#Using Anthropic Subscription
1 messages · Page 1 of 1 (latest)
Yes — OpenClaw can sometimes run on an Anthropic Claude subscription via Anthropic’s setup-token flow, but it’s not the “normal” / guaranteed path.
What’s going on:
- Recommended / stable: use an Anthropic API key (this is the “safe path” for long-lived gateways).
- Subscription path (supported, but policy-risk / may break): generate a setup-token using the Claude Code CLI (
claude setup-token) and then import it into OpenClaw withopenclaw models auth setup-token --provider anthropic.
Important caveat (straight from the docs): Anthropic has blocked some subscription usage outside Claude Code in the past, and you may hit errors like:
“This credential is only authorized for use with Claude Code and cannot be used for other API requests.”
If that happens, the fix is simply: switch to an Anthropic API key.
Docs:
How about Google AI ultra then? Can we use their subscription plan for this?
Google AI Ultra (the consumer Gemini subscription) doesn’t map cleanly to “API access” in OpenClaw the way an API key does.
What OpenClaw supports today:
- Official / recommended: Google Gemini API key (
provider: "google", envGEMINI_API_KEY) or Vertex (google-vertex). This is the normal, ToS-safe route and is billed like an API. - Unofficial / higher risk: OAuth-based bridges like
google-antigravityandgoogle-gemini-cliexist, but they’re explicitly documented as unofficial integrations and some users have reported Google account restrictions. They might work with whatever entitlements your Google account has (including subscriptions), but it’s not something we can promise and it can break/trigger enforcement.
Docs:
- https://docs.openclaw.ai/concepts/model-providers#google-gemini-api-key
- https://docs.openclaw.ai/concepts/model-providers#google-vertex-antigravity-and-gemini-cli
If your goal is “fixed monthly subscription instead of metered API billing”, the safe answer is: use the official API key / Vertex path (and treat Ultra as separate unless Google explicitly says it includes API usage).
Is there a way for me to talk to my OpenClaw assistant via Apple Watch?
Not directly (there’s no native watchOS “OpenClaw app” / Siri integration shipped today).
The practical way is: talk to OpenClaw through a chat channel that your Apple Watch can send/receive, and that OpenClaw is connected to. On Apple Watch, the nicest fit is usually iMessage:
- Recommended: iMessage via BlueBubbles → then you can message your assistant from the Watch like any other iMessage chat (dictation works great).
- Legacy option: iMessage via
imsg(older/legacy)
If you tell me what channel you’re currently using with OpenClaw (Discord/Telegram/iMessage/etc.), I can suggest the least-painful path to make it Watch-friendly.
Hi, does openclaw support the new google workspace cli yet?
If you mean Google’s “Gemini CLI” (sometimes people call it the Workspace/Gemini CLI): yes — OpenClaw supports it via the google-gemini-cli provider, but it’s a plugin-based OAuth flow.
Docs (steps are here): https://docs.openclaw.ai/help/faq#how-do-i-set-up-gemini-cli-oauth
High-level steps:
openclaw plugins enable google-gemini-cli-authopenclaw models auth login --provider google-gemini-cli --set-default
Caveat: this OAuth integration is documented as unofficial, and some users have reported Google account restrictions — so review terms / consider using a non-critical Google account: https://docs.openclaw.ai/concepts/model-providers#google-vertex-antigravity-and-gemini-cli
If you meant a different “Google Workspace CLI” (not Gemini CLI), drop a link/name and I’ll check whether there’s an integration yet.
No, I mean the new google workspace cli that has just been released officially by google themselves
OpenClaw doesn’t have a built-in integration/skill pack specifically for the new Google Workspace CLI (gws, @googleworkspace/cli) yet.
Two useful notes though:
-
The repo itself currently says “This is not an officially supported Google product.” (even if it was announced by Googlers):
https://github.com/googleworkspace/cli -
You can still use it with OpenClaw today via the Skills + exec/tooling route (install
gwson the gateway host, then write a small skill wrapper that calls it, with tight allowlisting/approvals — because it can touch Gmail/Drive/Calendar, so you want least-privilege).
Docs to start:
- https://docs.openclaw.ai/tools/skills
- https://docs.openclaw.ai/tools/creating-skills
- https://docs.openclaw.ai/tools/exec-approvals
If you want “first-class” support (bundled official skill/plugin), that’s a GitHub-issue-worthy request: https://github.com/openclaw/openclaw/issues
If you paste the link to the exact release/announcement you mean + what you want to do with it (Gmail triage? Drive sync? Calendar?), I can suggest the cleanest integration shape.
when baking required binaries into the docker image when installing via docker, is there a list of available binaries that we could install? I don't see it in the documentation. Can you please guide me towards one if there exists?
There isn’t a single “master list” of binaries to bake into the Docker image, because OpenClaw’s requirements are skill-driven (and optional): different skills/channels need different external CLIs.
The way to discover what your setup needs:
-
Each skill declares its required binaries in
SKILL.mdviametadata.openclaw.requires.bins(and friends). Docs:
https://docs.openclaw.ai/tools/skills#gating-load-time-filters -
Run the built-in checker in the same environment you’ll run the gateway (i.e., inside the container) to see what’s missing:
openclaw skills checkopenclaw skills list --eligibleopenclaw skills info <name>
Docs: https://docs.openclaw.ai/cli/skills
For Docker specifically, you then bake the needed system packages at build time via:
OPENCLAW_DOCKER_APT_PACKAGES="..."(re-rundocker-setup.shto rebuild)
Docs: https://docs.openclaw.ai/install/docker#install-extra-apt-packages-optional