The disturbance observer (DOB) technique is widely adopted to reject unknown disturbances for accurate and robust control performance with numerous variants. However, studies on the versatile behaviors of DOBs themselves in terms of various constraints have been minimally conducted. To address this, this letter presents a new framework called constrained disturbance observer (CDOB), a strategy that imposes constraints on a DOB through the use of quadratic programming. To accomplish this, we first introduce a variant of DOB that incorporates the simulation of the nominal system. Subsequently, we present a formulation that facilitates the simulation of the nominal system while complying with the user-defined constraints. Consequently, the CDOB behaves identically to the conventional DOB when the constraints are not active, while restricting disturbance rejection when the constraints become active. Numerical simulations are conducted to validate the CDOB under various scenarios.