Helpers: AI specialists in your conversations
Helpers, also called subagents, handle specialized parts of a task inside your Hatz conversation. You describe the outcome; Hatz can delegate research, file reading, calculations or document work and bring the results back into the answer. Helpers are optimized for their tasks, use fewer credits and do a better job on those tasks.
How helpers work
Start or continue a chat and describe the result you need.
Hatz identifies focused work that an available helper can handle.
The helper works with instructions and tools suited to that task.
Its result becomes part of the conversation for you to review and use.
Examples
Searching the web: “Find the current published pricing for these services and include source links.”
Researching in depth: “Compare these options against our requirements and explain the tradeoffs.”
Reading your files: “Summarize this project brief and list its deadlines.”
Pulling out the details: “Extract the invoice numbers, dates and totals into a table.”
Checking your connected apps: “Find the project’s open items in the connected app and summarize the next steps.”
Documents, presentations and data: “Prepare a project update from this brief and check the totals against the attached spreadsheet.”
Access and controls
A helper is different from an Agent saved in Workshop. Helpers operate within the parent task’s permitted tools and resources. Your connections, resource boundaries and any required approvals still apply. A helper label describes the work, not the underlying model.
Describe the outcome and let Hatz divide the work among the helpers available for the task.
Review results and usage
Check the returned sources, files and calculations. Helpers make specialized work more efficient; the overall task’s usage still depends on its scope and steps. Review the same workspace, task and time range in Credits & Usage.
Troubleshooting
If a helper cannot complete its part, read its result or error and check the relevant input, connection or resource permission. A task may have completed other parts already. For an uncertain external action, establish its outcome before repeating it. Give Support the visible chat link, time and exact error when you need help.
Watch: follow helpers and check their work

