A vehicle agent system controls operation of relevant in-vehicle devices as execution of a practical response for acting for a driver. When a command is inputted by the driver or a notification is inputted from an in-vehicle network, an assumed response is selected from driver information stored in a RAM, based on the command or the notification. The selected assumed response including parameters is presented to the driver. When the driver inputs a command for permitting the selected assumed response or the driver inputs no command for a predetermined period, the selected assumed response is determined to be the practice response. When the driver inputs modification, the selected assumed response is modified into the practical response.