Troubleshooting
Models Troubleshooting
Fix model and provider authentication issues.
Model issues are usually caused by provider mismatch, missing credentials, or active engine confusion.
Checklist
- Confirm the active engine.
- Confirm the provider is connected on that engine.
- Confirm the selected model belongs to that provider.
- Send a short test message.
- Check the error returned by the provider.
Common Problems
| Problem | Cause | Fix |
|---|---|---|
| OpenAI API key does not unlock Codex | Codex uses account auth | Use OpenAI Codex login/auth profile |
| Codex selected but no response | Auth profile missing on active engine | Reconnect Codex while that engine is active |
| Provider works locally but not on VPS | Key saved only locally | Save key on VPS engine |
| Ollama works outside xCloud but not inside | Wrong base URL or engine cannot reach Ollama | Use a URL reachable from the active engine |
| Model list is outdated | Engine config or provider registry stale | Restart gateway or update OpenClaw |
Remote Engines
Remember that credentials live with the active engine.
If xCloud is connected to a VPS, the model request is made by the VPS engine. The VPS needs the credentials, network access, and provider config.