Anthropic OAuth Guide

📋 What You Need

🔧 Step 1: Install Claude Code CLI

Claude Code CLI is the tool that generates the setup-token. Install it on any machine (your laptop, not necessarily your server).

macOS / Linux

# Install via npm (requires Node.js 18+)
npm install -g @anthropic-ai/claude-code

# Verify installation
claude --version

If you don't have Node.js

# macOS (Homebrew)
brew install node

# Ubuntu/Debian
curl -fsSL https://deb.nodesource.com/setup_22.x | sudo -E bash -
sudo apt-get install -y nodejs

# Then install Claude Code
npm install -g @anthropic-ai/claude-code

Windows

# Install Node.js from https://nodejs.org (LTS version)
# Then in PowerShell or CMD:
npm install -g @anthropic-ai/claude-code

First-time login

After installing, run Claude Code once to authenticate with your Anthropic account:

claude

# This opens a browser window to log in with your Anthropic account
# After login, Claude Code is authenticated and ready

🔑 Step 2: Generate the Setup-Token

On the machine where you installed Claude Code CLI:

claude setup-token

This outputs a token string. Copy it — you'll paste it into OpenClaw next.

🔌 Step 3: Connect to OpenClaw

On your OpenClaw server, paste the token. Choose one method:

Option A: During onboarding (first-time setup)

openclaw onboard

# When prompted for auth, choose one of:
#   "Anthropic token (paste setup-token)" — paste the token from Step 2
#   "Anthropic OAuth (Claude Code CLI)"  — if Claude Code is installed on this machine

Best path if you're setting up OpenClaw for the first time. The wizard walks you through everything.

Option B: Generate + paste on the same machine

If Claude Code CLI is installed on the same machine as OpenClaw, generate the token and feed it directly:

# Generate the setup-token
claude setup-token

# Then pass it to OpenClaw
openclaw models auth setup-token --provider anthropic

Option C: Paste from a different machine (most common)

You generated the token on your laptop but OpenClaw runs on a remote server:

# On your server, paste the token you copied from Step 2
openclaw models auth paste-token --provider anthropic

⚙️ Step 4: Configure Your Model

In your openclaw.json5, set your default model:

{
  agents: {
    defaults: {
      model: {
        primary: "anthropic/claude-opus-4-6"
      }
    }
  }
}

Available models with your subscription:

✅ Step 5: Verify the Connection

# Check that Anthropic shows up with valid credentials
openclaw models status

# Detailed view with auth profile info
openclaw models status --json

You should see:

• Anthropic provider listed as active
• An auth profile with a valid expiry date
• Your chosen model available

🎛️ Configuration Options

Thinking Mode (Claude 4.6)

Claude 4.6 defaults to adaptive thinking in OpenClaw. Override in config or per-message:

// In openclaw.json5
{
  agents: {
    defaults: {
      models: {
        "anthropic/claude-opus-4-6": {
          params: {
            thinking: "adaptive"  // or "on", "off"
          }
        }
      }
    }
  }
}

// Per-message: /think:on or /think:off

👥 Multiple Accounts

Separate agents (recommended)

For completely isolated credentials (personal vs work):

openclaw agents add work
openclaw agents add personal

# Then run onboarding for each agent separately

Multiple profiles in one agent (advanced)

Switch between profiles per-session:

/model Opus@anthropic:work

🔍 Troubleshooting

# Re-generate the token
claude setup-token

# Paste it into OpenClaw
openclaw models auth paste-token --provider anthropic

# Restart the gateway
openclaw gateway restart

# Re-run onboarding for that agent, or paste a token
openclaw models auth paste-token --provider anthropic

# Generate a fresh token
claude setup-token

# Paste it
openclaw models auth paste-token --provider anthropic

# Restart
openclaw gateway restart

# Check what's happening
openclaw models status --json

# Look for auth.unusableProfiles
# Add another profile or wait for cooldown

🔴 The Honest Risk Assessment

Quick Decision