Troubleshooting Center
Check HTTP status first, then Base URL, model ID, key, and provider style.
1. Quick diagnosis
| Error | Common cause | Fix |
|---|---|---|
| 401 Unauthorized | Wrong key, expired key, missing Bearer | Copy the key again and check headers |
| 403 Forbidden | Account or model permission issue | Confirm the account has access |
| 404 Not Found | Wrong Base URL or endpoint | GPT uses /v1; Claude Code uses root URL |
| model not found | Wrong or unavailable model ID | Use the model matrix IDs |
| 429 Rate limit | Too many requests or insufficient quota | Lower concurrency and check quota |
| 500 / 502 / 503 | Temporary upstream or gateway issue | Retry and provide request details to support |
| timeout | Network, proxy, or oversized input | Reduce input, check proxy/network |
| stream interrupted | Proxy interrupted streaming | Disable proxy buffering or test non-streaming |
2. GPT check order
export OPENAI_API_KEY="your_synterolink_key"
curl -i https://api.synterolink.com/v1/models \
-H "Authorization: Bearer $OPENAI_API_KEY"
If /v1/models fails, fix key, network, or Base URL before editing VS Code settings.
3. Claude Code check order
export ANTHROPIC_BASE_URL="https://api.synterolink.com"
export ANTHROPIC_AUTH_TOKEN="your_synterolink_key"
claude --model claude-sonnet-4-6 -p "Reply exactly: OK"
If it still uses the official Anthropic route, environment variables are usually not active in the current terminal.
4. VS Code / Cursor issues
- Check provider type.
- Check Base URL.
- Check exact model ID.
- Restart the extension window or VS Code.
- Prove the API works with curl first, then debug the plugin.
5. What to send support
Do not send real keys. Provide:
- Tool name: VS Code / Cursor / Claude Code / CLI
- Provider style: OpenAI-compatible or Anthropic-compatible
- Base URL
- Model ID
- HTTP status
- Error snippet
- Approximate time