claude-max
Skill4 starsChanged 3 months agoMITView on GitHub
Configure and activate the Claude Max subscription proxy so the agent runs on Claude Opus 4 at flat-rate cost instead of per-token Azure billing. Use when the user says "configure Claude Max", "use my subscription", "set up Claude Max", or asks why the agent is using GPT/Azure instead of Claude.
- Installs packages
In this repository38 other files it gives its agents.
- skills/take-ownership/AGENTS.mdAGENTS.md
- bitwarden-cliSkill
- blogpostSkill
- build-in-public-reelSkill
- chrome-devtools-remoteSkill
- chrome-useSkill
- claude-sessionsSkill
- dev-stackSkill
- dev-workflowSkill
- dev-workflow-supervisorSkill
- duckdns-domainSkill
- flutter-android-releaseSkill
- github-copilot-sessionsSkill
- git-worktreeSkill
- google-workspace-cliSkill
- hyperagent-eval-skillSkill
- macro-panelSkill
- mermaid2excalidrawSkill
- naste-slaste-styleSkill
- no-github-backlogSkill
- no-pr-backlogSkill
- opencode-apiSkill
- opencodeSkill
- overflowSkill
- prompt-hermes-aiSkill
- reportSkill
- retrospectiveSkill
- secretsmctlSkill
- solo-founder-meta-agentSkill
- solo-founderSkill
- spaceship-dnsSkill
- take-ownership-meta-agentSkill
- take-ownershipSkill
- telegram-cliSkill
- vercel-lfsSkill
- write-agents-mdSkill
- write-skillSkill
- yc-pitch-deckSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 3 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 38
Wait for the user's reply before continuing.
- Very specific: line 55
If it prints FAIL, tell the user the code may have expired and restart from Phase 1.
- Distinctive: line 73
If the script exits 0, confirm to the user: "Done — your agent is now running on Claude Opus 4 via your Max subscription."
- Distinctive: line 100
If the user wants to go back to the LiteLLM-hosted models:
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Claude Max Subscription Setup
Connects this OpenClaw tenant to the user's Claude Max subscription via the
claude-max-api-proxy (a local OpenAI-compatible server that routes through
the Claude Code CLI). After setup the default agent model switches to
Claude Opus 4 and all agent usage is billed to the flat-rate subscription
rather than per-token.
Prerequisites
- Claude Max or Claude Pro subscription at claude.ai
claude-max-api-proxyandclaudeCLI are pre-installed on this VM (which claude-max-api && which claude— both should resolve)
Setup Flow
The setup has three phases. Work through them in order.
Phase 1 — Start auth (get the URL)
bash ~/.openclaw/skills/claude-max/auth-flow.sh startThis prints a single https://claude.com/cai/oauth/authorize?… URL.
Send that URL to the user and ask them to:
- Open it in their browser (logged in to their Claude account)
- Click Authorize
- Copy the authentication code shown on the next page and reply with it
Wait for the user's reply before continuing.
Phase 2 — Submit the code
Once the user sends back the code, run:
bash ~/.openclaw/skills/claude-max/auth-flow.sh submit <CODE_FROM_USER>Expected output (success):
OK logged_in=true email=<user@example.com> subscription=maxIf it prints FAIL, tell the user the code may have expired and restart from
Phase 1.
Phase 3 — Enable the proxy and switch the model
bash ~/.openclaw/skills/claude-max/auth-flow.sh enable claude-opus-4This:
- Starts the
claude-max-apisystemd service (proxy onlocalhost:3456) - Health-checks the proxy (
/health→{"status":"ok"}) - Updates
models.providers.openaito point atlocalhost:3456 - Sets
agents.defaults.model.primarytoopenai/claude-opus-4 - Validates the config and restarts the gateway atomically
If the script exits 0, confirm to the user: "Done — your agent is now running on Claude Opus 4 via your Max subscription."
Check status at any time
bash ~/.openclaw/skills/claude-max/auth-flow.sh statusChoosing a different model
Pass a model ID to enable:
| Model ID | Notes |
claude-opus-4 | Most capable, default choice |
claude-sonnet-4 | Faster, good for most tasks |
claude-haiku-4 | Fastest, lightweight tasks |
bash ~/.openclaw/skills/claude-max/auth-flow.sh enable claude-sonnet-4Reverting to the hosted model catalog
If the user wants to go back to the LiteLLM-hosted models:
openclaw config set agents.defaults.model.primary "litellm/gpt-5.3-codex"
openclaw config set models.providers.openai.baseUrl "${LITELLM_BASE_URL}"
openclaw config validate
bash ~/.openclaw/skills/openclaw-config/safe-restart.sh --reason "revert to LiteLLM"
sudo systemctl stop claude-max-apiTroubleshooting
| Symptom | Fix |
which claude fails | npm install -g @anthropic-ai/claude-code |
Proxy starts but authentication: FAIL in logs | Re-run Phase 1+2 |
curl localhost:3456/health times out | sudo systemctl status claude-max-api — check logs |
| Gateway unhealthy after enable | auth-flow.sh auto-rolls back config; check journalctl -u openclaw-gateway -n 50 |
| Code expired or wrong challenge | Always use the code from the most recent start run |
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

