ChatGPT-5 Fails and Falls Back to Claude

[dense quest]

I’m running Clawdbot locally on Windows 11 and trying to use ChatGPT-5, but it fails to initialize and automatically falls back to Claude (Anthropic). Since I don’t have an Anthropic API key configured, all tasks fail with FailoverError: No API key found for provider "anthropic".

[grand umbra]

Based on the FAQ, Andy's issue is a model failover problem. When ChatGPT-5 fails, Clawdbot falls back to Claude (Anthropic), but since Andy doesn't have Anthropic configured, everything fails.

Here's the fix for Andy:

Option 1: Add an Anthropic API Key

Andy needs to add Anthropic credentials. The easiest way is to set an env var or add to ~/.clawdbot/.env:

ANTHROPIC_API_KEY=sk-ant-api03-...

Or configure it via the onboarding wizard:

clawdbot onboard

Option 2: Disable Failover to Anthropic

If Andy only wants to use OpenAI models, they should configure Clawdbot to not fall back to Anthropic:

{
  agents: {
    defaults: {
      model: {
        primary: "openai/gpt-5.2",
        fallbacks: []  // Disable fallbacks entirely
      }
    }
  }
}

Option 3: Check Current Auth Status

Andy can run this to diagnose: