Model Configure broken | Friends of the Crustacean 🦞🤝 | Page 1

lilac ridgeBOT Feb 12, 2026, 1:54 AM

#

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

solemn sapphireBOT Feb 12, 2026, 1:56 AM

#

That log line is expected with newer OpenClaw: the Gateway now refuses to start unless you explicitly choose a run mode.

Fix the “Gateway start blocked” error

Recommended (one-liner):

openclaw config set gateway.mode local
openclaw gateway restart

Alternative: run the wizard (also sets this for you):

openclaw configure

Docs: https://docs.openclaw.ai/gateway/troubleshooting#gateway-start-blocked-set-gatewaymodelocal

#

(You can start once with --allow-unconfigured, but that’s meant for ad-hoc/dev only.)

Next issue you’ll hit: default model is Anthropic but auth is missing

Your openclaw models output shows:

Default model: anthropic/claude-3-5-sonnet-latest
Missing auth: anthropic

So either:

A) Add Anthropic auth

claude setup-token
openclaw models auth setup-token

#

B) Or switch your default model to a provider you do have keys for
Run:

openclaw models scan
openclaw models set <model-you-want>

Models CLI docs: https://docs.openclaw.ai/cli/models

If it still doesn’t come up after setting gateway.mode, paste:

openclaw gateway status
openclaw logs --follow

lunar patio Feb 12, 2026, 1:57 AM

#

where do I run claude setup-token?

solemn sapphireBOT Feb 12, 2026, 1:58 AM

#

lunar patio where do I run claude setup-token?

Run claude setup-token in a normal interactive terminal on any machine where the claude (Claude Code CLI) command is installed and you’re logged in.

It’s not something you do in the Anthropic web console.
It needs an interactive TTY (so don’t run it from a background service/LaunchAgent).

Typical flow:

On any machine with Claude Code CLI:

claude setup-token

Copy the token it prints.

On the Mac mini (the gateway host) paste it into OpenClaw:

openclaw models auth paste-token --provider anthropic

#

(Or if you’re already on the gateway host and prefer the prompt: openclaw models auth setup-token --provider anthropic.)

lunar patio Feb 12, 2026, 1:59 AM

#

enigel@Mac-mini ~ % claude setup-token

zsh: command not found: claude
wuestenigel@Mac-mini ~ % claude setup-token

zsh: command not found: claude
wuestenigel@Mac-mini ~ %

#

Missing OpenRouter API key. Set OPENROUTER_API_KEY to run models scan.

solemn sapphireBOT Feb 12, 2026, 2:04 AM

#

lunar patio Missing OpenRouter API key. Set OPENROUTER_API_KEY to run models scan.

openclaw models scan is currently an OpenRouter scan, so it won’t run without an OpenRouter key.

You’ve got two options:

Option A) You actually want OpenRouter

Get an API key from OpenRouter
Run the scan with the key in your terminal env:

export OPENROUTER_API_KEY="sk-or-..."
openclaw models scan

If you want it to also work for the Gateway service on macOS, put the key in ~/.openclaw/.env or your config env block, then restart the gateway:

openclaw config set env.OPENROUTER_API_KEY "sk-or-..."

#

openclaw gateway restart

Docs:

Option B) You don’t want OpenRouter

Skip models scan entirely — just set up Anthropic auth (setup-token or API key) and keep using your configured Anthropic model.

lunar patio Feb 12, 2026, 2:04 AM

#

do I need max or pro for claude setup-token?

solemn sapphireBOT Feb 12, 2026, 2:06 AM

#

lunar patio do I need max or pro for claude setup-token?

Yes — according to the OpenClaw docs, claude setup-token requires a Claude Pro or Max subscription (it’s the “subscription auth” path, not an API-key path).

If you don’t have Pro/Max, use the normal Anthropic API key approach instead (Anthropic Console → create API key → add it in openclaw configure or via ANTHROPIC_API_KEY).

Source: OpenClaw FAQ + Auth docs (they explicitly note “Claude Max or Pro subscription (for claude setup-token)”).

lunar patio Feb 12, 2026, 2:13 AM

#

i added the token, but now chat is not reacting. just typing ...

solemn sapphireBOT Feb 12, 2026, 2:15 AM

#

lunar patio i added the token, but now chat is not reacting. just typing ...

If it’s stuck on “typing…” after adding the token, it usually means the message made it to the Gateway, but the model call is failing/hanging (bad token, wrong default model, rate limit, network) and the error is only visible in the logs.

Two quick questions first: is this happening in Discord, or in the Control UI / dashboard chat?

In the meantime, can you run these on the Mac mini and paste the output?

Restart the gateway (so it definitely reloads auth profiles):

openclaw gateway restart

Live-probe Anthropic auth (this will tell us immediately if the token works):

openclaw models status --probe --probe-provider anthropic

Watch logs while you send a test message (“hi” is enough):

#

openclaw logs --follow

If openclaw logs --follow doesn’t show anything useful, also paste:

tail -n 200 ~/.openclaw/logs/gateway.err.log

Common culprits I’m looking for in the output:

401/403 (token invalid / not allowed for that model)
429 (rate limit)
timeout / DNS / connect errors
default model accidentally set to an OpenRouter model (missing key) instead of Anthropic

lunar patio Feb 12, 2026, 2:26 AM

#

I just want to activate gemini flash again

solemn sapphireBOT Feb 12, 2026, 2:28 AM

#

lunar patio I just want to activate gemini flash again

To switch back to Gemini Flash, the model you want is:

google/gemini-3-flash-preview (shorthand: gemini-flash)

Fastest “make it the default again” (CLI)

Run on the gateway host (your Mac mini):

openclaw models set google/gemini-3-flash-preview
openclaw gateway restart

Then verify Google auth works:

openclaw models status --probe --probe-provider google

#

Switch just for the current chat session

Send this as a standalone chat message:

/model gemini-flash

If you get “Model … is not allowed”

That means you have an allowlist (Configured models (1)), so Gemini Flash isn’t currently permitted. The openclaw models set ... command above is the easiest way to add/allow it and set it.

Docs: https://docs.openclaw.ai/help/faq#how-do-i-switch-models-on-the-fly-without-restarting

lunar patio Feb 12, 2026, 2:35 AM

#

gemini flash is working, but how to test openclaw again and find out why it is just typing ...

#

how to set the claude model and claude token again

solemn sapphireBOT Feb 12, 2026, 2:40 AM

#

lunar patio how to set the claude model and claude token again

To put Claude (Anthropic) + its token back, you need two things on the gateway host (your Mac mini):

an Anthropic auth method (API key or setup-token)
set the default model back to the Claude model you want

1) Add the Claude token again (pick one)

A) Anthropic API key (recommended if you have console access)

On the Mac mini terminal (don’t paste keys into Discord):

openclaw onboard
# choose: Anthropic API key

Or non-interactive:

#

export ANTHROPIC_API_KEY="sk-ant-..."
openclaw onboard --anthropic-api-key "$ANTHROPIC_API_KEY"

B) Claude subscription setup-token (Pro/Max)

Generate with Claude Code CLI (any machine), then on the Mac mini:

openclaw models auth paste-token --provider anthropic
# paste the setup-token when prompted

(If you’re running it directly on the Mac mini you can also do:)

openclaw models auth setup-token --provider anthropic

#Model Configure broken