Troubleshooting
Common issues and resolutions for AgentCube deployments.
Connectivity Issues
| Symptom | Cause | Resolution |
Health endpoint returns unhealthy | Connector cannot reach Oracle | Verify ESSBASE_SERVER_URL or PLANNING_URL is correct and accessible from the container's network |
| Health endpoint not responding | Container not running or not exposed | Check container status and ingress configuration |
test_datasource returns connection error | Oracle credentials invalid or expired | Verify username/password; check for account lockout |
Authentication Issues
| Symptom | Cause | Resolution |
| SSO popup doesn't appear | OIDC_BASE_URL mismatch | Ensure OIDC_BASE_URL exactly matches the connector's public URL |
| SSO login fails | Identity provider misconfiguration | Verify app registration, redirect URI, and client credentials |
| Oracle token exchange fails | JWT assertion rejected by OCI | Verify signing key, certificate alias (AGENTCUBE_SIGNING_KID), and OCI confidential app configuration |
| User gets "access denied" from Oracle | User not in OCI Identity Domain, or username mismatch | Ensure user exists in OCI Identity Domain with username matching corporate email |
| Account locked in OCI | Repeated failed authentication attempts | Reset the user's password in OCI Identity Domain |
| Symptom | Cause | Resolution |
| Tools don't appear in Claude.ai | Connector added mid-conversation | Start a new conversation |
| Stale tool definitions in Copilot Studio | Copilot caches MCP tool definitions | Disconnect and reconnect the MCP connector in agent Actions settings |
| Connection errors from Copilot Studio | Compatibility issue | Verify the connector is accessible from Microsoft's network |
Container Issues
| Symptom | Cause | Resolution |
| Container restarts frequently | Resource limits too low or application error | Check container logs; ensure at least 0.25 vCPU and 0.5 GiB memory |
| Image update doesn't take effect | Azure Container Apps caching | Use a unique --revision-suffix when deploying |
| Environment variable changes not applied | Container not restarted | Restart the container after changing environment variables |
| Users must re-authenticate after restart | Expected behavior | OAuth state is stored in memory; users re-authenticate on container restart |
Log Access
Azure Container Apps
az containerapp logs show \
--name {app_name} \
--resource-group {resource_group} \
--tail 50
Increasing Log Verbosity
Set AGENTCUBE_LOG_LEVEL=DEBUG for detailed logging. Reset to INFO after debugging to reduce log volume.
Getting Help
If you cannot resolve an issue using this guide, contact Caprus AI support with:
- The health endpoint response (
/health) - Container logs (with
AGENTCUBE_LOG_LEVEL=DEBUG) - The specific error message or behavior observed
- Your connector version and authentication mode