#Anthropic Subscription CLI Allowed?

1 messages · Page 1 of 1 (latest)

plain warren
#

I saw this message on the openclaw docs
"Anthropic staff told us OpenClaw-style Claude CLI usage is allowed again, so OpenClaw treats Claude CLI reuse and claude -p usage as sanctioned unless Anthropic publishes a new policy.
For long-lived gateway hosts, Anthropic API keys are still the clearest and most predictable production path.
Anthropic’s current public docs:
Claude Code CLI reference
Claude Agent SDK overview
Using Claude Code with your Pro or Max plan
Using Claude Code with your Team or Enterprise plan"

Does this mean that we can use our pro/max subscriptions in the same way you can use the openai/codex subscription through OAuth and not get your account banned/blocked? I heard that there was a huge crackdown, but I haven't seen anything else about this change.

Where can I get more information on this instead of just "source: trust me bro"

hollow flume
#

I have been using my Pro Sub with Claude-CLI until OpenCLAIM changed their architecture or runtime.

In my case, I have multiple providers: OpenAI, GitHub-Copilot, and Claude. OpenAI's Codex and GitHub-Copilot work fine on pi runtime, but for Claude, it needs "claude-cli" runtime.
Meaning you can't switch providers in between a conversation or use the same agent. Works fine (but a PITA) by creating a dedicated agent that only runs on claude-cli runtime.

So if you only have Claude provider it will work.

Basically install claude code, login

claude auth login claude auth status --text claude setup-token (copy this token)

openclaw models auth login --provider anthropic --method cli --set-default openclaw models auth setup-token --provider anthropic (paste that token here)

In openclaw.json

make sure your claude-cli is on the top of anthropic:default auth order

"auth": { "order": { "anthropic": [ "claude-cli", "anthropic:default" ] } },

All documented here https://docs.openclaw.ai/gateway/cli-backends

fallen tulip
#

Will this not use extra usage?

hollow flume
#

Only if you have that enabled in your claude account. It will use subscription first and then fallback to extra usage.

fallen tulip
#

Have you seen any differences between April 23 - May 4 in updates?

#

Has openclaw performed better or worse in your experience

hollow flume
#

After 4.23, 5.3 has been the most stable for me. It's way faster at restart etc.

fallen tulip
#

Okay I will try it, the main issue I had was claude sub not working properly after the prefix went from claude-cli to anthropic and kept giving me API errors despite connected using CLI

#

I will update and try your method

hollow flume
#

Make sure you have done
openclaw models auth setup-token --provider anthropic
and set your auth order correctly

#

Also if you have a billing error in auth-state.json remove this so Openclaw won't blindly ignore that provider and it will try and use claude-cli

fallen tulip
#

5-4 is just 5.3-1?

#

So far so good with the new updates, Claude-cli is working

hollow flume
#

Do you only use claude?

#

In my case I found gpt 5.5 high works way better then Opus 4.7.
I can have a genuine long running sessions and still have plenty of weekly capacity. With Opus I can maybe as like 10-15 questions every 5 hours.
What's your experience with it like?

fallen tulip
#

Do you have $20 subscriptions for both gpt and opus?

#

I have claude max (200) with gpt plus (20)

#

I hit my limits very quickly on the 20 subscription on gpt