Common tasks and questions

Start with the task you need to complete. The linked guides contain the configuration details; the checks below help you choose the right approach.

How do I build and run my first Hyperflow?

Create a short flow with a known input and a result you can inspect. Add actions in order, map variables explicitly, and run it in a test environment before adding external writes. Use Build a Hyperflow, actions and variables. Visual actions handle many tasks; custom scripts are an option where the integration needs code.

Which trigger should I choose?

Choose the event that should start work, then verify its payload. A webhook receives an external event; an API trigger exposes a callable operation; a schedule runs at configured times. Email-driven flows also need the mailbox connection and polling configuration. The email setup described in the trigger guide includes Host Address, Connection Method, Port and Interval Type. Filter on the intended sender, subject or content before creating a downstream record. See triggers, webhooks and APIs and the Email library.

Where do application credentials belong?

Create the appropriate application credential in Vault, then select that profile in the action or agent configuration. Do not place secrets in prompts, descriptions or example payloads. The fields depend on the selected authentication method. For example, the documented Xurrent OAuth client-credentials profile uses API OAuth Service URL, API URL, Account Name, Client ID and Client Secret. Verify the endpoint and account before troubleshooting the flow. See application connections, Vault and Xurrent.

How do I map or transform data between systems?

Inspect the actual source payload, select the relevant values and map them to the target action's fields. Data Transform can use a variable or the result of an earlier action. Use its preview to inspect data, then configure the required key, type and transformations. Validate a small sample before processing a batch. A field that looks like a date or number may still be a string. See Data Manager and Data Transform.

How do I reject incomplete or invalid input?

Add explicit conditions before the write action: required values, expected formats and valid relationships between fields. Send invalid input to a clearly defined outcome or correction path. Validation is part of your flow design; it does not happen automatically for every business rule.

What should I inspect when a run fails?

Find the run in Monitoring, open its log and locate the failing action. Record the upstream status and error separately from the overall run status. Check the input mapping, credential, target permissions and target response. Configure retries, fallback actions and notifications deliberately through error handling. Before retrying a write, check whether the first attempt created or changed a record.

How do I avoid a ticket-sync loop?

Store the relationship between the source and target records and recognise events created by the integration. A webhook acknowledgement alone does not prevent duplicate creation. The Jira–Xurrent example explains source routing, origin markers, correlation and duplicate-event tests.

Can I use an AI Agent for support requests?

Yes: configure its model, knowledge, skills and channel, then test the specific operations it can perform. Searching knowledge and creating a request are different capabilities with different permissions. Start with the ServiceNow, Xurrent or ServiceDesk Plus example. A successful chat response is not evidence that a downstream write succeeded; return the system's record identifier after checking the action result.

How should I handle conversational context in an existing My ChatGPT flow?

The existing library documents Thread Scope → Message → Run and a Custom Thread ID. Use a conversation-specific identifier when reusing context; do not share one fixed thread between unrelated users. This is a library workflow pattern, separate from configuring an AI Agent or connecting an AI client to MCP. Check the My ChatGPT reference for the installed library and provider API requirements.

How do I turn an example into a repeatable service?

First establish its input contract, required permissions, result and failure behaviour. Then decide which parts become reusable Work Items and how deployments bind customer-specific values. Follow the Service Automation model and MSP guide. A single successful integration run is a useful building block, not a complete deployed service.