Skip to main content
Once Claude or ChatGPT is connected, you can ask about your environment in plain language. The assistant picks the right tools, combines the results and explains them. Below are requests that work well, with the tools the assistant typically uses. Mention the name or identifier of a route or configuration when you know it. The assistant then finds the right data faster.

Investigate failed tasks

Which tasks failed in the last 24 hours? Group them by route and tell me the most common error.
The assistant lists the failed tasks with list_tasks, reads the logs of a few of them with get_task_logs and summarises what went wrong. Ask a follow-up such as “Show me the data of the first failed order” to see the entity data of a single task with get_task.

Find out why a run failed

The incoming configuration shop-orders didn’t import anything this morning. What happened?
The assistant looks up the latest runs with list_runs, checks their status and task counts and reads the warnings and errors of the failed run with get_run_logs. This is often the quickest way to spot an expired credential, an unreachable system or an unexpected response.

Understand a configuration

Explain step by step what the route shop-orders does, from the incoming configuration to the outgoing one.
The assistant reads the route and the configurations it uses with get_configuration and describes the flow in plain words. This helps when you take over an integration someone else built, or when you write documentation for it. Secrets are always hidden.

Check environment variables

Which environment variables relate to the webshop? Does the HTTP connection of the route shop-orders use them?
The assistant lists matching variables with list_environment_variables and opens the connection with get_configuration to see which placeholders it uses. This helps before you change or remove a variable. The values of encrypted variables stay hidden.

Retry or resolve tasks

Retry all failed tasks of the route shop-orders from today.
This needs change access and the Editor role. The assistant first counts the tasks that would change with change_task_status and asks you to confirm, for example “This retries 37 tasks”. Only after you agree does it retry them. In the same way you can skip tasks that should not run, or resolve tasks you handled by hand.
Retried tasks run again and can send data to the connected systems again. Check the count and the filter before you confirm.

Clear stuck queue jobs

Are there jobs that have been running for more than an hour? Stop them.
The assistant lists the jobs with list_queue_jobs, shows which ones can be stopped or killed and, after your confirmation, uses control_queue_jobs. You can also skip scheduled jobs, for example during maintenance on a connected system.

Get a daily health summary

Give me a short status report of my environment: failed tasks, failed runs and jobs in the queue since yesterday.
The assistant combines several read-only tools into one overview, without changing anything.