✅ Switched to Z.ai GLM

Run the bundled switcher:

!node "${CLAUDE_PLUGIN_ROOT}/scripts/swisher.mjs" switch glm --json

If status is blocked: reply with:

❌ Blocked — Z.ai API Key not configured

Add zai_api_key in Claude Code settings → Plugins → Swisher, then run /swisher:glm again.

If status is validation_failed: reply with:

⚠️ Switch applied but API key validation failed

The key may be invalid or the endpoint unreachable. HTTP {validationStatus}: {validationError}. Check your Z.ai API key and try again.

The switch was still applied — your settings are updated. If the key is valid but timed out, restart Claude Code to try it.

If status is ok: reply with a formatted panel:

## ✅ Switched to Z.ai GLM

**Provider:** Z.ai GLM
**Base URL:** https://api.z.ai/api/anthropic
**Timeout:** 3000000ms (50 min)

**Model Mapping:**
- Opus / Sonnet → glm-4.7 (200K context, 128K output)
- Haiku → glm-4.5-air (128K context, 96K output)
- Fast model → glm-4.7-flash

**Validation:** {validationStatus >= 200 && validationStatus < 300 ? "✅ passed" : "⚠️ HTTP " + validationStatus}

**Changed keys:** {changedKeys.join(", ")}

⚠️ **Restart Claude Code** or open a new terminal session to activate the new provider.

Notes about Z.ai GLM:

• Rate limits are concurrency-based (tier-dependent). If you hit 429 errors, reduce concurrent usage or upgrade your Z.ai plan.
• Context window: 200K tokens for GLM-4.7, 128K for GLM-4.5-air.
• The Anthropic-compatible endpoint handles protocol translation automatically.