Choose a starting point
Open Agents. Choose Browse library to inspect ready-made agents, or Create agent to describe your own. A library entry can be installed with Get agent; Make it yours lets you describe a change before building your version. The creation page also offers example prompts for onboarding, time off, expenses, and attendance. Use one only if its job matches yours. You can also ask the main agent in Chat to help draft a specialised agent. Say what should happen, who or what starts the work, and which connected tools it may need. The main agent helps prepare the saved agent; it does not replace the agent’s own settings page.Review the saved definition
Open the agent and read Instructions. Check the actions, recipients, and destinations. Then inspect Triggers for the event or schedule, Tools for the actions it can call, and Permissions for what data it can read or change. If it uses an integration, connect that provider and check the provider’s Tool permissions too. Only one agent can run on each trigger. If another agent owns the trigger, Worklayer can ask to archive the holder before this one takes over. Read that confirmation and the older agent’s run history before replacing it.Save and follow the first run
Use Save after reviewing the agent. A saved active agent runs when its trigger fires. You can use Pause agent to stop new runs, Resume agent to start them again, or Archive agent to stop it and free its triggers. Open its run history after the first event to inspect what actually happened.Changing instructions and permissions together can open Review permission changes. If you lack permission to grant the requested access, someone with access-management permission must save those changes.