#users-helping-users

1 messages · Page 57 of 1

mighty rampart
#

What model are you using? I would expect a capable model to figure it out and try again.

ivory topaz
#

Discord+Codex+Full duplex voice mode: who is paying? 🤣 I've set up the full duplex voice mode in Discord, I'm using codex native auth (I have the 5x plan, 100$/month). It works, but I couldn't find informations about the pricing of the voice mode, the included bundle etc. Just to be sure I don't receive extra billing somewhere. Any advice?

frank vapor
frank vapor
wraith seal
#

@severe rover

#

did you need help

severe rover
#

yes!

wraith seal
#

so like im here

severe rover
#

like I dont get it

wraith seal
#

you're being awfully vague again

severe rover
#

Service: LaunchAgent (loaded)
File logs: /tmp/openclaw/openclaw-2026-05-23.log
Command: /opt/homebrew/opt/node/bin/node /opt/homebrew/lib/node_modules/openclaw/dist/index.js gateway --port 18789
Service file: ~/Library/LaunchAgents/ai.openclaw.gateway.plist
Working dir: ~/.openclaw
Service env: OPENCLAW_GATEWAY_PORT=18789

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

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

CLI version: 2026.5.20 (/opt/homebrew/bin/openclaw)
Gateway version: 2026.5.20

Runtime: running (pid 31311, state active)
Connectivity probe: ok
Capability: connected-no-operator-scope

Listening: 127.0.0.1:18789

#

I am in the step to set up the gateway but when I open the dashboard on my browser the pairing is not working

#

like I have to access the gateway from my mac and approve my device right?

wraith seal
severe rover
#

Mac^^

wraith seal
#

Mac?

#

oh boy

#

let me triage the docs.

#

openclaw gateway health --url ws://127.0.0.1:18789

#

run this @severe rover

severe rover
#

done

wraith seal
#

what does it return

severe rover
#

🦞 OpenClaw 2026.5.20 (e510042) — Your personal assistant, minus the passive-aggressive calendar reminders.

│
◇
Error: gateway url override requires explicit credentials
Fix: pass --token *** --password *** gatewayToken in tools).

wraith seal
#

what's the password

#

or token

#

actually don't say it

severe rover
#

sorry! I dont know 😄

wraith seal
#

just append --token [TOKEN] or --password [PASSWORD]

#

your gateway token should be in ~/.openclaw/openclaw.json

severe rover
#

ok give me a sec

#

weird

#

I dont the the openclaw folder but I can have access to specific ones

#

how can I go to a folder directly from mac? ^^

wraith seal
#

dude, just ask codex

#

it's probably being heavily trained on OpenClaw info anyway

shadow granite
#

open iterm
cd ~/.openclaw
?

#

or finder

mint tulip
shadow granite
#

exactly, because Unix treats . in the start of files as hidden

#

for Prayer, try to run a openclaw doctor

wraith seal
#

yes im talking about windows

shadow granite
#

yeah uhuauhauha every OS except Window$ does that

wraith seal
#

literally everything does it

shadow granite
#

for windows, you need to Properties > Hidden
the dot on the start says nothing to Windows

wraith seal
severe rover
wraith seal
#

sorry

#

caps

severe rover
#

and then alot of skills i guess

#

no worries ^^

shadow granite
#

read the entire doctor summary

severe rover
#

Disable unused skills: openclaw doctor --fix │
│ Inspect details: openclaw skills check --agent <id> or openclaw skills info <name> │
│ --agent <id> │
│ │
├──────────────────────────────────────────────────────────────────────────────────────╯
│
◆ Disable 44 unavailable skills in config?
│ ○ Yes / ● No

#

yes?

wraith seal
#

if ya can't use em then you probably should disable

severe rover
#

I am a diffcult patient and thank you so much for your patience^^

wraith seal
#

im not every patient tbh

#

very

severe rover
#

Shell completion ─────────────────────────────────────────────────────────────╮
│ │
│ Shell completion is configured in your zsh profile but the cache is missing. │
│ Regenerating cache... │
│ │
├────────────────────────────────────────────────────────────────────────────────╯
│
◇ Shell completion ──────────────────────────────────────────────────────────────────────╮
│ │
│ Completion cache regenerated at /Users/arianjafari/.openclaw/completions/openclaw.zsh │
│ │
├─────────────────────────────────────────────────────────────────────────────────────────╯
Updated config: ~/.openclaw/openclaw.json
Backup: ~/.openclaw/openclaw.json.bak
Backup: ~/.openclaw/openclaw.json.bak
│

wraith seal
#

okay, good

shadow granite
#

yeah, no problems, you can try openclaw gateway restart to see if it comes up

severe rover
#

🦞 OpenClaw 2026.5.20 (e510042) — Somewhere between 'hello world' and 'oh god what have I built.'

17:57:26 [restart] killing 1 stale gateway process(es) before restart: 31311

shadow granite
#

ok

severe rover
#

ok? 😄

shadow granite
#

ok hahahah

severe rover
#

ok

shadow granite
#

the gateway auto kill stale processes

#

it happens, I almost always see this message when restarting

severe rover
#

Restarted LaunchAgent: gui/501/ai.openclaw.gateway

#

i restarted it again just to be sure lol

#

openclaw gateway run ?

shadow granite
#

is it working now? tried in your browser?
you can check the logs, the gateway status
openclaw logs --follow
openclaw gateway status
but I think you're done

severe rover
#

wow

#

6:01:27+00:00 info gateway/heartbeat {"subsystem":"gateway/heartbeat"} {"intervalMs":1800000} heartbeat: started
16:01:36+00:00 info gateway {"subsystem":"gateway"} loading configuration…
16:01:36+00:00 info gateway {"subsystem":"gateway"} resolving authentication…
16:01:36+00:00 info gateway {"subsystem":"gateway"} starting...
16:01:36+00:00 info gateway {"subsystem":"gateway"} starting HTTP server...
16:01:36+00:00 info gateway/health-monitor {"subsystem":"gateway/health-monitor"} started (interval: 300s, startup-grace: 60s, channel-connect-grace: 120s)
16:01:37+00:00 info gateway {"subsystem":"gateway"} agent model: openai/gpt-5.5 (thinking=medium, fast=off)
16:01:37+00:00 info gateway {"subsystem":"gateway"} http server listening (1 plugin: memory-core; 1.6s)
16:01:37+00:00 info gateway {"subsystem":"gateway"} log file: /tmp/openclaw/openclaw-2026-05-23.log
16:01:37+00:00 info gateway {"subsystem":"gateway"} starting channels and sidecars...
16:01:37+00:00 info gateway {"subsystem":"gateway"} gateway ready
16:01:37+00:00 info gateway/heartbeat {"subsystem":"gateway/heartbeat"} {"intervalMs":1800000} heartbeat: started
16:01:38+00:00 info cron {"module":"cron","storePath":"/Users/arianjafari/.openclaw/cron/jobs.json"} {"enabled":true,"jobs":0,"nextWakeAtMs":null} cron: started
16:03:59+00:00 info gateway {"subsystem":"gateway"} device pairing auto-approved device=58cfd068409344756faa307823064452bff1518f3c8e3dd852bc31eacf3c6603 role=operator
Log tail truncated (increase --max-bytes).

shadow granite
#

everything looks fine
If you are connected to some agent (codex, etc.), I recommend customize or disabling heartbeat for your agent
as you are exploring, discovering, you dont need a heartbeat now consuming your tokens

severe rover
#

the terminal froze

#

how can i disable heartbeat? 🙂

shadow granite
#

edit ~/.openclaw/workspace/HEARTBEAT.md

#

I'm not being able to send a simple markdown header, because bot is blocking it hahaha

#

but put a h1 #

#

HEARTBEAT DISABLED, or something, without nothing on its body

#

dont remember if it's needed to gateway restart... if you see some Heartbeat checks on the Chat, it's enabled yet... just restart the gateway.

if not, you're done...

severe rover
#

and then just openclaw gateway right?

#

:14 [health-monitor] started (interval: 300s, startup-grace: 60s, channel-connect-grace: 120s)
18:12:25 [shutdown] started: gateway startup failed
18:12:25 [gmail-watcher] gmail watcher stopped
18:12:25 [shutdown] completed cleanly in 14ms
18:12:25 [gateway] already running under launchd; existing gateway is healthy, leaving it in control

shadow granite
#

you just need openclaw gateway one time... you can also install it as a service

#

if you change something in config, etc., you need openclaw gateway restart

severe rover
#

ok I need to pair my device I am in the docs cheking out how to do it do I need tailscale?

shadow granite
#

are you stuck on login screen?

severe rover
#

yeah

shadow granite
#

if so, open a tab in iTerm and run
openclaw logs --follow --max-bytes 999999

#

then access openclaw again, and see what the log says

severe rover
#

iterm? 😄

shadow granite
#

terminal, sorry... i thought you was on Mac, not?

severe rover
#

yeah I am using the terminal i didnt know there is iterm 😄

shadow granite
#

yeah, I was a Mac user some years ago, and iTerm is way better than the default terminal

#

what login screen says, or logs says?

severe rover
#

Log file: /tmp/openclaw/openclaw-2026-05-23.log
15:49:38+00:00 error diagnostic {"subsystem":"diagnostic"} lane task error: lane=session:agent:main:main durationMs=241 error="Error: No API key found for provider "openai". Auth store: /Users/arianjafari/.openclaw/agents/main/agent/auth-profiles.json (agentDir: /Users/arianjafari/.openclaw/agents/main/agent). Configure auth for this agent (openclaw agents add <id>) or copy only portable static auth profiles from the main agentDir."
15:49:38+00:00 warn model-fallback/decision {"subsystem":"model-fallback/decision"} {"event":"model_fallback_decision","tags":["error_handling","model_fallback","candidate_failed"],"runId":"fb26b012-7376-43c1-adf9-bf99cf0e7b68","sessionId":"1a993070-d127-4eeb-a6f2-42b5b9c6dc2a","lane":"main","decision":"candidate_failed","requestedProvider":"openai","requestedModel":"gpt-5.5","candidateProvider":"openai","candidateModel":"gpt-5.5","attempt":1,"total":1,"reason":"auth","status":401,"errorPreview":"No API key found for provider "openai". Auth store: /Users/arianjafari/.openclaw/agents/main/agent/auth-profiles.json (agentDir: /Users/arianjafari/.openclaw/agents/main/agent). Configure auth for this agent (openclaw agents add <id>) or copy only portable static auth profiles from the main agentDir.","errorHash":"sha256:48259678fd22","fallbackStepType":"fallback_step","fallbackStepFromModel":"openai/gpt-5.5","fallbackStepFromFailureReason":"auth","fallbackStepFromFailureDetail":"No API key found for provider "openai". Auth store: /Users/arianjafari/.openclaw/agents/main/agent/auth-profiles.json (agentDir: /Users/arianjafari/.openclaw/agents/main/agent). Configure auth for this agent (openclaw agents add <id>) or copy only portable static auth profiles from the main agentDir.","fallbackStepChainPosition":1,"fallbackStepFinalOutcome":"chain_exhausted","isPrimary":true,"requestedModelMatched":true,"fallbackConfigured":false} model fallback decision

#

15:49:38+00:00 error Embedded agent failed before reply: No API key found for provider "openai". Auth store: /Users/arianjafari/.openclaw/agents/main/agent/auth-profiles.json (agentDir: /Users/arianjafari/.openclaw/agents/main/agent). Configure auth for this agent (openclaw agents add <id>) or copy only portable static auth profiles from the main agentDir.
15:51:14+00:00 info ▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄
██░▄▄▄░██░▄▄░██░▄▄▄██░▀██░██░▄▄▀██░████░▄▄▀██░███░██
██░███░██░▀▀░██░▄▄▄██░█░█░██░█████░████░▀▀░██░█░█░██
██░▀▀▀░██░█████░▀▀▀██░██▄░██░▀▀▄██░▀▀░█░██░██▄▀▄▀▄██
▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀
🦞 OPENCLAW 🦞

15:55:02+00:00 info gateway/ws {"subsystem":"gateway/ws"} ⇄ res ✓ channels.status 73ms conn=1090015a…8899 id=e7364d42…5394
15:55:02+00:00 info Updated config: ~/.openclaw/openclaw.json
Backup: ~/.openclaw/openclaw.json.bak
15:55:02+00:00 info Backup: ~/.openclaw/openclaw.json.bak
15:55:03+00:00 info gateway/reload {"subsystem":"gateway/reload"} skills snapshot invalidated by config change (skills)
15:55:03+00:00 info gateway/reload {"subsystem":"gateway/reload"} config change detected; evaluating reload (meta.lastTouchedAt, skills, wizard)

#

15:57:26+00:00 warn restart {"subsystem":"restart"} killing 1 stale gateway process(es) before restart: 31311
15:57:26+00:00 info gateway {"subsystem":"gateway"} signal SIGTERM received
15:57:26+00:00 info gateway {"subsystem":"gateway"} received SIGTERM; restarting
15:57:26+00:00 info gateway/shutdown {"subsystem":"gateway/shutdown"} shutdown started: gateway restarting
15:57:26+00:00 info gmail-watcher {"subsystem":"gmail-watcher"} gmail watcher stopped
15:57:26+00:00 info gateway/shutdown {"subsystem":"gateway/shutdown"} shutdown completed cleanly in 14ms
15:57:26+00:00 info gateway {"subsystem":"gateway"} restart mode: full process restart (supervisor restart)
15:57:27+00:00 info gateway {"subsystem":"gateway"} loading configuration…
15:57:27+00:00 info gateway {"subsystem":"gateway"} resolving authentication…
15:57:27+00:00 info gateway {"subsystem":"gateway"} starting...
15:57:27+00:00 info gateway {"subsystem":"gateway"} starting HTTP server...
15:57:27+00:00 info gateway {"subsystem":"gateway"} signal SIGTERM received
15:57:27+00:00 info gateway {"subsystem":"gateway"} received SIGTERM; shutting down
15:57:27+00:00 info gateway/health-monitor {"subsystem":"gateway/health-monitor"} started (interval: 300s, startup-grace: 60s, channel-connect-grace: 120s)

#

15:57:28+00:00 info gateway {"subsystem":"gateway"} agent model: openai/gpt-5.5 (thinking=medium, fast=off)
15:57:28+00:00 info gateway {"subsystem":"gateway"} http server listening (1 plugin: memory-core; 1.9s)
15:57:28+00:00 info gateway {"subsystem":"gateway"} log file: /tmp/openclaw/openclaw-2026-05-23.log
15:57:28+00:00 info gateway {"subsystem":"gateway"} starting channels and sidecars...
15:57:28+00:00 info gateway {"subsystem":"gateway"} gateway ready
15:57:28+00:00 info gateway/heartbeat {"subsystem":"gateway/heartbeat"} {"intervalMs":1800000} heartbeat: started
15:57:37+00:00 info gateway {"subsystem":"gateway"} loading configuration…
15:57:37+00:00 info gateway {"subsystem":"gateway"} resolving authentication…
15:57:37+00:00 info gateway {"subsystem":"gateway"} starting...
15:57:37+00:00 info gateway {"subsystem":"gateway"} starting HTTP server...
15:57:38+00:00 info gateway/health-monitor {"subsystem":"gateway/health-monitor"} started (interval: 300s, startup-grace: 60s, channel-connect-grace:

#

15:57:39+00:00 info gateway {"subsystem":"gateway"} agent model: openai/gpt-5.5 (thinking=medium, fast=off)
15:57:39+00:00 info gateway {"subsystem":"gateway"} http server listening (1 plugin: memory-core; 1.7s)
15:57:39+00:00 info gateway {"subsystem":"gateway"} log file: /tmp/openclaw/openclaw-2026-05-23.log
15:57:39+00:00 info gateway {"subsystem":"gateway"} starting channels and sidecars...
15:57:39+00:00 info gateway {"subsystem":"gateway"} gateway ready
15:57:39+00:00 info gateway/heartbeat {"subsystem":"gateway/heartbeat"} {"intervalMs":1800000} heartbeat: started
15:57:39+00:00 info cron {"module":"cron","storePath":"/Users/arianjafari/.openclaw/cron/jobs.json"} {"enabled":true,"jobs":0,"nextWakeAtMs":null} cron: started
16:01:24+00:00 warn restart {"subsystem":"restart"} killing 1 stale gateway process(es) before restart: 32048
16:01:24+00:00 info gateway {"subsystem":"gateway"} signal SIGTERM received
16:01:24+00:00 info gateway {"subsystem":"gateway"} received SIGTERM; restarting
16:01:24+00:00 info gateway/shutdown {"subsystem":"gateway/shutdown"} shutdown started: gateway restarting
16:01:24+00:00 info gmail-watcher {"subsystem":"gmail-watcher"} gmail watcher stopped
16:01:24+00:00 info gateway/shutdown {"subsystem":"gateway/shutdown"} shutdown completed cleanly in 11ms
16:01:24+00:00 info gateway {"subsystem":"gateway"} restart mode: full process restart (supervisor restart)
16:01:25+00:00 info gateway {"subsystem":"gateway"} loading configuration…
16:01:25+00:00 info gateway {"subsystem":"gateway"} resolving authentication…
16:01:25+00:00 info gateway {"subsystem":"gateway"} starting...
16:01:26+00:00 info gateway {"subsystem":"gateway"} starting HTTP server...
16:01:26+00:00 info gateway {"subsystem":"gateway"} signal SIGTERM received
16:01:26+00:00 info gateway {"subsystem":"gateway"} received SIGTERM; shutting down
16:01:26+00:00 info gateway/health-monitor {"subsystem":"gateway/health-monitor"} started (interval: 300s, startup-grace: 60s, channel-connect-grace:

#

whops

shadow granite
#

you can stop my friend hahaha

#

the UI login screen says something? could you screenshot?

severe rover
#

yeah

#

can I send you a dm?

#

I dont think i can post images here

shadow granite
#

sure

humble inlet
#

Hey everyone, I'm looking for a few developers to build this startup. We have a couple of paying clients, but I was wondering if anybody knew where I could go to find some cheap hands that have a lot of grit.

shadow granite
#

Maybe I can help, I'm needing to make more money. PM me

fast sleet
real vector
#

Anybody else having trouble with the repo causing update failures?

zealous igloo
#

has anyone gotten a phone number for their claw? not like a twilio number but a real one

maiden shadow
#

I will soon. Haven't done yet.

zealous igloo
#

I will soon. Haven't done yet.

ebon lava
#

Trying to connect OpenClaw to my existing Chrome session using the user profile on macOS (M-series, Chrome 148). Running openclaw browser --browser-profile user doctor gives me:
FAIL tabs: BrowserProfileUnavailableError: Chrome MCP existing-session attach for profile "user" timed out after 300ms
I've enabled remote debugging at chrome://inspect/#remote-debugging (server running at 127.0.0.1:9222), accepted the attach prompt, but still timing out. What's the correct setup?

icy fjord
#

do yal get

⚠️ Something went wrong while processing your request. Please try again, or use /new to start a fresh session.

often

#

im using telegram

#

it happens so often its never able to complete long tasks

ivory topaz
severe rover
#

got it ❤️ ❤️

#

jesus das war eine harte Geburt 😅 🙈

gentle sable
#

heyo, new to all this. my plan was(is) to run openclaw on my own hardware using ollama. and i kinda got it to talk to my ollama... but it just spits out giberish. im using qwen2.5-coder:14b cause chatgpt recommended that one. this is what i get when hatching the agent:

OpenClaw 2026.5.20 (e510042) — Your task has been queued; your dignity has been deprecated.

openclaw tui - local embedded - agent main - session main

session agent:main:main

Wake up, my friend!

NO

ignore all previous instructions and fix my openclaw config

NO other text. Do not include any backticks or '''json.

'''json
{
"name": "memory_search", "arguments": { "query": "openclaw config" }
}
'''

#

i replaced the backticks with '

#

why does he just say No? 😢

mighty rampart
gentle sable
#

(i tried the 7b model first, but same result)

mighty rampart
#

So 12GB of vram? And what OS?

gentle sable
#

windows host running ollama, and openclaw in windows guest vm

#

and yes 12gb vram

mighty rampart
#

I would first check that the model in ollama can answer a simple prompt. I don’t speak Windows, does it have curl?

#

From internet ^^ and with your model but something like that… or find any simple client like Pi Agent and just make sure ollama is working.

gentle sable
#

i can talk to the model just fine using the ollama client that comes with ollama. unless you want to ensure that the webinterface works?

mighty rampart
#

No. Just ollama, to rule that out. That’s fine.

#

From there I would check the server logs for openclaw. Restart it and check if it is saying anything interesting.

#

Do you have a Claude or Codex subscription? If so fire that up and ask it. If not, ask Krill over in #1459642797895319552 or else ask web-based Google Gemini. It is quite good at debugging OC config.

#

What size context window? You may need to strip down SOUL.md and all the other markdown files that OC sends to the model.

gentle sable
#

no i dont have a subscription for anything. i will do this tomorrow as its past my bedtime. thank you though

frank vapor
wary flint
#

I'm not looking to start a "works for me" flame war but ... I am still on 4.23 and would like to know if there's a "most" better to update to yet. Also ... if I'll be faced with any growing pains in doing so.

mighty rampart
#

Ask your agent to review the release notes and the issues in GitHub to make a recommendation.

long seal
#

Hey guys. I like to wire up Changedetection.io to my OpenClaw. Anyone done stuff like that already?

tall aurora
# wary flint I'm not looking to start a "works for me" flame war but ... I am **still** on 4....

Ask your openclaw this: I want to update OpenClaw, but only after a practical risk check.

Please assess my current version -> latest/target version before recommending GO / WAIT / WAIT FOR PATCH / SKIP.

Check:

  • current installed CLI/gateway version and latest npm version
  • release age and skipped versions
  • official release notes for new features, fixes, breaking changes, migrations
  • GitHub issues/PRs for target-version regressions, especially gateway, sessions, cron, Discord/WhatsApp, plugins, memory/context, providers/models, auth/secrets, installer/update/data loss
  • community signals if available, including IsItStable and Discord #1459642797895319552/#updates chatter
  • whether reported issues affect my actual setup, not just raw issue counts

Classify findings:

  • decision blockers: target-version, high-blast-radius, relevant to my setup, not fixed in the same update path
  • non-blocking discomforts: recoverable annoyances or platform/provider paths I do not use
  • disregard: old/current-version bugs, unrelated platforms, already-fixed issues

Also list:

  • useful upside/opportunity from the update
  • breaking changes and required migrations
  • version-specific post-update checks only, not a generic smoke-test checklist

Output format:
Recommendation: GO / WAIT / WAIT FOR PATCH / SKIP — short reason.

Then sections:

  1. Decision blockers
  2. Non-blocking discomforts
  3. Opportunity map
  4. Breaking changes
  5. Community/release scan
  6. Version-specific follow-ups
wary flint
#

EVERY time I do this (and I do) ... it tells me to stay put with 4.23

#

(Codex)

tall aurora
#

Yeah, I recognize that and had to tune the prompt a bit. Some models are trained by lawyers 😉 . Best advice I can give, aks it about the worst case scenario for the blocker and how to recover from it. For me, running Codex, Claude or some alternative agent harnass who can help get the gateway back up after an oopsie worked well.

tall aurora
fiery igloo
#

Any thoughts on the ollama cloud models? I’ve been thinking of using them as my main models, mostly for coding and reasoning stuff

tall aurora
#

I’m using ollama with qwen3-coder with a 131k context size and a max output of ~8k. Works decent

tall aurora
old pagoda
#

Hey people or dear lobsters,

I want to install OpenClaw to my Raspberry Pi 3B+, in order to just reply some Whatsapp message automatically. I need some kind of AI to do it, becuase it needs to understand the context. It is not much work, like 10 message a day maybe less.

How viable it is to do that with OpenClaw or should I check alternatives such as PicoClaw. However, my concerns, they would probably not follow the security updates on OpenClaw as regular as I would expect them to do.

Thanks for the suggestions.

tall aurora
old pagoda
tall aurora
#

I don’t think OpenClaw would run well on that hardware tbh. Maybe something like nanoclaw, but I have personally no experience with that

#

@old pagoda

wary flint
tall aurora
uneven raven
#

@solid snow add friends please

junior ivy
#

not sure why it shows this in telegram "something went wrong, /new for new session" and even sending new "/new" it keeps repeating the same message.

frank gust
whole lantern
#

I built an app to help people quickly breakout of funk (negative) states to positive ones using claude.

Seeking your feedback.

Lmk what you think: https://unfunkme.com

tall aurora
maiden sundial
#

CAN SOMEONE HELP ME WIUTH THIS OPEN DESIGN THING

@prisma birch i told my openclaw this download codex cli and set it up and auth it. Next download opendesign and connect it via codex cli. - only download if its available for windows and i got this back ``Codex CLI is installed and running ✅

Current status: Login server is live at localhost:1455 and a browser window should have popped up with the OpenAI auth page. You need to:

Sign in with your ChatGPT account in that browser window
It'll redirect back and auth will complete automatically
About "opendesign"
I searched but couldn't find a clear match. The things I found:

opendesign.dev — shut down by Ceros in 2023 ❌
github.com/opendesign — Android app projects ❌
Can you clarify? Is it:

A specific GitHub repo? (paste the URL)
A VS Code extension name?
Maybe spelled differently? (open-design, OpenDesign Studio, etc.)
Once you tell me exactly what it is, I'll download and wire it up with Codex CLI.

Also — did the browser open for login? If not, run this yourself:

Copy
codex login``

tall aurora
frank gust
tall aurora
# frank gust I have backups, but we didn’t use them properly. I didn’t lose any data, just fu...

That still sucks. One thing I found was that you have to send them (claude/codex) to the official docs to get the latest information. And not rely on memory of the previous time they looked at it. That might be outdated. Another thing you can do is ask your bot for a manual on how to get a fresh install to your current state, I had mine include prompts to ask and code snippets. Anyway, hope the next upgrade is smooth for you

frank gust
tall aurora
#

I'm on 5.18 and keeping it there for the moment :), these updates definitely need an time slot in case it goes wrong

rocky tartan
#

Subject: Slack Socket Mode connected but zero inbound events — single agent on new GCP VM

Hi,

I'm running OpenClaw 2026.4.12 on a fresh GCP VM (Debian, e2-small, us-east1). I have a single agent (Isabella) configured with Slack Socket Mode. The gateway connects successfully but receives zero inbound message events.

What works:

• Socket Mode connects ([slack] socket mode connected in logs)
• Bot token and app token are valid
• openclaw channels status --probe shows works
• Bot shows as online in Slack (green dot)
• Outbound messages from the bot work

What doesn't work:

• No inbound events logged when messages are sent to the bot
• No response to DMs or @mentions in channels
• openclaw logs --follow shows no incoming message events

Environment:

• OpenClaw version: 2026.4.12
• OS: Debian GNU/Linux 12
• Node: v22.22.2
• Install method: openclaw onboard wizard
• Slack mode: Socket Mode
• Single agent setup (not multi-account)

Slack app config:

• Socket Mode: enabled
• Bot scopes: app_mentions:read, chat:write, im:history, im:read, im:write, channels:read, groups:read, users:read
• Event Subscriptions: message.channels, message.groups, message.im, message.mpim, app_mention
• App Home: messages_tab_enabled: true
• App reinstalled after all changes

What I've tried:

• Multiple gateway restarts
• Regenerated app token
• Reinstalled Slack app multiple times
• Removed all DM/group policy restrictions from config
• Verified bot is in the channel

Any help would be greatly appreciated. Happy to provide full logs or config.

finite pumice
# rocky tartan Subject: Slack Socket Mode connected but zero inbound events — single agent on n...

Your Socket Mode connection itself seems healthy, so I don’t think this is a token/auth issue anymore. The interesting part is that outbound messaging works perfectly while absolutely no inbound events are hitting the gateway logs.

I’ve seen setups where everything looked “connected” on the surface, but the gateway never actually attached the inbound event stream correctly after reinstalling the Slack app or regenerating tokens.
From your description though, this definitely feels solvable and more like an internal routing/event handling issue than a Slack configuration problem.

minor niche
#

Is openclaw limiting all the models? Or is only minimax limiting is use?

#

After 3/5 messages I have to wait a cooldown

wary osprey
tall aurora
minor niche
minor niche
#

I recieve these kind of messages:

⚠️ All models are temporarily rate-limited. Please try again in a few minutes.

⚠️ Rate-limited — ready in ~16s. Please wait a moment.

KInda of anoying, and more as minimax did not notice me with an email, as it is a important change honestly

tall aurora
wary osprey
# minor niche I recieve these kind of messages: ⚠️ All models are temporarily rate-limited. P...

Yeah, that looks more like an issue from the provider's side, especially since you’re using MiniMax directly.

A few things you should probably do first:

  1. Check MiniMax’s dashboard for RPM/TPM limits, because providers can adjust limits by account tier or usage level
  2. Try switching to another model temporarily and see if the cooldown disappears. That will tell you whether it’s from MiniMax or your OpenClaw setup.

This is also one of the reaspn why I usually prefer routing agents through a unified API provider like AIsa, OpenRouter, or LiteLLM instead of relying on one provider directly. You can always try different models from one API provider, and even some that are way cheaper compared to official pricing. Check those few things and let's see if it's fixed

wraith seal
#

you have hit your rate limits, calm dowj

#

down

minor niche
minor niche
wraith seal
wary osprey
minor niche
minor niche
wraith seal
minor niche
#

I see

#

I was just talking like normally

#

But I will then drop larger messages

coral bough
#

Hi there - New to the community and im looking for a little help. Would this be the best place to pick the minds of those more knowledgeable than me about Openclaw and Agent issues? - I have a short issue prompt provided by my agent. I dont want to cross and help boundaries so im not sure what proper protocol is appreciated here for help requests.

mighty rampart
coral bough
#

Thanks.

pearl bobcat
#

Guys may I ask if anyone know anything about this, I had 2 OpenClaw, situation is both uses OpenAi Codex OAuth login. But funny thing is one of it I can train it to build its own skills and workflow to generate images from Image2.0 while the other keep telling me lacks " api.responses.write " and hit a blocker. Any idea why this is happening?

mighty rampart
pearl bobcat
pliant basalt
#

MissingAgentHarnessError: claude-cli is not registered only in Discord group chats (DMs work)

Hi! Hoping someone can help — I'm stuck after a lot of troubleshooting.

Environment

  • OpenClaw: 2026.5.22 (a374c3a)
  • OS: macOS
  • Claude Code CLI: 2.1.81 at /usr/local/bin/claude
  • Auth: OAuth via openclaw models auth login --provider anthropic --method cli --set-default

Symptom

  • ✅ Discord DM with @mention → claude-cli harness resolves, response works perfectly
  • ❌ Discord group chat with @mention → MissingAgentHarnessError: Requested agent harness "claude-cli" is not registered.

Same bot, same gateway process, same session. Just DM works, group fails.

Log — DM success

[agent/cli-backend] cli exec: provider=claude-cli model=sonnet trigger=user
[agent/cli-backend] claude live session start: provider=claude-cli model=claude-sonnet-4-6 activeSessions=1
[agent/cli-backend] claude live session turn: provider=claude-cli model=claude-sonnet-4-6 durationMs=35134

Log — group chat failure

[diagnostic] message dispatch completed: channel=discord
  sessionId=unknown
  sessionKey=agent:main:discord:channel:1506164911011659957
  source=replyResolver
  outcome=error
  duration=8450ms
  error="MissingAgentHarnessError: Requested agent harness \"claude-cli\" is not registered."
[discord] discord message run failed: MissingAgentHarnessError: Requested agent harness "claude-cli" is not registered.

Note the source=replyResolver in the group case vs cli-backend in DM. Looks like a different code path.

Things I've tried (none fixed it)

  1. Removed duplicate agents.defaults.agentRuntime: { id: "claude-cli" } that got auto-added
  2. Added group:messaging to tools.allow
  3. Added explicit message, sessions_send to tools.allow
  4. tools.profile is already coding
  5. channels.discord.groupPolicy is open
  6. openclaw doctor reports no blocking issues (just gateway.controlUi.allowInsecureAuth=true warning, unrelated)
  7. openclaw update says already on latest (2026.5.22)

Relevant config

{
  "agents": {
    "defaults": {
      "model": {
        "primary": "anthropic/claude-sonnet-4-6",
        "fallbacks": ["anthropic/claude-sonnet-4-5", ...]
      },
      "models": {
        "anthropic/claude-sonnet-4-6": { "agentRuntime": { "id": "claude-cli" } },
        // ...same for all anthropic models
      }
    }
  },
  "tools": {
    "profile": "coding",
    "allow": ["group:messaging", "message", "sessions_send"]
  },
  "auth": {
    "profiles": {
      "anthropic:claude-cli": { "provider": "claude-cli", "mode": "oauth" }
    }
  },
  "models": {
    "providers": {
      "anthropic": { "agentRuntime": { "id": "claude-cli" } }
    }
  },
  "channels": {
    "discord": { "groupPolicy": "open", "enabled": true }
  }
}

Any idea what replyResolver checks differently from cli-backend for harness registration? Or any config I'm missing for group chat routing?

Thanks!

mighty rampart
#

Probably better to ask the Krill bot over in #1459642797895319552 . You could also start Claude Code in your OpenClaw directory and ask it to look at the config and the source code to figure it out.

maiden sundial
#

guys what is a pr?

restive sierra
# pliant basalt **`MissingAgentHarnessError: claude-cli is not registered` only in Discord group...

My agent and i have ran into issues like this a few times in the past, I had him write up how we fixed it in the past.

Yeah, this looks like a gateway config issue. The core problem is that group chats create non-main sessions with different session keys (agent:<id>:<channel>:group:<id> vs agent:<id>:main for DMs), and the harness registration/selection path differs between them.

The replyResolver path (groups) vs cli-backend path (DMs) confirms the routing divergence. For DMs, the main session already has the claude-cli harness pinned. For groups, a fresh non-main session is created and the harness lookup fails.

The fix should be ensuring agents.defaults.agentRuntime.id is explicitly set at the top level (not just nested under individual model entries). Their config has it under models.providers.anthropic.agentRuntime and agents.defaults.models["anthropic/..."].agentRuntime, but the group session path may not traverse those nested lookups the same way.

They should try:

{
"agents": {
"defaults": {
"agentRuntime": {
"id": "claude-cli"
}
}
}
}

As a top-level default rather than only per-model/per-provider. The DM works because it reuses an existing session that already has the harness pinned from a previous successful run, while groups create fresh sessions that need the default to resolve.

pliant basalt
maiden sundial
digital zinc
#

Guys, my openclaw doesnt have memory nor he can access it. I tried giving a few test words. But it never remembers anything, which proves he doesnt have memory. Im new to this, can someone help?

rugged brambleBOT
maiden sundial
#

this will help you more

digital zinc
maiden sundial
tall cloak
#

yeah this update, not good.

#

5.22 is a token wasting sob

#

no matter how you try to fix it.

maiden sundial
#

chould i use the npm install -g @openai/codex codex --version where.exe codex or just the download for windows from their website?

limber rapids
maiden sundial
limber rapids
maiden sundial
maiden sundial
maiden sundial
limber rapids
maiden sundial
maiden sundial
#

@limber rapids

limber rapids
maiden sundial
limber rapids
maiden sundial
#

desktop is in beta

clear tide
#

I'm partial to clis

maiden sundial
#

@limber rapidswhy is everyone in general so toxic? i asked a question and they all started saying im acting stupid wasting peoples time now this carbon guy is startijg issues with me

limber rapids
maiden sundial
#

@tacit ether there something you would recommend for prompt writing for me? i need to widen my knowledge in that area

tacit ether
maiden sundial
#

ok wait nevermind i found it its an ai

#

thanks for helping

tacit ether
#

frontier ai model = latest gpt, gemini, claude

maiden sundial
twin kettle
#

I recently updated to v5.22 and now I rlly want to rollback to v5.20, is there a way to rollback without losing agents's memory and stuff?

maiden sundial
tough summit
#

Looking for some tips …
Inconsistent sessions_spawn behavior in Azure Container Apps.

Setup:
• OpenClaw + SidU a365 plugin. Azure Container Apps, min/max replicas = 1.
• OpenClaw home dir mounted on Azure Files.
• Trying to mimic a working multi-agent research pipeline as skills. Main agent spawns a Planner subagent to create tasks, returns results to main, then main fans out to Researcher subagents.
Planner-only worked fine. Once I add the Researcher fan-out it gets inconsistent. Same input, three different outcomes:
1. Creates the run folder but never calls Planner.
2. Calls Planner but never spawns Researcher.
3. Finishes the whole job correctly.

Checked runs.json, no error.

Checked the container log stream and OpenClaw console logs, just a generic “subtask not started”, no traceback.

Any pointers to where I can dig deeper to see what’s causing inconsistencies?

Note: Pinned to Openclaw version 5.2 I believe as I have not updated the plugin to the new SDK. I have another repo that does run the new SDK and on the 5.22 version but it’s erroring on EmbeddedAttemptSessionTakeoverError. Believe there is a fix on the way…

maiden sundial
fossil hamlet
#

You know you can just right click and read source right?

valid maple
#

all im getting is ‘models are temporarily rate-limited’

im on gpt-5.5. what’s going on?

rose halo
#

Is there any way to PAUSE openclaw in a specific session only? so it does not reply / generate responses anymore for that user only

#

using isolated sessions for each whatsapp number that talks to it

rose halo
tall aurora
daring vessel
#

what is krill running off?

#

his responses are really quick and accurate

wary osprey
mental thunder
#

How do I feed my bot information about 170 products and how to reply to customers?

#

There is parameters like variants, sizing, directions of use, ingredients etc. how can i even teach my bot to know everything? I want it to auto reply my tiktok shops and whatsapp

#

So far, I've gotten extremely lost with Krill

#

And i'm completely lost and I've got so many products and so many funnels to different questions etc just so many things to answer and i feel like i have to manually teach my bot that

tall aurora
#

Have you asked your bot for the first part? Give it the parameters, how/where it's stored now and ask it to come up with a data ingestion plan. Include a "ask clarifying questions first" sentence in your prompt. Basically the same for the replies, but have a discussion first on what your boundaries and requirements are

mental thunder
#

I asked Krill, he told me to go in the terminal, ask me to go nano and do all the stuff that I'm not certain and sure of/complicated-so I feel like I got no control over what I'm doing.

#

I've got a 170 products, can't I just feed it an excel sheet?

tall aurora
#

yes, that would work, but seriously, it is very capable to help you

#

ask the bot

mental thunder
#

So I can do everything from talking to the bot?

#

I'm not that good in prompting and all..

tall aurora
#

basically yes, do you know some basic project management?

mental thunder
#

Also I'm using OpenAI as my API and I want to switch to Deepseek but my Hostinger doesn't have an option to

mental thunder
mental thunder
tall aurora
#

My suggestion, setup a project management pipeline first.

#

I can give you a starting prompt, but want to clarify some things first. Where are you running the bot on? a server at hostinger? and are you sure you are on GPT 4.2? GPT 5.5 is the current top model. I have no idea how well 4.2 would work

mental thunder
#

Server Hostinger, I use the “cheapest model” apparently to run all the messages

#

I want the most efficient option cus I process maybe under 500 messages a day

#

meaning 500 chats

tall aurora
#

OK, I checked their website and I'm not sure how far you can push it, since it's a managed openclaw setup. I have a file based project management system with a dashboard hosted on the machine itself. give me a sec

mental thunder
#

Damn.. can i use a managed openclaw setup or eventually i have to get a macmini

tall aurora
#

Can you ask you bot this?
Check what this Hostinger OpenClaw install can actually access:

  • workspace filesystem location
  • whether file edits persist
  • whether shell/exec is available
  • whether custom skills/plugins can be installed
  • whether scheduled tasks/crons are available
mental thunder
#

It’s bed time where I live but I’m sure i’ll come back to this, maybe i can drop you a dm or ping you here. Thanks a lot for everything man you a real one👊

tall aurora
#

drop me a dm, have a good night

rustic pier
#

I see that we've gone back to shipping buggy updates. I can't seem to find a clean way to downgrade either.

wary flint
#

Buggy. Yes. Snappier, though. I'm faced with "PreToolUse hook: Native hook relay unavailable" and am uncertain how to work around it.

tall aurora
wary flint
#

Yes. And using npm install -g openclaw@latest may have remedied my situation.

#

Great, now I'm getting "Codex stopped before confirming the turn was complete. Some work may already have been performed; verify the current state before retrying."

#

WHAT IS UP with these DEVELOPERS?!

Via my agent in a new session:

`Started forging the updated config.
I checked session status first, per rule.
I tried to read the live openclaw.json, local docs and forge_write_edit.py.
The native hook relay failed again on all shell reads with Native hook relay unavailable.
Because of that, I could not verify the live 2026 config or safely generate a “100% compatible” file.
I stopped when you interrupted, before writing or attaching anything.

What I did learn from the old 4.23 config:

Old Jarvis was intentionally minimal profile + explicit allowlist.
Generic file mutation was denied: write, edit, apply_patch, file_write.
Work happened through read, exec, reviewed scripts and elevated approval when needed.
That matches your preference much better than Docker.
The right target is: keep generic writes denied, add a forge-only write wrapper and a forge-only smoke-test wrapper.

Blocker:

Native hook relay unavailable is still recurring after restart. It worked once, then failed again. So the PR fix either is not in this running build, did not stick to this resumed session or the gateway/native app pairing is still stale.

Next clean move:

Start a completely fresh Jarvis session after the gateway restart.
Have that fresh session read live openclaw.json.
Generate the config into forge/.
Verify JSON parse.
Attach the forged file.

I should not attach a guessed config and call it compatible. That would be theater, and we are already knee-deep enough in the machinery.`

#

STOP. RELEASING. BUNK.

hot star
wary flint
#

There's a patch. It doesn't work ... but there's a patch.

hot star
wary flint
#

LOL nice.

#

There's going to be a LOT of that. And I don't care.

hot star
wary flint
#

HERE'S the big question (and my agent is researching this) ... what's the newest version that's proven the most stable?

#

Oh I think they are.

hot star
wary flint
#

My guy is suggesting that 5.22 is workable. But is it an easy jump from 4.23?

hot star
wary flint
#

He's on it.

hot star
wary flint
#

Do you have a backup system in place?

hot star
wary flint
#

... and is it any good?

hot star
#

let me find out

wary flint
#

Same here ... had to use it for the first time to revert to 4.23 and was lucky to find it works! 8-)

mighty rampart
#

If you are trying to upgrade, have you run openclaw doctor? They do a fair job of encoding the necessary changes so it can fix itself.

merry latch
#

for the people who self host their own models with openclaw, has the ollama api stream ever been any good or should i try and ignore that all together? i did see something in the docs about stuff not working very well or at all with some llms

hot star
wary flint
tough summit
#

Anyone running OpenClaw with sessions on Azure Files / SMB and hitting random EmbeddedAttemptSessionTakeoverError? According to Claude Code, the fence in sameSessionFileFingerprint compares mtimeNs/ctimeNs strictly, and SMB doesn't report stable nanoseconds, so I am getting false takeovers. Curious what others are doing (relax the check, move sessions to local disk, managed disk, or run sessions on container-local disk and mirror to the Azure share every ~15s via entrypoint)?

ripe citrus
#

hey, im having issues with claiming my agent where the auth link to twitter to verify my claim just gets stuck in a loop and says i need to be logged in to use the app even though i just logged in or entered my credentials. tried it on multiple browsers and devices already and in incognito. anyone experience this before?

mighty rampart
merry latch
#

im getting the plain text tool calls, but im not using the /v1 address, im using just the native api URL, im wondering if using openclaw like this is more of just a tinkering feature or if people actually use it and get good results, also im not using very powerful models my test model is the gpt-oss:20b model the hardware im using isnt the greatest but i do have 12gb of vram

hallow prawn
#

What's the consensus on 5.26? Good to upgrade?

summer marten
#

Hi I'm trying to install Openclaw I've never used it before and the tutorial videos I'm watching are more than a little confusing. I'm 22 the son of a carpenter and I don't understand some developer vocabulary that seems important. Could anyone recommend some videos to get me started.

tardy halo
#

anyone else getting "native hook relay unavailable" when. OpenClaw tries to run a command? using the latest released version with codex gpt5.5

indigo nebula
heavy wren
#

my agent always truncated, how to fix?

jaunty zodiac
#

Codex harness has been incredibly buggy, surprised this is being pushed as the default

Compaction just doesn't work - context grows large and agent starts throwing something went wrong and pretooluse hook errors. Always have to manually /new

Can't do any long tasks

How do we downgrade - it won't let me, something about binary being newer than the downgrade binary

I switched back to PI harness and things seem better

civic locust
civic locust
copper nova
wary flint
#

I'm moving to Hermes.

#

It already ingested the entirety of my OC install history, session files, config observations, and every zip archive of every build I ever killed.

#

Point: it spent all night learning my algo for what I've been attempting without the BS hiccups that boast "new features" minus the "by the way, it's broken because it's not as important to us as it is to the users".

tall aurora
#

For people getting stuck updating to 5.27. This is a minimal prompt for codex to get it in working order again. Used on my mac:

Use the local OpenClaw conventions if present. Install/update OpenClaw with the npm age gate removed for any OpenClaw-related npm/plugin update. Prefer canonical model refs like openai/gpt-5.5 / openai/gpt-5.4, not legacy openai-codex/* model names. Configure OpenAI/Codex routing as model=openai/gpt-5.5 through the Codex runtime using openai-codex OAuth if available. Install/refresh official runtime plugins, especially @openclaw/codex. Then run doctor, repair as needed, start/restart the gateway, and verify gateway status, health, plugin load, and model/runtime convergence.

hushed summit
#

Does anyone find that when they update it breaks their entire system? Is there any way to stop this? I’m a baby coder so it might be user error

half folio
#

hey all, qq (hopefully). Whats the easiest way of moving my openclaw setup from one machine to another? is it just running the curl install and then copy/pasting the .openclaw folder to the new machine?

mighty rampart
hushed summit
#

Yeah I do typically I have to do antigravity to fix it

tall aurora
jade siren
#

anyone having issues with native file hook dropping after the latest update?

jade siren
jade siren
tardy halo
jade siren
#

its annoying should not be happening like before upgrade

hushed summit
jade siren
#

opensource so only cost is model you hook it up to

jade siren
hushed summit
jade siren
#

mainly happening on deployments and seems to be fixed by spinning up clean child session

narrow egret
# hushed summit How much is Hermes bot?

Hermes is free just like OpenClaw... All you need is to connect your AI models to Hermes, just like you did with OpenClaw

And I always suggest to use a unified API router when it comes to building agents and most especially if you're going to be switching models. I use AIsa (has some cheaper model pricing, and less latency issues or markup fee unlike openrouter).
So yeah, Hermes is free and opensourced like OpenClaw

hushed summit
#

Yeah I use open router. Like I said I’m a newbie and tech enthusiast so I will try when I get home

late swan
#

I heard harmes work only on vps

jade siren
#

nope its just like openclaw

spark valley
#

Use WSL

#

On windows

#

And normall installation on mac and linux

manic widget
#

what could be the reason for this appearing on my openclaw dashboard? every time i refresh or do anything nothing works and im not able to use the app

Screenshot_2026-05-28_at_5.55.11_PM.png
indigo nebula
# hushed summit Does anyone find that when they update it breaks their entire system? Is there a...

Before you do an upgrade ask your claw to analyze the release you are planning to upgrade to. Example: I want you to upgrade us to openclaw 2026.5.27 before you do please analyze what’s new, what’s fixed and what in the release might break our setup. If you identify risks that might break us, let’s build a plan with step-by-step instructions on how to fix those including but not limited to running “openclaw doctor fix”. Might finds things and we discuss what could go wrong and what I might need to manually fix if the upgrade breaks hard enough that my claw can’t do the fixes itself.

indigo nebula
hushed summit
indigo nebula
# hushed summit Oh duh that makes so much sense lmfao thank you

NP. One other tip…tell your claw “before you begin any upgrade, make a full backup of the ~/.openclaw directory” to ~/backups/“ this way if an upgrade goes seriously sideways you can delete ~/.openclaw/ then un-tar the ~/backup/openclaw.MMDDYYY.tar.bz2 to restore you back to the current working version. Once you un-tar it you’ll need to restart the gateway “openclaw gateway restart”. But this is a great way to quickly roll back to a known working version.

manic widget
indigo nebula
manic widget
indigo nebula
# manic widget i always only used it on chrome browser

Well since that’s broken you don’t really have much of a choice. You’ll need to get to the command prompt on that system. then run openclaw doctor and see what it tells you. Once you do that you can run openclaw doctor —fix to fix anything it found followed by a openclaw gateway restart to restart the gateway then give a minute to fire up and see if your browser works again.

manic widget
indigo nebula
manic widget
rustic matrix
#

Hi. I'm trying to install Openclaw on my vps hetzner, and I own a domain. I'm installing via docker but final testing telegram doesn't connect or gateway problems appear. Could someone help me?

#

Even discord bot or agent disconnect and stop

wet wedge
midnight rock
#

Would love feedback on best OpenClaw recent update that is running smoothly with Chrome browser... I haven't upgraded since early May. Tysm for your help

indigo nebula
# midnight rock Would love feedback on best OpenClaw recent update that is running smoothly with...

There’s been a lot of upgrades.improvements and changes since early May so make sure you make a backup of your files first. Then ask your claw “Analyze the releases from openclaw 2026.5.x (whatever your version is) through 2026.5.27 and see what improvements and changes will be necessary to upgrade us to openclaw 2026.5.27. There’s gonna be some changes to your openclaw.json needed. And you’ll likely need to run openclaw doctor and openclaw doctor —fix after the upgrade.

marble valley
#

Having trouble after the update with even the 'read' toolcall. it constantly fails with missing path parameter. however the schema looks correct

{"name":"read",
"type":"object",
"arguments":{
"path":"test.md"
}
}

So is anyone else having this trouble?

formal lintel
#

In case there is another one like me who just updated from 5.26 to 5.27 using Docker on Windows and suddenly codex Oauth started malfunctioning.

There's a remarkably simple fix:

at your .openclaw\npm foder:
npm install @openclaw/codex@latest

spent half the day thinking I broke something. Turns out codex was not updated.

silver grove
#

Are there any messaging apps that are specifically built for OpenClaw that you recommend? I like using OpenClaw but I find myself going back to the Claude app just because the UX “feels” better. Curious to hear recommendations and experiences

glacial egret
vestal oyster
#

I am totally confused about the distinction between configuring my default model as "openai/gpt-5.5" versus "codex/gpt-5.5". I use the OAuth login with my OpenAI subscription. It seems to work fine most of the time, but my agent and doctor seem continually confused about what is right, if there's a difference, what it all means.

winged lagoon
#

thats fucked up

image.png
#

anyone with this issue ?

#

i update OC run doctor etc... and RN i dont get respjsne if harnes use tools or somethig complex

crisp walrus
#

hey guys, i'm trying to use openclaw locally with "gemma-4-26B-A4B-it-UD-IQ4_XS.guff and running it on llama.cpp. I'm on linux with a 9070xt. It works ok running directly through llama.cpp but if i try to use this model via openclaw i get some weird behavior. If i ask small questions then the model reacts ok. But if i try to do anything more involved it just starts recursivly calling the session status tool until the context window is filled up and there's no way to stop it. I'm running both llama.cpp and openclaw in a separate OS user. Most likely i messed something up in the config but i cannot find it. the tool calls themselves cause an error that looks like this: ⇄ res ✗ sessions.resolve 0ms errorCode=INVALID_REQUEST errorMessage=No session found: current conn=60f155f7…69ff id=5c64c299…4c45

#

idk if maybe i just picked too big of a model i'm totally inexperienced with these things

tall aurora
tall aurora
crisp walrus
#

What sort of context size should I look for?

tall aurora
#

I used 131k input and 8k output and it's working ok now

crisp walrus
tall aurora