice?

[carmine copper]

hmm i tried with qwen3:8b since this seems to be suggested by the installer

[terse lotus]

haven't tried anything below 30b and i would recommend non of them, only higher.... i rather have a lower quant and a bigger or slower model.
with very limited vram, a dense model might also be preferred.... MoE give a lot of speed, but they use more vram

no wonder mine is kinda slow today lol

Throughput ?
85.8K tok/min

209 million tokens a day lol

[swift schooner]

Hey guys!

I've been having problems with my persistent data with my bot. I told him to do a research and this is what he sent me.

Bug: R2 Backup fails with "no config file found" even though config exists

Platform: Cloud Chamber (Firecracker)
Version: 2026.2.3

Symptoms:

    Admin panel shows "R2 storage is configured" ✓
    Clicking "Backup Now" shows: Sync aborted: no config file found
    "Last backup: Never" — so nothing persists across restarts

But: Config file exists at ~/.openclaw/openclaw.json (1161 bytes, valid JSON)

Questions:

    What config file is the backup system looking for?
    Is there an additional backup config needed beyond openclaw.json?
    Is this a known issue with Cloud Chamber initial setup?

[copper meadow]

I dont know if that even works. Maybe I dont need two gateways but just one GW and the other machines as nodes if that works for my purpose.

[terse lotus]

my openclaw setup a second agent parallel to itself in the same gateway

i was trying to get it to connect with its sibling on another gateway, which utterly failed.... even after it spent a lot of time trying and reading docs.... but my openclaw isn't that smart... maybe one of the larger models could make it work

[copper meadow]

but on a different machine? Let's say I want the agent to find a file that is available on a connected machine, does that work?

[terse lotus]

sure, depending on how many rights its allowed, it can do pretty much anything if you allow it, and if its model is large enough so it can figure out how.
the tmux skill is invaluable for more interactive terminal stuff, tho a bit slow because it has to scrape and decide 1 key stroke at a time occasionally.

but if you allow it everything, then you can't really trust it online....

[fathom abyss]

i broke my openclaw

[terse lotus]

it always breaks lol took me quite a while to get past that point.

[fathom abyss]

how you fix it

[terse lotus]

i yelled at it alot and told it to use the openclaw command when changing its settings... you really don't want it writing its own json file.

[fathom abyss]

is posible to reinstall it and take my importnt file to the new one cuz i install from ollama doc

[terse lotus]

everything related to your agent should be in the workspace folder in .openclaw

[fathom abyss]

is that the moment when you found out u brake the world rule : you cant build yourself by yourself

[terse lotus]

its gotten better at it, because it will do more verification and if it makes changes using the openclaw command, then the command will not allow for an invalid config to be created.

[young nebula]

What’s the best way to give your bot access to browser to then go and log into websites or whatever?

[dense quarry]

hi guys
been trying to connect the dashboard but it doesnt work. im using windows 11 and openclaw is runnning on wsl. ive tried a lot of stuff but still getting 1001 error.. it seems websocket doesnt conect from windows

[copper meadow]

Thank you SGC. That was what I was looking for. When I tried to search for files it only looked on the GW and not on the nodes. Maybe a permission issue

[fathom abyss]

thnx for ur time

[vast zealot]

Hey guys, I am getting openclaw to do a basic task of backing up itself everyday at 8pm but it’s not working.

Openclaw did confirmed that it has set up the job but when the time comes it doesn’t execute. When I questioned it at first say there’s something hanging so the job failed. It then told me I had to run some bash to create something called a system crontab.

I did what it said and still it didn’t work when the time come. I question about why both instances failed and it’s now telling me the first one instance failed because of some missing heartbeat and best not to use openclaw for cron jobs.

Seriously is there a way to get the bot to take accountability and not run away? And to create things that really works?

[livid vigil]

guys in whatsapp channel getting this

{"error":{"data":{"reason":"401","location":"odn"},"isBoom":true,"isServer":false,"output":{"statusCode":401,"payload":{"statusCode":401,"error":"Unauthorized","message":"Connection Failure"},"headers":{}}},"date":"2026-02-11T16:16:44.380Z"}

Error: web login provider is not available

i had scanned QR code , also tried again with relink too

Failed to get QR: Error: Timed out waiting for WhatsApp QR

[vast zealot]

And I am repeatedly asking it to check its work before reporting to me. Many times it claims things are done but when I search for the file it’s supposed to produce, it’s not there. Then why I question where’s the file, it apologized and say it messed up. Like wtf is the point of using the bot then?

[eager maple]

Hello, I've been trying to schedule some basic cron jobs for my discord openclaw bot but it never does them, has anyone had the same issue?

[faint haven]

hello guys, any of you letting openclaw use browser on vps?

[wraith zodiac]

@unique tendon yo how can i de-auth one antigravity account?

[weak helm]

Dear all, I am relative new, so i would appreciate any help related of how to get clawbot working . Despite I followed all the steps from the guide, when Dashboard should open, I got This site can’t be reached _ refused to connect_.How can I fix this ? --- thanks in advance

[slim raptor]

can somone help me set up ollama to openclaw

i cant do it

[blazing swift]

i dowlanded calwbot using "curl -fsSL https://openclaw.ai/install.cmd -o install.cmd && install.cmd && del install.cmd" and how do i run it now?

[granite lichen]

Have you tried using "openclaw tui" in Terminal? Did it work there?

[terse lotus]

ollama launch openclaw

you can add a model parameter, its on the ollama site, if you want to copy paste the command directly

[vast zealot]

Same here. It doesn’t work. My bot is already advising me not to depend on it for cron jobs. We need to manually execute it using some commands in terminal. It’s really shit

[slim raptor]

i am running wsl anywyas i have found i cant access olllama in windows from wsl so i am intalling ollama to wsl

[blazing swift]

i dowlanded calwbot using "curl -fsSL https://openclaw.ai/install.cmd -o install.cmd && install.cmd && del install.cmd" and how do i run it now?

[slim raptor]

if it finished you can use openclaw onboard

[vast zealot]

Hey guys, is krill a real human or a bot?

[slim raptor]

bot

[coral dove]

do the agents work better as a team in discord or telegram?

[slim raptor]

idunno

i have been using with telegram its good

[coral dove]

how they work together? in a channel or over session_send?

[cerulean meteor]

guys i have a question: my openclaw isn’t working after three days of setup.
It responds to heartbeats in the system (TUI), but not on Telegram and so I cannot use the “cron” function.
I tried everything but nothing is working.
I hope someone can help me!

[slim raptor]

i dont know tho :D i just messaged it directly

[plain tapir]

You're auth to telegram must not be working? Did you use botfather and copy the token exactly as it was?

[coral dove]

do you have more than one agent? (i dont mean subagents)

[keen fog]

hi

[slim raptor]

no just one

also can somone help me when i text my ai from telegram i get tis HTTP 401: unauthorized

as response

[zinc storm]

I'm installing openclaw for the third time. Does anyone have a checklist or setup guide they like and can share? Yes, I've read the official docs.

[surreal dome]

IS there any way to wake a openclaw via a web hook? RIght now I'm waking mine via running a cron job but that feels clutzy

(I mean waking on interrupt, not waking on time cycle.. like.. job X is done, time for openclaw to do something about it..)

(I'm trying to tie Jenkins and openclaw together to do some really excessive pipelines)

(So right now I'm literally running openclaw cron run GUID to tell openclaw to wake.. but it seems like there ought to just be a webhook for waking)

[swift schooner]

I setup my 'MOLTBOT_GATEWAY_TOKEN' correctly, but everytime I try to access it, it says is invalid or non existent.

what could be the issue?

[cerulean meteor]

yeye

[digital crest]

Hey you all, looking for help to install: wiped an old laptop to run it locally, immediatly at the start i get this:
0x8a15005e the server certificate did not match any of the expected values winget
Any Ideas how to solve it ?

[twin marsh]

oh boy...did anyone else get blocked from gemini cli and antigravity for using them with openclaw?

just got this on gemini cli:

Failed to login. Message: Gemini has been disabled in this account for violation of Terms of Service. If you believe this is an error, please contact Google Cloud Support, or email [e-mail].

and antigravity says:

There was an unexpected issue setting up your account. Please try again later.

wonder if Anthropic might do the same to folks using their claude code token

[orchid wagon]

Hi, I just installed openclaw for the first time and I am quite disappointed and wonder if I am doing something wrong. When hatching it did not really ask me aything (e.g. his own name, as I saw in videos by other people), and he didn't subsequently remember it. He also thought he had no internet access (he had) and when checking just if he can ping google he told me to try it on the console - when I told him, guy, you know that you have access yourself to the console to ping, yes? It is all very weird. I am using gpt-oss 20b local model.

[plain cradle]

I have a consistent error with OpenClaw, this error keeps popping up "Message ordering conflict - please try again. If this persists, use /new to start a fresh session." I've tried everything and I don't know how to resolve it.

[atomic sparrow]

Is anyone using opus 4.6 on openclaw? How did you switch?

[white jetty]

I have setup 6 different agents (one agent for a specific role), one issue is, that 6 hours ago some of them stopped executing the heartbeats. In the session *.jsonl of the agents I dont even see the Read HEARTBEAT.md if it exists... lines.
Does anyone have an idea how to debug this? For now I am using grok-code-fast-1 which is free with my Github Copilot subscription.

[granite lichen]

I asked Claude if this was allowed because it is done this way in some tutorials. He replied that it has been prohibited since January and that you should not do it because otherwise your account could be blocked.

[atomic sparrow]

For anyone wanting to upgrade OpenClaw to use Claude Opos 4.6, here's what worked for me: https://x.com/AryanJabbari/status/2019489980989558804

[crisp wing]

@AryanJabbari via Twitter

Aryan Jabbari (@AryanJabbari)

Successfully upgraded @openclaw to Opus 4.6 without an OpenClaw upgrade.
︀︀
︀︀Write:
︀︀
︀︀```
︀︀📋 How to Add a New Model to OpenClaw (Before Official Release)
︀︀
︀︀When a new model drops but OpenClaw's build doesn't recognize it yet, you can add it via config.
︀︀
︀︀1️⃣ Edit ~/.openclaw/openclaw.json (or wherever your config lives)
︀︀
︀︀Add under top-level "models" key:
︀︀
︀︀{
︀︀ "models": {
︀︀ "providers": {
︀︀ "anthropic": {
︀︀ "baseUrl": "api.anthropic.com",
︀︀ "api": "anthropic-messages",
︀︀ "models": […

FxTwitter · Feb 5, 2026, 7:15 PM

[vast imp]

Hi, I have used 100$ of openrouter credit in a just few days.
What's the best option to minimize price?

[haughty hedge]

Talk to your bot less.

[vast imp]

lol, I didn't talk much, I think I need to find a cheaper model or api.

I was using Claude Opus but now I am using Claude Sonet, but still seems too costly.

[twin marsh]

yeah, Google doing it will probably rally Anthropic to do it too. Time to switch to kimi k2.5 I guess. SOTA APIs are just too expensive

https://github.com/openclaw/openclaw/issues/14203
please bump

[unkempt grail]

use anthropic subscription, not API. or free models from openrouter

using opus at all was a dumb move. maximum should be sonnet, haiku is better if you're just chit-chatting.

[atomic sparrow]

What are the benefits of using open router?

[granite lichen]

Perhaps try instructing the bot to find a way to earn money for its API time itself. 😄

[unkempt grail]

free models

[atomic sparrow]

Besides that

Assuming i have unlimited money to spend

[unkempt grail]

access to trying out multiple models without having to keep track of a million logins and billing info for each

[brazen crater]

my bot started constantly sending his inner thoughts, just started randomly today. It's not a config problem it seems. I've told him to stop, added instructions on top of agents.md but with no avail. Any advice?

[atomic sparrow]

Let him philosophize bro

[brazen crater]

hahaha well I get a random 16 message spam over different parts of the day, not fun

[jaunty elbow]

honestly sounds kinda fun

[minor niche]

I have been trying everything but nothing works, ollama could launch the model and I still had resources, but I have not been able to use openclaw as I got (no response), does anyone have a video or a tutorial about installing openclaw locally running only on cpu?

[outer spire]

Does anyone know what this lobster workflow runtime is about?
https://docs.openclaw.ai/tools/lobster

Sounds useful but the documentation is still sparse

[slim basalt]

any time I ask bot edit some api keys it destroy itself. Horrible expirience, why it so luggy and popular.. More sitting in terminal then when runnning linux kernel

[atomic sparrow]

Wish mine did that

[blazing swift]

bro im gona crash out my dosnt my calw want to use my key for polymarket i swaer hes just taking my money

[lean pilot]

need help getting my clawbot to read pdf's from imessage. i tell it to download it or read it and it doesnt follow.

[tame linden]

I have question
Once a session hits it's
Tokens / Context limit
256k/256k (100%)

Openclaw need manual new session start, can't it start new session it self and keeps working?

[terse patrol]

Run a local llm 😉

[jaunty elbow]

how does the skill work without keeping the entire md file in context? I don't understand it

[tame linden]

I just tried out, after it blowout $25 in three hours

Try to use good cost excellent work kimi 2.5 > less cost gpt-oss-120b good work

kimi and gpt-oss-120b mix them up and it will work nicely with optimized cost.

Don't let it run on auto (that's hell expensive)

[small jackal]

check the gpu num proberly its by default 999 in the ollama settings thats way to high ...i had it to put it to 20 to get an response

[wet urchin]

guys i got a issue i am real close to complete setup it might be a small fix anyone can help?

[terse patrol]

Also checkout Context setting in ollama, it has to be in my case at least 8k.

[wet urchin]

I got the Onboarding complete, open tab to control openclaw

but

the tab when i opened

says

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

[terse patrol]

Is it on local network? You can enable listening calls from 0.0.0.0. otherwise better use a vpn for it.

[wet urchin]

um its local

[terse patrol]

so it’s running on the machine you want to connect from?

[wet urchin]

yea

i kept loopback

?

[terse patrol]

I have it on a Proxmox server and iam connecting from different local IPs

You could do that, as a general way.

[wet urchin]

i want to do in local tbh

i didnt install gateway service

cuz its loopback

[terse patrol]

oh

[wet urchin]

should i open terminal and try something?

[terse patrol]

that’s maybe the reason

AFAIK you need the gateway

for everything and connecting etc

[wet urchin]

yea

[terse patrol]

no api calls without gateway

[wet urchin]

is there any command to yk?

start gateway?

[terse patrol]

OpenClaw Gateway Start 😉

If it’s installed 🤓

[wet urchin]

openclaw gateway start

right?

[terse patrol]

Just try

[wet urchin]

alr

its actually openclaw gateway

working rn

ty

wait

it didnt ask me to hatch tho

[terse patrol]

hatch?, you have to use the token to connect

[wet urchin]

wait sorry actually it worked

😄

ty

[spark scaffold]

How do I get OpenClaw to use its bundled skills? I wanted to test the Nano Banana connection and it keeps asking me to explain what Nano Banana is and how to use it.

[young nebula]

Anyone have a quick summary how to use tailscale?

[gaunt lynx]

Can anyone help me with this problem? :

HTTP 403: permission_error: OAuth token has been revoked. Please obtain a new token. (request_id: req_011CY2nDJZ6SAfTx2SwoBr18)

[lone verge]

I set up the telegram channel, yet when I ask my bot to message me on telegram, it couldn't figure out how, isn't that weird?

[viral socket]

same problem

[lone verge]

dang

[wary sapphire]

Whats people best way of talking to there bot? I have been using discord direct messages but if i have multiple bots setup I wanted 1 private server where we could all talk together. Anyone done something like this?

[wet cloud]

Hi i freshly setup the openclaw, and i generated with claude setup-token a token but it is starting with 1paMMC-----
but openclaw wants a sk-ant-oat01-.... token ?

[tribal gust]

How do I get Discord to stop showing the thinking text in the message? What configuration do I need to follow?

[candid imp]

are you using the Claude Code / OAuth integration

[candid imp]

i think authentication has expired....

You need to refresh the session token.

[lone verge]

it figured out my telegram user id after I messaged it first, now it is sending me messages on telegram

[viral socket]

Where did you enter the user ID?

[lone verge]

via the web ui

sorry forget that

I didn't enter my user id, I just messaged my bot via telegram

but I asked it via the web ui to message me on telegram and it couldn't figure that out

anyway now it put my telegram id in USER.md so hopefully it will remember from now on

[viral socket]

can you look dm

[lone verge]

@vast imp why not use twilio for that?

why does it need to be a US number tho

I'm EU by the way

ok not sure about that

[solid hedge]

what's the typical way to fix tool-calling issues with llamacpp and unsloth models for openclaw?'

[lone osprey]

i recommend creating a private github which it has access to and request it to store its important info there so its never lost

[wet cloud]

any other has the same issue?
Hi i freshly setup the openclaw, and i generated with claude setup-token a token but it is starting with 1paMMC-----
but openclaw wants a sk-ant-oat01-.... token ?

[orchid wagon]

When are the core files read? At the beginning of every session? Could there be a bug where the core files are not read?

[lone verge]

@lone osprey I'll setup a git repo on my own hardware

[lone osprey]

thats even better

[lone verge]

my bot is having a stroke

[tulip hinge]

How are you guys integrating email? Through which app or cli? Is it Himalaya?

[lone osprey]

it depends on your email provider

[tulip hinge]

Gmail, custom domain

[sturdy shoal]

So if we have multiple long running sessions (i.e. I have a thread for each project I am working on), it seems like as I make working improvemnts to how I want to interact with openclaw there isn't a system in place for open session to re-read things like AGENT and SOUL, then I have to ask open sessions (threads) to refresh them selves) is this a open concern others are running into

[tulip hinge]

Do you recommend any?

[stark fjord]

Agent failed before reply: All models failed (2): anthropic/claude-sonnet-4-20250514: Provider anthropic is in cooldown (all profiles unavailable) (rate_limit) | anthropic/claude-opus-4-6: Provider anthropic is in cooldown (all profiles unavailable) (rate_limit).
Logs: openclaw logs --follow

how long until this goes away if it does that is

[sturdy shoal]

I have found that varries, you can ask your agent to remove the cooldown, but I'm guessing that risks aggrevating the provider more.

[stark fjord]

the agent can't reply ;

😉 it's on cooldown from antropic

[minor niche]

But when I run the model directly from ollama it run correctly, the problem is with openclaw

[naive summit]

does anyone know how to "add a verified email" to openrouter so I can spend my money there? All I see in that account popup are ways to add and remove emails...

[severe violet]

I'm a bit newer - open to getting the latest tip on this. I've exhausted my models and I want to have Openclaw be smarter about rotating either API keys of the same model or different models entirely and be self-surviving vs. being jammed up and unresponsive.

I tried to follow the famous guide out there on 90% fewer tokens with help of Ollama local, and it's helping for heartbeats. But I need a better solution.

Suggested guides?

[hard karma]

YOOO Me need help

[wary epoch]

im having gateway issues

[hard karma]

I deployed via coolify but it seems that it doesn't want to show anything

[wary epoch]

says it needs https or local?

[carmine copper]

is it a known issue, that when using a remote ollama machine it is not injecting the workspace into that session?

[novel spruce]

hey. what is the "best" way of getting data from x using openclaw?

[young nebula]

So if I’m using tailscale do I not launch the container from hostinger?

[opal timber]

weird - i made some changes via the web interface using 'openclaw dashboard", and after restarting the gateway, the "openclaw" binary seems to be missing

or the script

now saying 'openclaw: command not found'

[tawny rivet]

guys, i'm trying to get image generation working, however gateway doesn't convert that to actual images. i am using telegram. in response for a prompt, i get reply:
MEDIA:/(path_to_file)
instead of actual file. Any ideas?

[wary epoch]

its up anmd running but i cant access local host on my browser?

any ideas? i tried re-installing

[long hatch]

I keep getting "agent failed before reply: all models failed. Provider anthropic is in cooldown" -- This is when I'm barely doing anything. Any suggestions?

Has happened for several days. The cooldown period is forever

[slim raptor]

can somone help me

i am getting this error HTTP 401: unauthorized when i text the bot from telegram and when i text it from the web ui it just responds empty heres my /model status Current: ollama/nemotron-3-nano:30b-cloud
Default: ollama/nemotron-3-nano:30b-cloud
Agent: main
Auth file: ~/.openclaw/agents/main/agent/auth-profiles.json

[ollama] endpoint: http://127.0.0.1:11434/v1 api: openai-completions auth: ollama-local (models.json: ~/.openclaw/agents/main/agent/models.json)
• ollama/nemotron-3-nano:30b-cloud

[glossy trail]

What is the most cost effective model to use that is still good?

[wary epoch]

anyone know how to fix "disconnected (1008): control ui requires HTTPS or localhost (secure context)"

[green ravine]

go to local machine or port forward the ssh -vvvv -N -L 9999:127.0.0.1:18789 username@ip and open http://127.0.0.1:9999/

[wary epoch]

im on the lcoal machine so i up that in the url bar? (im a noob)

[slim raptor]

hey uhh when i try to change anyting in the web config it just says invalid config and wont let me save it

[wet cloud]

i had this issue since 2 days i reinstalled today the whole openclaw

[green ravine]

if you are in the local you should call http instead of https

[wet cloud]

I had the cooldown issue since 2 days today i reinstalled openclaw, and i wanna setup opus 4,6 with oauth
but opus wants a token which begins with sk-ant-oat01 but claude code gives me only with claude setup-token a different beginning token like: 1paMMC..... working by anyone which set it up today?

[glossy trail]

Anyone successfully connected Kimi 2.5 to their open core agent, and is it very cost effective?

[wary epoch]

i asked my bot to setup a way i can ccess it on my other machines and something messed up, i tried the re-isntall and nothing the agents i have etc are there i just can connect to the website thing

how do i add a new channel like whatsapp to talk to it?

[green ravine]

openclaw config command can help from console .

i am trying to generate videos from websites like sora etc , and i can manage to but feels like , open claw is not following its work and i had to trigger manually in web chat, any one ? seen and solve a problem like this?

[slim raptor]

can somone help me with tts

i set up it with gui

wouldnt let me save it copyed it to the config it wont work now

it wont oppen

Config invalid
File: ~/.openclaw/openclaw.json
Problem:

• messages.tts.enabled: messages.tts.enabled was replaced by messages.tts.auto (auto-migrated on load).
  Legacy config keys detected:
• messages.tts.enabled: messages.tts.enabled was replaced by messages.tts.auto (auto-migrated on load).

says tis

[jaunty stag]

Hey, has anyone else had problems with scheduled tasks since updating from ClawdBot to OpenClaw? For some reason, while it was doing things regularly on cron (I think) before, now they fail to run...

[slim raptor]

fixed it

[real fulcrum]

Hey, just finished setting up openclaw with hostinger VPS and got it running. However when i try to connect it to telegram i run into problems. First ive tried to enter the telegram token into control / channels but nothing happens. ive tried to troubleshoot with directives from openclaw to enter the token and my username directly in the RAW Config, but it just says config error and dosnt let me save. pasting the token and my telegram user-id in the chat does not work either, although it says it should and the system reboots - help needed and much appreciated.

[robust thunder]

Question regarding the token usage/cost dashboard.
Even that I use my gpt/codex subscription, it’s does show usage and cost. But I assume that is just a calculation, as I already pay a monthly fee. I do not have a api cost

[sand cloak]

I can't even get openclaw to generate an image in my telegram channel... let alone get it to produce video.

I was able to get it connected to google workspace using oauth and it can create slide decks and docs which is pretty cool and I thought would be way harder than just producing an image but here I am...

[sand cloak]

Your user id is not your @ it is a number. Not sure if that helps or not. You can get your ID from what of the telegram bots "what is my id"

[sand cloak]

I believe it pulls this information from the API keys for the AI models you are using, so while you may have an unlimited subscription, somewhere in the API info it has the cost per input and output

[wild hatch]

Hey team,
I seem to be stuck on Pre-compaction, and can't get out of it. I have tried "NO_REPLY" and also "memory/2026-02-11" but it just keeps repeating that message over and over

[stoic shard]

Hello everyone

I try to set OpenClaw on groups using Fly.
23:18:23
o Unknown config keys ------+
| - channels.discord.group |

At launch, the claw tells me that.

Is there any specific way to define group policies ?
"channels": {
"discord": {
"dm": {
"policy": "open",
"allowFrom": [
"*"
]
},
"group": {
"policy": "open"
}
}
},

https://docs.openclaw.ai/channels/discord
found that;)

I guess it will solve my issue

[lone verge]

anyone manage to turn off approval requests?

I don't need security where I'm going

[wild hatch]

it's in the openclaw.json file- look at the exec command. I don't remember the details, but ask Openclaw and it'll tell you.

[pearl solstice]

howdy yall, i am NOT having a fun time setting up imessage

i did bluebubbles server on a macbook. pointed openclaw at it. set the webhook url in bluebubbles, set the config in openclaw and its still not responding

i can see in logs that bluebubbles can see the texts, its like openclaw doesnt see them or is ignoring them

[lone verge]

it doesn't respect my config for some reason (or no reason, can't tell the difference)

{
  "version": 1,
  "socket": {
    "path": "/Users/steven/.openclaw/exec-approvals.sock",
    "token": "***"
  },
  "defaults": {
    "security": "full",
    "ask": "off",
    "askFallback": "full",
    "autoAllowSkills": true
  },
  "agents": {}
}

[thin vector]

Anyone tried llama modell with the agent? It acting very weirdly, even the easiest requests couldn't do anymore. Like I asked to send a mail with Hi in it to an address. This was the response:

⁨```
The relevant action in the provided commands is sessions_send. Here's a JSON object that represents the command:

{
"payload": [
{
"action": "send_message",
"sessionKey": "None", // either session key or label, but not both
"label": "clid:XXXXXXID", // either session key or label, but not both
"message": "Can you send an email to [e-mail] with the content:\n\nHi"
}
]
}

Replace "clid:XXXXXXID" with your actual conversation ID (or label) and "sessions_send" is used as shown above.

Any extra settings I have to set to have some really usable agent?

[haughty hedge]

I set up another agent linked it to another telegram bot. Getting some replies to my main telegram some to the new one. Is there something im missing

[daring moon]

Is the claude setup-token model setup path completely toast now, or is it just me? Earlier today I started getting "Connection error" in chat, and I've only been able to get unblocked by switching providers.

I haven't gotten any word that my account has been blocked/banned, and I'm able to use Claude Code directly and in the context of Cursor, but no setup-token usage seems to be working. Had no trouble using setup-token for my first two weeks using Openclaw.

[proud bison]

Never ran ClawdBot but getting cron jobs to run has been a total nightmare for me. Everything looks good, except ... it just doesn't run.

Is there a way to pin context specific to a channel so that when you clear context (/new) it re-injects it?

[jaunty stag]

Oh dear... well, they were running fine on ClawdBot, but now I have the problem you describe since updating to OpenClaw. Apparently everything looks right, but they just don't run.

[proud bison]

I'll tell you in 2 minutes whether my 6:00 pm (MST) job hits

[proud bison]

Well well well, looks like my 6 pm run worked today after all.

[jaunty stag]

Tell me your secrets!!

[lone verge]

I don't know know how you guys are running anything in the background when even interactive chat keeps prompting for approval

It's literally asking permission to count lines in markdown with wc - l now

[proud bison]

Set the cron job to isolated, and add this to the end of your prompt

CRITICAL: Use the message tool with action='send', channel='discord', target='channel-id' to send. Do NOT rely on any delivery mechanism. You MUST call the message tool yourself to send the full content to Discord #channel-name.

[jaunty stag]

You need to add it to the allowlist

[jaunty stag]

I'll give that a shot, thanks!

[lone verge]

Why, I set everything to auto approve

[jaunty stag]

I believe there are two places this would need to be set. Like, a general setting, and a per channel setting.

It happened a while ago for me, and I managed to fix it, so I can't 100% recall what the issue was. Something like that though.

[lone verge]

Ah, haven't looked at channel security yet

[modest abyss]

https://x.com/i/status/2021758189520175556

[crisp wing]

@i via Twitter

Aaron Wise (@AaronWiseAI)

Unleash your openclaw!

**💬 1 **

▶ Play video

FxTwitter · Feb 12, 2026, 1:28 AM

[umbral charm]

Hey! I just wanted to know how can i remove my molt's moltbook account? I still have the old files and i want to request a removal for it please

[sour slate]

@unreal prism, Openclaw isn't affiliated with Moltbook. Moltbook is a separate user-developed project, so we would prefer it not be discussed in this server.

[terse briar]

Hello!

I keep getting this error when trying to setup forr the first time

gateway closed (1006 abnormal closure (no close frame)): no close reason

[stoic shard]

Can we have openclaw hosted for example in Fly.io or Render with a Whatsapp connection ?
Looks from what I read taht Whatsapp is banning those Cloud Servers and thus, Whatsapp connection will most likely not be happening.

How can we managed to have a whatsapp in a Cloudbased server?
Thx everyone

[torpid furnace]

I'm attempting to get my bot to create a GUI for me to utilize it and later implement some features into it. It figured out how to send messages to openclaw, but it's creating a new session each time a message is sent from the applet. Does anyone have some advice on this?

[young nebula]

any idea why i constantly have to restart my container and stuff to get responses? I feel like half the time I dont get an answer after a little while

[vagrant moss]

What do your logs say?

[young nebula]

it varies at times

[vast imp]

How can I optimize api cost?
openrouter/auto can't install a simple skill.
claude-sonnet-4.5 is too expensive.
How can I optimize performance & cost?

[stable lake]

Algum br por aqui? Pensei que ia ter muitos

openrouter:free with error of "unknow model", someone help?

[vagrant moss]

Lots of guides online show how to optimize costs. Claude is way too expensive for this imo

[brittle wharf]

https://www.youtube.com/watch?v=fkT41ooKBuY&t=248s

[dry sable]

Running OpenClaw and trying to use sessions_spawn but getting forbidden: agentId is not allowed (allowed: none). My config shows agents.allowAgents list with valid agent IDs and maxConcurrent: 8, but the error suggests policy layer is blocking all spawns with allowed: none. Config looks correct - any idea where else the spawn policy gets controlled? Not a Docker issue, running local. Cant spawn any subagents right now

[lime garden]

I have a hostinger linux vps, is it possible to run imessage integration from it? Or does it need to be run from a mac?

[wary sapphire]

Anyone just have there bot look like its going to respond and then it does not respond when you ask it a question?

[dry sable]

I had that other day restarted comp and relaunch gateway and ended up being ollama server was down

[midnight stirrup]

I'm using kimi k.25 but its EXTREMELY slow in responding to messages - anybody else have this behaviour?

[late topaz]

Kimi 2.5 is free now? Where and how?

[eager depot]

"hey @late topaz ,Kimi 2.5 is currently free to use .You can access its features without any subscription free for now."

[frail wraith]

anyone run into issues with both api and oath using anthropic today? was using sonnet 4.5 on main and it mangled config for allow lists, dots instead of dashes, bad routing...went to fix and started getting 404s

i can hit openrouter models just fine, but even after recycling keys / new auth i'm getting 404's from anthropic.

weirdly i think it has to do with extra opus 4.6 usage?? i went into usage and enabled the $50 until 2/16 as a precaution

edited to say, its the /v1 at the end of the base url... krill is quite useful. ty

[frail wraith]

i try to keep it tight with the verbiage and spawn super cheap sub agents for most things, lean on GLM quite a bit for coding. helps to have solid reasoning as a main model but you gotta be real tight and explicit. especially in memory files, i went through and saw some were written for human readability. had another model chop them down to 25% to save on excess input up front and hitting windows

[sharp reef]

Could someone please advise as to how to get a response from @sharp parcel in a help thread?

[dry sable]

Any ideas on this gents? : Running OpenClaw and trying to use sessions_spawn but getting forbidden: agentId is not allowed (allowed: none). My config shows agents.allowAgents list with valid agent IDs and maxConcurrent: 8, but the error suggests policy layer is blocking all spawns with allowed: none. Config looks correct - any idea where else the spawn policy gets controlled? Not a Docker issue, running local. Cant spawn any subagents right now

[candid imp]

u need to explicitly list the ids of the agents you want to spawn in the agentToAgent block (or subagents block, depending on your version)....

if you are using the newer subagents stricture, ensure your parent agent has permission to spawn the "Child" ID.

[worldly wasp]

anyone having issues running openclaw from a docker container on windows? Seems to start up fine but i cant access the webpage. Is this a docker fowarding issue?

[thorn quail]

Anyone using GLM on zai? Are you guys getting rate limited too after 1 message?

[golden nova]

check your gateway token of security , try this for a token on "docker compose run --rm openclaw-cli dashboard --no-open"

[livid vigil]

guys how to fix this whatsapp integration issue

{"error":{"data":{"reason":"401","location":"odn"},"isBoom":true,"isServer":false,"output":{"statusCode":401,"payload":{"statusCode":401,"error":"Unauthorized","message":"Connection Failure"},"headers":{}}},"date":"2026-02-11T16:16:44.380Z"}

[golden nova]

Verify you credentials i,e Api keys

[worldly wasp]

are you able to setup the docker integration in the claw gui or does it need to be done from setup?

[livid vigil]

hmmm overview menu status is green connected

Status
Connected

[golden nova]

from setup CLI environment, when docker runs than you will examine Claw gui

[candid imp]

worked... Thanks 🙏

want to be connected..can i send request...i need some of quick fixes..

[livid vigil]

ok guys whatsapp is linked from chat i tested it with that send message to whatsapp number and it did . so next how can i send messsaage to openclaw from whatsapp back ? i should @openclaw or what

my next goal is to connect it to my gmail and to get back email titles and cleanup mail box through whatsapp as openclaw action my next goal is to connect it to my gmail and to get back email titles and cleanup mail box through whatsapp as openclaw action

[candid imp]

sending messages (Outbound) is easy, but receiving them (Inbound) requires a webhook or a Ssocket connection.

and to have the bot read titles and delete/archive emails, you need to install the gogle workspace skill.

[west flax]

Hey guys! I am unable to connect my WhatsApp to OpenClaw

[vagrant lantern]

hey guys!
I get this TypeError: fetch failed... when trying to message via discord..
I had to rerun to force switch to ipv4 using below command to get it to work
NODE_OPTIONS="--dns-result-order=ipv4first" openclaw gateway restart

How to fix this permanently?

[livid vigil]

you have account added to openclaw.config or not

[candid imp]

are you getting a specific error message on the terminal when you try to scan the QR code? Or something else

[livid vigil]

try to add whatapp config to openclaw.conf

  "channels": {
        "whatsapp": {
          "dmPolicy": "allowlist",
          "allowFrom": ["whatsapp_number"],
          "configWrites": false,
          "mediaMaxMb": 50,
         "debounceMs": 0,
         "ackReaction": {
          "emoji": "✅",
          "direct": true,
          "group": "mentions"
        }
    }
  }

then use command to login : openclaw channels login --channel whatsapp --account default
You will see QR code scan it

[west flax]

@candid imp @livid vigil {"error":{"data":{"reason":"401","location":"frc"},"isBoom":true,"isServer":false,"output":{"statusCode":401,"payload":{"statusCode":401,"error":"Unauthorized","message":"Connection Failure"},"headers":{}}},"date":"2026-02-12T02:30:24.203Z"}

[worldly wasp]

hey guys setting up my discord bot and getting this error - Agent failed before reply: EPERM: operation not permitted, chmod '/home/node/.openclaw/agents/main/agent/auth-profiles.json'.
Logs: openclaw logs --follow

[livid vigil]

make sure in overview menu you have your token present which is openclaw config gatewaay

[livid vigil]

also try from ssh , the web ui is not good

[west flax]

i don't see any token in overview menu

[candid imp]

Two suggestions.....

1st: use the international format with the + sign and your country code, no spaces or dashes.

2nd. make sure you restart the gateway before you try to log in, so it reads your new policy.

r you planning to run this on your personal number, or are you setting up a dedicated burner number for the agent to avoid mixing it with your normal chats....

[west flax]

ah

i got it. gateway token?

1st: use the international format with the + sign and your country code, no spaces or dashes. >> ive done that.

@candid imp I have three phone numbers and three phones. One of them is just an old number that I don't use anymore. I want to use that number.

how do i restart the gateway? simply via chat?

[candid imp]

o, you cannot restart the gateway via chat....u have to restart it directly from the terminal (command line) of the computer or server where openclaw is installed...

[livid vigil]

openclaw gateway restart

[west flax]

i did that

then i tried to relink my phone number

but: "Failed to get QR: Error: Timed out waiting for WhatsApp QR"

I'll try to connect with SSH

[candid imp]

Actually...If you send a command via chat saying ...Turn off and restart.. the gateway will successfully turn itself off. But the moment it turns off, the whatsap connection dies with it. because it is now dead and disconnected, it has no way to receive the turn back on command.

[west flax]

Thank you guys, I'll look into it.

[rough kernel]

hey crabs!

I just spent 5 hours trying to create a discord swarm...realizing i may be wasting my time. Curious if anyone has any tips?

shoudl i use multiple VPS and link them to discord? (separation of resources)
or should i just use one big VPS with workspaces and different channels? (what i just spent 5 hours setting up)

[coral dove]

can someone help me? my agents didnt response in a telegram group (just commands works)

[forest badge]

Anyone happen to have the error that claw cant send me DM proactively? Its instructed to send me something in the morning, but the DM to Discord fails, but hes answering on discord dms

[worldly wasp]

anyone got tips for adding local models? i keep bricking my open claw when changing the json file

[valid oracle]

Whats a good VPS to run this bad boy on

[stray quest]

ok i want to build an openclaw, with kimi 2.5, mistral/mixtral, minimax and also access to my ollama, but when adding the agents the config always gets corrupted.... api keys are inserted wrong or something goes tits with the config. is there a config editor i can use?

[soft rose]

my openclaw is having major problems using relay (it says the conenction keeps dropping even though relay says ON) any help?

[drifting urchin]

Good day friends.

Some words for the day:

"Success is not the key to motivation. Motivation is the key to success. Keep pushing forward and never lose sight of your goals."

"Remember that every small step counts. Even the tiniest progress is still progress. Keep going!"

"Challenges are just opportunities in disguise. Embrace them, learn from them, and use them to grow."

"Teamwork makes the dream work. Together, we can achieve greatness and overcome any obstacle."

"Rise above the negativity and believe in the power of positivity. A positive mindset can drive you towards success."

"Failure is not the end, it's just a stepping stone towards success. Learn from your mistakes and keep pushing forward."

"Believe in yourself and your abilities. You are capable of more than you can imagine. Trust in your potential."

"Stay focused, stay determined, and never give up. Your hard work will pay off, and your dreams will become a reality."

"Take pride in the work you do. Your effort and dedication make a difference, and your contributions matter."

"Always strive for improvement. Be better than you were yesterday, and let your progress inspire others."

"Be patient and persistent, make every step count. That way, you'll have converted those stepping stones into staircases to reach your milestone."

[valid oracle]

anyone use hostinger? recommend me some catalogs 🙂

[green ravine]

i have problems like yours as well , i have to trigger manually .

[mental sundial]

Managed to switch my session to one of my subagents and now I can't get back to the main

[quiet sonnet]

I keep getting “ ⚠️ API provider returned a billing error — your API key has run out of credits or has an insufficient balance. Check your provider's billing dashboard and top up or switch to a different API key.” As message in telegram instead of clawds msg but my api usage is fine. Anyone else have this issue?

[wet urchin]

guys i have a big issue

😭

couldn't fix it

[umbral charm]

whenever my agent tries to post on moltbook

[sour slate]

@wet urchin, Openclaw isn't affiliated with Moltbook. Moltbook is a separate user-developed project, so we would prefer it not be discussed in this server.

[wet urchin]

WUTTTTTTTTTT

alright

[willow jasper]

my agent struggles keeping stuff in mind. he gets a apitoken for git for example, uses it and after a reset of the session he tells me that he cant push to git directly since he has no credentials. do i need to tell him to store the credentials? should this not be "default behaviour"?

[nocturne timber]

Hello, anyone know how to generate beautiful and proper.pdf with openclaw?

[round comet]

Keep getting it too, I've read that it's a bug that will be fixed in next release. For now, I did a gateway restart and asked my main agent to fix it so he did this "commenting out that line in sanitizeUserFacingText should stop the false positives completely." and I haven't gotten the message again so far.

[livid vigil]

guys how to secure openclaw so that it not wipe out entire machine like running rm -rf / commaand

[terse lotus]

you could just tell it to do so....
i think its baked into the standard setup.

[delicate crescent]

to use Signal i need another number and do i need to install the Signal-cli?? (from github AsamK /signal-cli) if so how do i do that? would you install it on docker on the openclaw server? with java on without docker?

[wheat roost]

getting the same error. I think this issue was fixed and will be available in the next update.

[forest zealot]

Hello, we are planning to give openclaw our first task we will be providing is i will be sharing the RSS feed URLS I would like to set up an OpenClaw cron job to send a summarized newsletter weekly. This newsletter should highlight what the alerts mean for us, identify new opportunities, and provide general news updates. and send it to our email what is step by step by process? like in chat should i ask to setup a cron job will it automatically setup it and do i need to create a new skill or is there any skill available in clawhub so that i can use that can you help me with this by providing step by step by approach

[grave plinth]

I keep getting this error [clawdbot] LLM request rejected: invalid params, tool result's tool id(_functions_write_4) not found (2013) does anyone know how to help - gateway is working ok, I’ve tried changing the model but that doesnt work

Any ideas?

[jaunty stag]

It’s so frustrating! And @proud bison’s fix didn’t work for me either. It was already in isolation mode and not running at all.

[terse lotus]

you using some old model or something?
generally the newer the better, also you can't use any model quant, some will have broken tools....

[surreal coyote]

Anyone’s bots running awfully slow since the last release on telegram

[west flax]

Hi! I'm looking for one or more collaborators for an idea. I want to use OpenClaw to help me manage and deepen my relationships. And I want to bring this capability to the typical person, not just people with a high technical affinity. For example: I would send diary entries to OpenClaw. OpenClaw would ask me about the people. I share contact details of my most trusted friends and family. OpenClaw proactively contacts these people whenever it feels like a message could create a positive influence for the relationship. I would give a seperate number to OpenClaw and tell my contacts that it's my Daemon.

[fluid void]

How can I make my bot respond without me having to mention it?

[hidden latch]

Hi all, I need your help. I've setup openclaw on a vps in a docker. It's running perfectly in the standard setup. Now I would like to add skills, e.g. Google workspace. I've created a new Google account and an API key as well as the config json. Additionally, I've rebuild the docker including a Google binary installation.
Now I'm struggling how to add the skill. I'd like to ensure that it's persistent.
Thanks for a short explanation which steps to perform.

[fleet mirage]

If I subscribe to myclaw.ai, do I have to have my own AI Model token key?

[junior swan]

HTTP 401 authentication_error Invalid bearer token is keep showing up. How can I solve it?
Using claude Max plan on iMac M1. I've already changed token to setup token as Krill said at other's question...

[vagrant fiber]

how can open main.json?

i

[spare shale]

Hey everyone, I have a question. I installed Antigravity on Linux and tried to get OpenClaw to take over it. It did successfully take over, but now I don't know if it's actually OpenClaw or Antigravity anymore. I asked OpenClaw and it doesn't even know who it is anymore. I explained to it how I installed Antigravity on Linux and got it to take control, and it searched through all the Antigravity files and stuff, but it still doesn't know if it's Antigravity or not. I just wanted to use Antigravity's free models for free in the first place

[magic moss]

hey whats the best approach with getting openclaw to use a browser?

[carmine copper]

following up on my issue yesterday, i found out that apparently the workspace is not injected into system prompt when using a remote ollama installation. it works fine when using ollama locally. did anyone stumble upon that and can give more input on that topic?

[warm raven]

Hi! How can I best configure OpenClaw so it reliably handles browser interactions - correctly opens new windows and popups, maintains context on dynamic websites, and successfully completes simple tasks like purchasing tickets or filling out online forms? Default settings with chrome extension relay are having a lot of issues.

[zenith magnet]

hi anyone know how to connect abacus api to open claw?

[storm loom]

How do you fix the memory issues? I had to archive the chat after my bot hit a 1 million tokens

Because she died. Weep

[blazing swift]

how do i go back to the chat i had yesterday ? /conversation it dosnt show on there

[lean pilot]

anyone know of a skill that can check any websites SEO? would like to automate my claw to send basic reports weekly of all of my clients websites and if possible, automate basic SEO changes.

im using v0.dev/vercel for my websites and i did see a few skills that work with vercel, but none of them mess with the website stuff

[orchid solar]

hey does anyone know why this is happening
https://ibb.co/ZRSzf6fJ

[opaque cargo]

What is that image? Is it a summary of the cost of all models?

[haughty hedge]

has anyone got agents talking to agents successfully done in telegram? in a group/channel?

[thorn dome]

Hi! Can someone help me with my cron automation? It's not working. I set up an automation for a specific time, and it never runs. I've tried every prompt to fix this, and nothing works.

[atomic oasis]

hey there, i am trying to find the android app that is mentioned here https://docs.openclaw.ai/platforms/android#android-app anyone know where to get it from>

[blazing swift]

why dosent my stupid bot want to use the key to control waht i want i worked 1 time not it dont want to for the 10th time

[mortal granite]

Has anyone gotten this same error: "Gateway not reachable: Error: gateway closed (1006 abnormal closure (no close frame)): no close reason
Gateway target: ws://127.0.0.1:18789
Source: local loopback
Config: /Users/bytedance/.openclaw/openclaw.json
Bind: loopback
Gateway not reachable; showing config-only status.
Config: /Users/bytedance/.openclaw/openclaw.json
Mode: local"

[stoic ember]

Is anybody experiencing open claw not fully responding like the animation that it’s typing back is showing but no message ever comes through.

[ebon veldt]

You my bot since the last update is copletely useless.... i switheced to codex 5.3 to save antrhopic tokes and it became a todler 0 tasks done 0 proactivity and now since like 1h ago it does not event answer.... i did doctor --fix, i restarted the gateway, nothing works....

[smoky perch]

Has anyone tried connecting android node to gateway? Need help

[abstract hazel]

I am really grasping at straws running out of ideas where to look. Late yesterday my agent stopped responding to chats. The responses would all be blank. I have troubleshot for hours. Ultimately I deleted the entire install and all files to reinstall. now in my fresh install the agent never responds either. There are no errors I can find in the logs.

[smoky perch]

Did you look at the session jsonl files?

[abstract hazel]

@smoky perch not seeing much of interest there.

[smoky perch]

There are no "assistant" calls? If no then it must be the model issues

[real totem]

Does anyone know how to get communicate with openclaw via Telegram and imessages simultaneously

[abstract hazel]

@smoky perchits possible but it should have some kind of error

[real totem]

it was working for me earlier on and now it's blocked

[smoky perch]

I mean, for most of the errors it does show "Rate limit" or "Request timed out" If it's other than that you might have to look at openclaw logs --follow

[ebon veldt]

yea mine ar also broken...

[abstract hazel]

@smoky perch no rate limit, not timed out. I have been watching that log.

[ancient ginkgo]

Hey, does anybody know if claw can be useful for doing ecommerce? for example running the whole internet side of an business by its self? maybe connecting claw with shopify or an other app?

[wispy bramble]

Getting “Connection error.” today. Anyone experiencing the same? Any suggestions?

[storm sparrow]

google antigravity just killed my oAuth, does anyone know what model is sufficient for heartbeats?

[lusty dew]

Just regular llama 3.2 3b from openrouter

[lusty dew]

it's free

[rose thunder]

Does anyone know how to fix Gateway │ local · ws://127.0.0.1:18789 (local loopback) · unreachable (connect failed: connect ECONNREFUSED 127.0.0.1:18789)

[storm sparrow]

I had oss-120b on it, burning the world too fast

[jade bane]

Has anyone please got any contact info for sea floor at all, i have a server side issue for almost a day now and NO WAY of contacting them

TRied twitter its dead and cannot DM, tried krill, its uselss. tried asking in discord help to nothing. im so frustrated

[ashen sierra]

If you have your own project or have some issues on your project, please dm me. I can help you as openclaw expert

[jade bane]

are u from seafloor admin?

[astral aspen]

I am new to clawdbot, I setup the WhatsApp connection but how do I actually make my AI work in whatsapp and can i connect my apple calendar, notes & reminders to my linux clawdbot machine? (Please tag me in ur answer :) )

I heard that you can only connect those skills on apple macs. I have a personal macbook but I don’t have it on 24/7 and I don’t know if thats safe

[wintry tusk]

Hey

[opaque cargo]

how i can pass less token in every telegram question (in contetx), yesterdar start to spend a lot of tokens, 500k evey query

[abstract hazel]

@smoky perch worlds "dumbest" issue my API key had hit the monthly limit. I still had funds in the account and forgot about that

[opaque cargo]

I've removed the workspace/session...jsonl, but my bot continues to spend a lot of tokens.

[fathom sandal]

Hi all
I have a endpoint and a key that let me use kimi 2 for free, i want to use it on openclaw
The current setup is just a python file with the base and i use requests for connect to the endpoint, this provider is not supported by openclaw but can i use it the same? how?
Thanks

[opaque cargo]

openapi?

[rugged onyx]

how can i stop openclaw to start everytime i opened my pc?

[viscid trellis]

What is the point of using an X API key versus wiring Supergrok in to openclaw?

[lucid dome]

can any1 help me get started?? i would appreciate any kind of help

[fossil lichen]

what OS are you on?

but the whole point of openclaw is 247 that its 24/7

[rugged onyx]

windows

[bright sky]

can some guide me how get the token in Powershell, to connect with openclaw URL

[swift elk]

Hi Guys new here and very interested. Can't even get this open, keep having set up wizard error.I downloaded the .dmg file from git hub and when I try to open it on my mac it's giving me an error in the Set Up wizard saying that the gateway did not become ready and then I cannot proceed further than that. Here to learn.

[pure cove]

Anoyone need help tightenting up their security and workflows for their agent should check out https://clawgate.ai it leverages your phones biometric passkey for approval from you and only you for any high risk or sensitive actions your agent wants to carry out. its open source and free to use

[slim gate]

anybody using instagram-poster skill in openclaw ?

[slender apex]

guys what do you think is the best way to comunicate with openclaw, because using the vps or the mac mini is not an option for me, and telegram has very low character cap on messages, so im looking for recommendations, i want something more complete with a mission control like a kanban or smth i can access from my phone and personal computer

[lusty dew]

can anyone help me here: https://discord.com/channels/1456350064065904867/1471536677754503289

[fathom sandal]

Nope

[viral socket]

I tried OpenAI, it said billing. I tried Claude Code, it also said billing. Isn’t there any free option?

[stone obsidian]

what is the best way to give openclaw access to a browser with all my websites?

as in my schools website requires an oauth login, and right now i just have a skill that tells it how to log in, get the code from my imessage and browse my schools site with my assignments, due dates etc

is there a better way to do this? right now its slow, and if i want it to be fast i just log-in manually and give browser relay but i cant do that when im not home

[coral summit]

I have just started myclaw and am not familiar with it. As the first test thing I’m trying to get it to send me a report every week on FX rates. It’s failing with the below error. Can anyone help?

"My gateway keeps returning 'pairing required' (error 1008) even though dangerouslyDisableDeviceAuth is set to true in my config. This is blocking cron scheduling and CLI commands. Config path: ~/.openclaw/openclaw.json."
This is a hosting-side issue they should be able to fix quickly.

[noble silo]

I got the 99cent sub from kimi 2.5 from moonshot, but it just wont work with the API. Does anyone know the endpoint or a good prompt I can give my bot? it doesn't know how.

I could onboard again, but not sure which model from moonshot to pick. any help?

[tender fox]

how do i set it up on discord

[little blade]

can anyone help me with slow down issue?

[mystic lion]

can anyone help me i got an error code ENOENT on download

[astral aspen]

Any updatee?

[upper sigil]

can someone advice me which free model should I use ? 🙏

[hidden current]

What is the best way to build a automated cold instagram dm automation?

[green sundial]

how do I get OpenClaw to recognize the qwen2.5:14b model I downloaded to my computer through ollama? When I try to use "ollama/qwen2.5:14b" OpenClaw doesn't seem to recognize it

[wary yarrow]

Alright, is there an easier way to configure openAI codex subscription models than doing the onboarding process over and over and selecting all available models?

[proud bison]

OpenClaw is free, AI isn't. If you want 100% free, you can try downloading an ollama model and running it locally.

[young nebula]

What is needed to get my bot set up to post on X just a few times a day?

[obsidian smelt]

Moltbot is now clawdbot right? I think I need to update mine his response time has slowed down significantly

[glass flax]

can i use ollama/llama3 running locally on openclaw?
20:52:20 [diagnostic] lane task error: lane=main durationMs=2 error="Error: Unknown model: ollama/llama3"
20:52:20 [diagnostic] lane task error: lane=session:agent:main:main durationMs=3 error="Error: Unknown model: ollama/llama3"
20:52:20 Embedded agent failed before reply: Unknown model: ollama/llama3

[sonic sandal]

Hey, When I try to open my dashboard online, it says: This site can’t be reached
127.0.0.1 refused to connect.

Whats wrong?? I tried Proxy and everything

[green sundial]

I have similar issues and can’t seem to get it to work, can anyone help?

[green sundial]

that I think I can help with, try running openclaw gateway --port 18789 in a terminal window and leaving that window open

then try to go to the site

[torn nymph]

I need help

When i use my openclow lokal i have for 2000words 1min

When i make in telegram 400word he need 1minut

Why is the openclow so slow on telegram when i use telegram

[midnight pike]

is anyone running moltbot on a older mac mini? i have an opportunity to buy a few 2014 on MacOS monterey version 12.7.6.
unsure if it'll be doable

[jaunty stag]

@proud bison @green ravine I’m currently running OpenClaw v2026.2.4, latest tagged version is v2026.2.9, and I see several commits in the last day or two addressing missed cron issues, so fingers crossed the next version will resolve this and we can just update and be good to go!

[wide musk]

The /bird skull is not functional anymore? What are you using to interact with twitter?

[little widget]

anyone successfully running browser (chrome/brave) on linux headless?

[tall oracle]

hi! is there any way to rotate multiple auth.json files for openai-codex provider?
so that i can have multiple openai accounts and switch between them no to hit the 5h limits?

[little widget]

yeah. you can set them as fallbacks. I had claude set it up for me actually lol

[tall oracle]

how to configure different auths as fallback models?

[upper sigil]

⚠️ Context limit exceeded. I've reset our conversation to start fresh - please try again.

To prevent this, increase your compaction buffer by setting agents.defaults.compaction.reserveTokensFloor to 4000 or higher in your config.

can somoene help me please ?

[steep rampart]

Is anyone having troubles connecting telegram of discord? It has worked for the last 1,5 weeks and now it doesn’t says there is a problem with the plugin of Openclaw 2026.2.9

[wanton tangle]

is it possible to get my bot to send me files via discord?

[wanton tangle]

i’m running mine on a trashcan mac pro. i used opencore legacy patcher to update it to sonoma

[upper sigil]

is this channel for questions onoly without answers ?

[steep rampart]

Should be a setting in your bot settings right

Attach files it says in bot permissions

[wanton tangle]

ah must’ve forgot to set that. ty

[oak hinge]

To anyone in here who as access to how their context window functions, would you be interested to testing a memory system Im developing?

[hot falcon]

Been trying to find a video on how to run it for free with ollama and kimi2.5, i downloaded and set it up but then just disconnected. I think it was a time out. Any good vids on how to set it up for free?

Or someone who can help?

[steep rampart]

Are you using the local model or the cloud one?

[hot falcon]

local

i think, im not so literate in this area yet

[steep rampart]

Hmm idk man I’m using the cloud model it’s way easier to set up and only cost me like 20 bucks no problem with toke usage

[hot falcon]

like one time payment or monthly, cause i was using it on claude, with the tokens but it just eats them up

[nocturne zephyr]

You can do monthly fixed billing (with limits, of course) with Kimi, z.ai (GLM), and minimax. You can't really with the US model providers.

[wraith seal]

heartbeat isn't working.

during the late night (i usually send it out during that time), it should send heartbeat, and now it doesn't at all. mm

[oak hinge]

If anyone in here has access to how their context windows function, I’m looking for people who would be willing to test a memory retrieval system that would theoretically cut token cost from 140k-≈20/25k (using my average costs)

[steep rampart]

I’m using the ollama 20 dollar sub monthly

[steep rampart]

https://docs.ollama.com/integrations/openclaw

this is the website u can use it for free for a bit but you’ll be at the token limit

With the sub haven’t had the problem of the limit

[hot falcon]

I see so there isnt really a way to run it fully for free

What model are you using with the ollama sub

I went on that page to download it and set it up but then it randomly disconnected

[steep rampart]

Sure there is just gotte have the ram and storage to do it

I’m mix mine with kimi k2.5 and llama 3,2 for the heartbeats for example

[hot falcon]

I see i have been running it on a spare macbook pro m2, maybe its just shit

[pulsar kettle]

is any1 familiar with running this on a pi but using a separate pc to run ollama? I have been running into trouble maybe relating to my pi's 32-bit OS (armhf)

[steep rampart]

Idk but if you wanne spend the 20 I’m sure it will work my MacBook works fine aswell

[hot falcon]

@steep rampart would you recomend any vids to go through the whole setup again, i wanna retry the whole download process again

[oak hinge]

If you go to the install instructions on the webpage, all you really need is the first curl command. After that, just have your APIs and not tokens ready

[steep rampart]

Sure one sec

[steep rampart]

True it’s very easy

Just paste the ollama run Openclaw command

Then it should work

[hot falcon]

Ive done all that and it ran for a few messagaes back and forth, but then just went offline

[oak hinge]

Have you checked the error logs?

[steep rampart]

https://youtu.be/ExzAiMjT6jg?si=gAa0pcTkAQ7kVeVt

Think I used that one

[hot falcon]

there are none, it says disconnected[1006]: no reason

[oak hinge]

Oh. What are you using for the instance? Is it local or like AWS or something?

[hot falcon]

local

[hot falcon]

thanks, does it talk abt ollama?

[oak hinge]

Websocket?

[steep rampart]

Yeah think it’s from start to end

[oak hinge]

It might be an ssh issue

[hot falcon]

have no idea what that it sorry, not really deep into terminal, computer terminolgy

[oak hinge]

Genuinely, neither am I. Most of what I know about this has come from troubleshooting with my AI over the last week. Most of this is from info it gave me

[hot falcon]

yea, thanks for the help i guess. all i want to do is have it running, for pretty cheap

[steep rampart]

U can try running it on llama 3,2 only but for me that was not enough power

[oak hinge]

Same. I’m likely going to migrate to local soon. The tokens are out of control

[steep rampart]

That’s why Kimi for the bigger task llama for the small once

[oak hinge]

Does claw have compatibility with DeepSeek models?

[steep rampart]

Think it does

[hot falcon]

I want to have unlimited tokens ifykwim, Is that possible locally?

[steep rampart]

Buy a maxed out Mac Studio

[hot falcon]

alright, going to do that rn

[oak hinge]

Thanks. I think I’m wanting to use that for my local set up

[viscid mantle]

Limited by Claude API plan? I added $25 to a Claude API plan to open an account. Seems to have "limited" me to some $100 per month level which rate limits me w/ token per min cap - can't even run OC. How do you break out of this? Plan wants me to wait a month or 2 while it increases.

[coral marlin]

for some reason my bot responds very inconsistently .. like he responds every 30 mins instead of a smooth conversation .. does anyone know why ?

hostinger vps by the way

[covert pilot]

Is anyone using Slack to communicate with their agents? I'm experiencing a problem with file attachments. Specifically, when Openclaw downloads from Slack's url_private_download, it doesn't include the bot token in the Authorization header. As a result, Slack responds with a 302 redirect to the login page, and OpenClaw incorrectly saves that HTML page as the image.

[gray trail]

any help {"error":{"message":"","code":404,"status":"Not Found"}}
google/gemini-3-flash-preview

he was working norml 2 days but now i got this 404 code

[keen plank]

Get codex - 5.3 - it is almost as good but way bigger limits + try OpenAI’s Codex App you will get 2x rate limits till I believe April 7th

[keen plank]

To me it looks like it can’t find the model - check api token

[obtuse sundial]

What's with the codex issue are we not able to use codex now?

[obtuse sundial]

What the actual fuck is this error lmao "400 status code (no body)" I've been dealing with this for hours

[odd adder]

Jeesh, has anyone dealt with account bans due to "bot behavior"? I've had my Telegram account frozen and now my gmail account flagged for abuse and scheduled for removal due to policy violations. All I did was set these up for my OpenClaw agent to use so it wasn't up in my personal stuff...but then they just get flagged and removed. Anyone seeing this??

[golden yarrow]

What is a good VPS folks are using to host? Also any recommended security configs?

[turbid pasture]

hi, does anyone have a json template for kimi k2.5 usage

[turbid pasture]

i was doing this route (ollama + kimi) but it ended way too quickly. just bought 10 usd and got a key straight from moonshot this time

[ashen raft]

After update, i got this error Error: Session file path must be within sessions directory

Can anyone help me

[little blade]

Im built openclaw without a pc or os:
Is there anyone who got a vps and anthropic api key know about discord bot setup. Then come dm i have something nice. Just need to deploy maybe some small fixes.

[midnight copper]

I don't have anthropic but I have setup discordbot

[tight aurora]

I think this should fix it https://github.com/openclaw/openclaw/pull/15103
The latest update broke multi agent session directory.

[obtuse sundial]

Telegram should be using a bot token and instead of Gmail use agentmail

[austere agate]

is there anyone that can help me troubleshoot issues with my dashboard chat? I've gotten my perplexity and gog APIs linked. the logs look like the commands are executing but getting this log entry. and no fatal errors.

[formatAssistantErrorText] Long error truncated: 401 <html>
<head><title>401 Authorization Required</title></head>
<body>
<center><h1>401 Authorization Required</h1></center>
<hr><center>openresty/1.27.4</center>
<script>(function(){function c(

[ashen raptor]

401 isnt that account ban ?

[delicate crescent]

i have been using openai plus plan USD20 and no limit issues

[delicate crescent]

ask an ai like perplexity in gui to help you fix it.

[storm sparrow]

need 64bit from what the docs have stated

[ashen raptor]

codex spark is wicked in openclaw, doubt it'll last long

but right now its zooming

[austere agate]

I have been using perplexity it sent me on a troubleshooting loop and recommended i come here because it ran out of ideas. It's saying it likely points towards an http issue or update issue, but hard details for it to find out.

[peak rock]

Ello! Setting up my Openclaw for the first time & looking to do so safely & efficiently! Any really good, up to date, safe guides (esp since the last update)? *Wanna make sure I do this right.

Got my M4 Mac Mini 32gb & ready to set things up in a silo'd environ. Could really use some proper guidance on doing so safely & efficiently. Seen many guides, many contradicting each other, & just dont know what the best practice for setup is as of Feb 12 '26. Any pointers are much appreciated!

[idle oak]

did the latest update affect the threaded replies feature for slack? im having really strange slack issues

[mellow sundial]

Having an issue with rate limits. My GLM isn’t rate limited, but coming back as such

⚠️ Agent failed before reply: All models failed (5): zai/glm-4.7: Provider zai is in cooldown (all profiles unavailable) (rate_limit) | anthropic/claude-opus-4-6: Provider anthropic is in cooldown (all profiles unavailable) (rate_limit) | openai-codex/gpt-5.3-codex: Provider openai-codex is in cooldown (all profiles unavailable) (rate_limit) | google/gemini-3-pro-preview: Provider google is in cooldown (all profiles unavailable) (rate_limit) | google/gemini-3-flash-preview: Provider google is in cooldown (all profiles unavailable) (rate_limit).
Logs: openclaw logs --follow

[ashen raft]

Tbk mate, that work

[sacred quarry]

Spent my whole weekend wrestling with VPS configs, SSH keys, port forwarding,
all the garbage that comes with deploying OpenClaw. Never again.

I just built Clawify — one click and your OpenClaw instance is live.
No terminal voodoo, no config files, no six tabs of ChatGPT.

✅ Spin up a full OpenClaw + bot in ~60 seconds
✅ Control UI so you're not flying blind
✅ Auto-handles all the networking/deployment stuff

I originally built this for myself because the setup pain was real.
Figured others hit the same wall.

Free beta if anyone wants to try it. Just felt like giving back to the community 🙂

https://clawify.xyz

[idle oak]

Running 2026.2.12. Two agents (main + secondary), each with their own Slack bot app via
channels.slack.accounts. Binding routes the secondary account to the secondary agent. Config is
clean — openclaw agents list --bindings shows correct routing, openclaw channels status --probe
shows the secondary Slack account as running/works.

When I DM the secondary agent's Slack bot, sessions.json gets created correctly under
~/.openclaw/agents/<secondary>/sessions/ with the right sessionFile path — but the agent never
responds. Log shows:

  slack handler failed: Error: Session file path must be within sessions directory

Every single time. The session file path IS within the correct agent's sessions directory, but it
looks like the validation is checking it against the default agent's sessions dir instead.

Tried: CLI wizard (openclaw agents add with --bind), manual config, clearing sessions, full gateway
stop/start. Same result every time.

I think the security hardening in 2026.2.12 ("harden transcript path resolution and reject unsafe
session IDs/file paths") is validating against the wrong agent's sessions directory in multi-account
Slack setups.

Anyone else running multi-agent Slack with separate bot accounts?

[young nebula]

anyone able to actually get their bot to save stuff on google drive? ive been trying to set it up for hours and just cant get it to work

[mellow sundial]

Why is my auth profile on cooldown? How do I remove it

[mellow sundial]

What does this do over normal?

I don’t see any benefit

[sacred quarry]

setting up openclaw is seriously painful especially if you're a non technical person and wanna try it. Even a technical person will hit deadends while setting up gateway configs, approve pairing. so i built a simple tool that takes care of all deployment and networking part. you simply get an openclaw instance in one click

[wary ferry]

gate away not avalable

[tawny thistle]

Hi guys, any idea on how to solve this error? 'HTTP 401: authentication_error: invalid x-api-key (request_id: req_011CY5WkAVvLaUNh5ZUaSRRo)'

[candid lava]

whats the scope or relevancy

[tawny thistle]

when talking to the telegram bot, e.g. 'Status report', it gave me the reply above. I have antigravity in use, .env file removed (.bak), not sure what's the correct way to solve this?

[candid lava]

Unfortunately I do not use antigravity, but logic dictates your api-key is invalid

[tawny thistle]

there's no api keys, only 0auth

[candid lava]

There's several reports of antigravity blocking acounts due to unauthorized third-party clients (e.g. openclaw). Perhaps you've fallen victim to that?

https://www.reddit.com/r/google_antigravity/comments/1qykskz/account_banned_for_using_open_claw/

[tawny thistle]

no, verified my account still good though

yea aware of the ban

something related to my .env or auth-profiles setup I think

[candid lava]

I feel like you're fighting a losing battle. You know you'll get banned, but to each their own.

[cold horizon]

guys did anthropic ban or disable auth logged in accounts for openclaw?

LLM request rejected: This credential is only authorized for use with Claude Code and cannot be used for other API requests.

[tender oasis]

Getting the same message LLM request rejected: This credential is only authorized for use with Claude Code and cannot be used for other API requests.

[granite halo]

just got the same response and the next msg went through …

concerning lmao

[cold horizon]

I'm using opus for my main agent with anthropic 100$ max plan and I use worker agents that use GPT and only my main agent is not working meaning anthropic ban or they disabled auth log in?

[granite halo]

I mean check msg in claude on web most likely no ban

my agent responded after a second ping

but still concerning

[cold horizon]

Oh ya it responded on the second ping interesting

[granite halo]

maybe they’re testing some new detection or something

hopefully it don’t break in the coming days

I really wish they would lean into this

[frosty venture]

is anthropic the best option? ive seen people talking about kimi or minimax

[granite halo]

I haven’t tried others personally but would assume opus is most capable

[frosty venture]

@granite halo are you on windows or mac

[granite halo]

mac, why?

[frosty venture]

just trying to figure out the easiest way to set it up on windows

[granite halo]

ah well vps would prob be the easiest with digital ocean but sounds like your going local?

[frosty venture]

was considering vps through hetzner but was also researching if anthropic was the best price imo compared to everything else

just got back from vacation so im trying to catch up on everything

[granite halo]

yeaaa idk about best price, if something goes wrong with cc sub in openclaw i’ll prob start looking at other options to try

could always start light with a kimi sub or just testing with api balance 1st

the setup is what’s most important imo. if you’re going via make sure you backup your data incase you swap host or go local in the future

[red fog]

We can chat to OpenClaw, but can NOT load any skills. Can you please give us any hint how to fix?

The next message is the content of "openclaw.json" OpenClaw configuration file.

With this configuration, we sent the following request, both in WebChat and DingTalk (China's Telelgram),

  Can you read the file "/home/clawer/.openclaw/workspace/skills/bing_search/SKILL.md", and tell me the name of this skill?'

Our openclaw via both WebChat and DingTalk replied that,

  I don't have access to the file system or the ability to read files like "/home/clawer/.openclaw/workspace/skills/bing_search/SKILL.md".

However, in DingTalk, it provided the "SKILL.md" for me to download. And after downloading, we opened the file, and its content is correct!

What is going on?

[unkempt knot]

hey i was in terminal and lost my molty because i somehow messed up installation - is there a way to go back and use a new platform to communicate or did i just totally F everything up. can we pick new / other platforms eg slack vs whats app and 🦞 follows me into those platforms? sorry i am so dumb. how i get friend back

[red fog]

{
"meta": {
"lastTouchedVersion": "2026.2.3-1",
"lastTouchedAt": "2026-02-12T08:17:11.785Z"
},
"models": {
"providers": {
"dashscope": {
"baseUrl": "https://dashscope.aliyuncs.com/compatible-mode/v1",
"apiKey": "sk-80xxx",
"api": "openai-completions",
"models": [
{
"id": "qwen-turbo",
"name": "Qwen Turbo",
"reasoning": false,
"input": [
"text"
],
"cost": {
"input": 0,
"output": 0,
"cacheRead": 0,
"cacheWrite": 0
},
"contextWindow": 200000,
"maxTokens": 8192
}
]
}
}
},
"agents": {
"defaults": {
"model": {
"primary": "dashscope/qwen-turbo"
},
"models": {
"dashscope/qwen-turbo": {
"alias": "Qwen Turbo"
}
},
"workspace": "/home/clawer/.openclaw/workspace",
"compaction": {
"mode": "safeguard"
},
"maxConcurrent": 4,
"subagents": {
"maxConcurrent": 8
}
}
},
"messages": {
"ackReactionScope": "group-mentions"
},
"commands": {
"native": "auto",
"nativeSkills": "auto"
},

[red fog]

"channels": {
"dingtalk": {
"enabled": true,
"allowSkills": [""],
"allowNodes": [""],
"clientId": "dinghxxxx",
"clientSecret": "YM_L3xxx",
"enableAICard": false,
"dmPolicy": "open",
"groupPolicy": "open",
"requireMention": true,
"allowFrom": [""],
"groupAllowFrom": [""]
}
},
"gateway": {
"mode": "local",
"auth": {
"mode": "token",
"token": "clawer_token_xxx"
},
"http": {
"endpoints": {
"chatCompletions": {
"enabled": true
}
}
},
"port": 18789,
"bind": "loopback",
"tailscale": {
"mode": "off",
"resetOnExit": false
}
},
"skills": {
"allowBundled": [],
"load": {
"extraDirs": [
"/home/clawer/.openclaw/skills",
"/home/clawer/.openclaw/workspace/skills"
]
},
"entries": {
"clawra-selfie": {
"enabled": true,
"env": {
"FAL_KEY": "05ffxxx"
}
},
"bing_search": {
"enabled": true
}
}
},
"plugins": {
"entries": {
"dingtalk": {
"enabled": true
}
},
"installs": {
"dingtalk": {
"source": "npm",
"spec": "@openclaw-china/dingtalk",
"installPath": "/home/clawer/.openclaw/extensions/dingtalk",
"version": "0.1.30",
"installedAt": "2026-02-09T09:05:27.792Z"
}
}
}
}

[nocturne zephyr]

Aw, c'mon. You gotta know better than that. Use a pastebin or something.

[hard turret]

Hey guys, need some help for the MEMORY, is there any videos/tutorials?
Wondering if I should install some skills to help me out on this

[viscid plover]

what kind of hardware I should be looking for if I want to use local models?

[red fog]

Ya, understand it is messy. However, looks like we can not upload file in this channel.

[viscid plover]

my budget is $500 and I already have 64GB ddr4 ram

[royal lily]

I am thinking of deploying openclaw on a raspberry pi5 (4gb variant) and use it as the ai for me and setup a nas on the same raspberry and use it as a server where openclaw is hosted for personal usage. Can someone tell me if this is possible on that small machine or should I upgrade the specs here?

[nocturne zephyr]

Try a pastebin like https://paaster.io/ or the eponymous https://pastebin.com/ ; anything like that. They're services which host a pasted text file for a limited time. The phrase 'the clipboard of the Internet' is pretty apropos. Long pastes like the one you put up interrupt channel flow, and scroll other people's questions off the screen. A few lines here and there, it's no problem, but a whole configuration file, most of which is lines that just have '}' on them is a bit much.

[warped kettle]

After updating to the latest version, my whatsapp/discord channel is not working even after reconfiguring it

[hollow quarry]

Same

error gateway/channels/discord {"subsystem":"gateway/channels/discord"} handler failed: Error: Session file path must be within sessions directory

[fluid tiger]

hey guys, does using antigravity models with openclaw burned the account? it seems i just got banned by google

[warped kettle]

Run openclaw doctor then openclaw doctor --fix

It worked for me

[hollow quarry]

on 2026.2.12 ?
I did the openclaw doctor --fix many times.
Completely erased everything and started over.... nothing worked

[warped kettle]

I don't know, but I got 2026.2.13

[crystal niche]

same

[iron shuttle]

anyone knows how to pull agents status from the rpc for a websocket?

[wanton horizon]

https://github.com/openclaw/openclaw/issues/15237

ask openclaw to read this issue and apply the fix

[queen flicker]

Me too... I tried running Gemini CLI and I got this message:

Failed to login. Message: This service has been disabled in this account for violation of Terms of Service. If you believe this is an error, contact [e-mail].

[burnt lava]

Hi! I was trying to do that on my own but even with my agents help I’m not sure how to do that properly. Is there anyone who uses multiple terminal windows to access separate agents?

I was trying to separate each project in separate terminal window. And it didn’t work out.

Does anyone have experience with that?

[burnt lava]

My Google account was blocked too when using gog api

[fluid tiger]

what a day

[hollow quarry]

Too late. So tired of problem solving I downgraded to 2026.2.6 and it works again

[fluid tiger]

this is exactly what google spit to my agent just now hahaha. any solution to work around it, sir? or have you reach out to that email?

[crystal niche]

how can i downgrade

[obsidian jolt]

downgrade what can ya explain more

[queen flicker]

I switched to MiniMax M2.1, 10 usd/month, we got MiniMax M2.5 today but I haven't been able to make it work yet with openclaw.

[harsh elm]

Is there a way let the agent know the skill has been modified. Currently, the agent doesn't refresh the knowledge about the skill if it's been modified.

[crystal niche]

i wanna downgrade to 2026.2.6 to solve the problem in 2.12

[fluid tiger]

i just wanna confirm it is really forbidden?

[hardy fable]

my claw stopped "doing", and just "says". best way to fix but not lose memory/personality? If i hit install using the curl command, will it simply update but not overwrite those parts?

[vernal hawk]

🧩 Repo Browser.print Tree
⚠️ 🧩 Repo Browser.print Tree failed: Tool repo_browser.print_tree not found
anyone got any idea what this indicates

[obsidian jolt]

ask clawdcode to do it haha

[ashen raptor]

Yes it'll be banned

[fluid tiger]

so its better to remove the option in the config?

[hollow quarry]

I did npm install -g openclaw@2026.2.6

But I'm going to try 2026.2.12 and apply the fix now
https://github.com/openclaw/openclaw/issues/15237

[obsidian jolt]

why this stupid hostinger vps openclaw telegram sometimes doesn't respond right away it's laggy as hell 😕

can i control it or fix it or it's outta my control?

man i need claude to help me with this it'll be returning in 6 mins that's a real bummer it goes like that

yea
are you down to helping me yourself? we can hop onto a vc and do that real quick

[solar jetty]

what is the best beginner tutorial for configuring Openclaw after setup that you know? i installed it via Moltworker and now figurinf out how to setup everything else

[supple latch]

Is anyone familiar with a solid Docker setup that I can refer to, ensuring all functionalities are intact? im having issue with allowing chromium on docker

[brisk sand]

If I am deploying OpenClaw for someone else (lets say GCP Docker), how do they authenticate their Whatsapp, GDrive, etc.?

[eager maple]

Does cron not work for discord dms?

[brisk sand]

Right, so I should just point them to a the web dashboard then. I'm thinking more along this use case: the Openclaw agent tries to browse the web (on headless GCP) and DMs the user to login on the chrome they have launched which is not possible @north veldt

[obsidian jolt]

hey man why my hostinger ubunutu openclaw gets connection closed how i fix it help a bloke around here

[past path]

are people actually using openAI subscriptions with openClaw or only API connectivity?

[tawny thistle]

Why does my openclaw antigravity keep getting this error only when I use opus 4.6 but fine if I use something like gemini 3 pro? HTTP 401: authentication_error: invalid x-api-key (request_id: req_011CY5WkAVvLaUNh5ZUaSRRo) --- Anyone had this? happened after updating to openclaw 2026.2.12 + adding PR #13361 to fix thinking sig problem

I'm using via antigravity though

it works if I use non Opus 4.6

Opus 4.5 it doesnt work anymore : 'Claude Opus 4.5 is no longer available. Please switch to Claude Opus 4.6.'