12. Troubleshooting
Use the checks below to diagnose common connection, authentication, configuration, rights, execution, and supervision issues.
12.1. Connection and authentication issues
*Check the HTTPS URL of the MCP Server.
*Check network access from the AI client.
*Check the TLS certificate, proxy, or gateway.
*With OAuth/DCR, check the authentication flow and the token.
*With API key mode, check server-side authorization and the x-api-key header.
For local usage, connection issues may occur when the AI client relies on Node.js and Node.js does not trust the certificate chain. In that case, set the NODE_OPTIONS environment variable to --use-system-ca so that Node.js uses the Windows certificate store.
12.2. Configuration and rights issues
*Check the headers configured in the AI client: x-schema, x-api-key, x-allow-mutation.
*Check the default GraphQL schema in the HAS console.
*Check the connection profile, Hopex rights, and the customized metamodel exposure described in section 5.2 Customized metamodel exposure when a custom schema is used.
*Confirm the exact header syntax in the AI client documentation.
12.3. Execution and performance issues
*Check GraphQL errors returned by the MCP Server.
*For mutations, check the configuration described in section 5.1.2 Mutation authorization.
*For deletions, check the explicit confirmation request.
*In case of latency, reduce the requested volume, explored relationships, or query depth.
12.4. Support information
*MCP Server URL and environment.
*AI client used, version, and connection mode.
*Authentication mode.
*Configured headers.
*GraphQL schema used.
*MCP session / mcpSessionId.
*Tool or resource concerned.
*Structured response returned.
*HAS supervision events.
*Client-side and server-side logs, when available.