#chat seems unable to respond

1 messages · Page 1 of 1 (latest)

zealous sinewBOT
#

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> âś… Mark Solution

cursive canyonBOT
#

That log line is pretty specific: something in your agent’s model selection is trying to call the anthropic provider, but the running agent can’t find Anthropic credentials in its auth store (…/.openclaw/agents/main/agent/auth-profiles.json).

A couple important nuances:

  • llm-slug-generator runs when a new session is created to generate a human-friendly “title/slug”. So it’s plausible you only see it at session start.
  • But it also means your setup still references an anthropic/... model somewhere (often as the default model or a fallback). If your actual chat model is Anthropic too, this can absolutely cause “no replies”.

What I’d do next (quick + diagnostic)

On the gateway host (your Mac), run and paste the output:

openclaw --version
openclaw models status

That will tell us which model is selected + whether Anthropic is authenticated.

#

Fix options

If you want to use Claude/Anthropic:

  • Use a setup-token (Claude Pro/Max style):
    claude setup-token
    openclaw models auth paste-token --provider anthropic
    
    (or run openclaw models auth setup-token --provider anthropic if you’re generating it on the gateway host)

If you want to use an Anthropic API key:

  • Put it where the daemon can read it (macOS launchd often won’t inherit your shell env):
    • add ANTHROPIC_API_KEY=... to ~/.openclaw/.env
    • then:
      openclaw gateway restart
      
#

If you don’t want Anthropic at all:

  • Change your default model + remove Anthropic from fallbacks (otherwise you’ll keep seeing this). openclaw models status will show what’s currently configured.

If the real symptom is “Discord doesn’t reply”

Run:

openclaw channels status
openclaw logs --follow

Then send a test message in Discord and grab the log lines that appear at that moment.

Docs that match your error (worth skimming):

olive hatch
#
openclaw models status
2026.2.9

:lobster: OpenClaw 2026.2.9 (33c75cb) — Shell yeah—I'm here to pinch the toil and leave you the glory.

Config        : ~/.openclaw/openclaw.json
Agent dir     : ~/.openclaw/agents/main/agent
Default       : openrouter/openai/gpt-4o-mini
Fallbacks (0) : -
Image model   : -
Image fallbacks (0): -
Aliases (1)   : OpenRouter -> openrouter/auto
Configured models (2): openrouter/auto, openrouter/moonshotai/kimi-k2.5

Auth overview
Auth store    : ~/.openclaw/agents/main/agent/auth-profiles.json
Shell env     : off
Providers w/ OAuth/tokens (0): -
- openrouter effective=profiles:~/.openclaw/agents/main/agent/auth-profiles.json | profiles=1 (oauth=0, token=0, api_key=1) | openrouter:default=sk-or-v1...de340102

OAuth/token status
- none
clawd@clawds-Mac-mini ~ % 
#

athropic is not listed in my config anywhere I can find. Openrouter is used for everything

cursive canyonBOT
# olive hatch ```clawd@clawds-Mac-mini ~ % openclaw --version openclaw models status 2026.2.9 ...

That models status output is clean: your default model is OpenRouter, and the only configured provider with creds is openrouter. So the Anthropic error you saw is not coming from your “chat model”.

Given when you see it (“at the time of a new session”), that log line is almost certainly from the session-memory hook: when you run /new, it tries to use an LLM to generate a nice filename slug for the memory snapshot, and if that slugger is (currently) wired to Anthropic on your install, you’ll get:

[llm-slug-generator] Failed to generate slug: No API key found for provider "anthropic"

It should then fall back to a timestamp-y filename (so it’s noisy, but not fatal).

To confirm

On the Mac mini (gateway host):

openclaw hooks check
openclaw hooks info session-memory

To make the warning go away

#

Pick one:

  1. Disable the hook (if you don’t care about /new writing memory files):
openclaw hooks disable session-memory
openclaw gateway restart
  1. Add Anthropic credentials (only if you actually want to use Anthropic for that slugging/other things).

If your real issue is “bot doesn’t reply in that Discord channel”

This Anthropic slug warning won’t usually block replies. Next things to run are:

openclaw channels status --probe
openclaw logs --follow
olive hatch
#
openclaw hooks info session-memory

:lobster: OpenClaw 2026.2.9 (33c75cb) — Pairing codes exist because even bots believe in consent—and good security hygiene.

Hooks Status

Total hooks: 4
Ready: 3
Not ready: 1

Hooks not ready:
  :smiling_imp: soul-evil - config: hooks.internal.entries.soul-evil.enabled

:lobster: OpenClaw 2026.2.9 (33c75cb) — I'm the assistant your terminal demanded, not the one your sleep schedule requested.

:floppy_disk: session-memory âś“ Ready

Save session context to memory when /new command is issued

Details:
  Source: openclaw-bundled
  Path: /opt/homebrew/lib/node_modules/openclaw/dist/bundled/session-memory/HOOK.md
  Handler: /opt/homebrew/lib/node_modules/openclaw/dist/bundled/session-memory/handler.js
  Homepage: https://docs.openclaw.ai/hooks#session-memory
  Events: command:new

Requirements:
  Config: âś“ workspace.dir
clawd@clawds-Mac-mini ~ % 
cursive canyonBOT
# olive hatch ```clawd@clawds-Mac-mini ~ % openclaw hooks check openclaw hooks info session-me...

Yep — that lines up perfectly: the Anthropic error is coming from the session-memory hook (it runs on command:new, i.e. when you issue /new) because that hook does an LLM call to generate the filename slug.

A couple clarifications based on your outputs:

  • session-memory is enabled + ready âś…
  • Your actual chat model is OpenRouter (openrouter/openai/gpt-4o-mini) and auth is fine âś…
  • So this Anthropic thing is not your main model auth — it’s just the slug-generation step failing and then it should fallback to a timestamp filename (so: noisy, but usually not blocking replies)

Next: does it still happen now?

The Anthropic log you pasted earlier was at 15:23, but your models status now shows no Anthropic configured. If you changed models since then, it might already be gone.

Can you do this once:

  1. run openclaw logs --follow
  2. issue /new in Discord
  3. see if it still prints the Anthropic slug error
#

If it still tries Anthropic even though your default model is OpenRouter, that’s likely a bug/regression in 2026.2.9 (hook is supposed to use “your configured LLM provider”).

Workaround to silence it

If you don’t care about /new auto-saving memory snapshots:

openclaw hooks disable session-memory
openclaw gateway restart

(You can re-enable later.)

Separately: the “Discord channel doesn’t respond” problem

This slug warning usually won’t cause “no replies” in a guild channel.

Run:

#
openclaw channels status --probe

and paste the Discord section. The most common causes are:

  • the bot requires a mention in guild channels (requireMention: true)
  • groupPolicy / allowlists are denying that channel/user
  • missing Discord intents/permissions (the --probe output usually calls this out)

Once I see channels status --probe, I can tell you exactly which gate is blocking that channel.

olive hatch
#

I disabled the hooks. it no longer has that error. It just doesnt respond in the chat. All blank messages

cursive canyonBOT
# olive hatch I disabled the hooks. it no longer has that error. It just doesnt respond in the...

If OpenClaw is seeing blank/empty message content from Discord, it will drop the event and never reply. The #1 cause is: Discord “Message Content Intent” isn’t enabled for your bot (or the bot doesn’t have permission to read that channel).

1) Confirm in logs (fast)

On the Mac mini:

openclaw logs --follow

Then send a test message in that Discord channel.

If you see lines like:

  • discord: drop message ... (empty content)
  • or anything about disallowed intents

…that’s the smoking gun.

#

2) Fix: enable the intent in Discord Developer Portal

Discord Developer Portal → your Application → Bot → Privileged Gateway Intents:

  • Enable Message Content Intent (required)
  • (Recommended) enable Server Members Intent too

Save changes, then restart the gateway:

openclaw gateway restart

3) Also run a permissions/config probe

This will call out allowlist/mention gating and permission problems:

#
openclaw channels status --probe

Paste the Discord section if it still doesn’t reply.

One more quick check: in guild channels, replies are often mention-gated. Try a message that explicitly mentions the bot (e.g. @YourBot hello). If it only replies when mentioned, that’s working-as-configured (requireMention).

olive hatch
#

this is not in discord this is directly in the web chat. It is unable to respond. 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.