This paper presents a method for learning whole-body force control for legged manipulators without requiring force sensing. The approach enables compliant interaction and whole-body force application, allowing for intuitive teleoperation by directly commanding the manipulator. The robot's body automatically adjusts to achieve the desired position and force. The method uses a policy trained in simulation and transferred to reality, relying on proprioceptive sensing to estimate and modulate forces. The policy is trained to track both position and force commands, enabling impedance control and gravity compensation. The system allows for a wide range of loco-manipulation tasks, including kinesthetic teaching and safe human-robot interaction. The paper demonstrates the effectiveness of the method through experiments on a quadruped robot with an arm, showing accurate force tracking and estimation. The results indicate that the learned policy can effectively control forces and positions, enabling the robot to perform tasks such as lifting and pulling objects, and maintaining compliance during manipulation. The method is particularly useful for scenarios requiring precise force control, such as kinesthetic demonstration and collaborative tasks. The paper also discusses the challenges of teleoperating quadruped manipulators and proposes a solution based on a policy that takes teleoperation commands as input and outputs joint position commands. The system is evaluated in both simulation and real-world scenarios, demonstrating its effectiveness in various tasks. The paper concludes that the proposed method provides a new approach for learning whole-body force control in legged manipulators, paving the way for more versatile and adaptable legged robots.This paper presents a method for learning whole-body force control for legged manipulators without requiring force sensing. The approach enables compliant interaction and whole-body force application, allowing for intuitive teleoperation by directly commanding the manipulator. The robot's body automatically adjusts to achieve the desired position and force. The method uses a policy trained in simulation and transferred to reality, relying on proprioceptive sensing to estimate and modulate forces. The policy is trained to track both position and force commands, enabling impedance control and gravity compensation. The system allows for a wide range of loco-manipulation tasks, including kinesthetic teaching and safe human-robot interaction. The paper demonstrates the effectiveness of the method through experiments on a quadruped robot with an arm, showing accurate force tracking and estimation. The results indicate that the learned policy can effectively control forces and positions, enabling the robot to perform tasks such as lifting and pulling objects, and maintaining compliance during manipulation. The method is particularly useful for scenarios requiring precise force control, such as kinesthetic demonstration and collaborative tasks. The paper also discusses the challenges of teleoperating quadruped manipulators and proposes a solution based on a policy that takes teleoperation commands as input and outputs joint position commands. The system is evaluated in both simulation and real-world scenarios, demonstrating its effectiveness in various tasks. The paper concludes that the proposed method provides a new approach for learning whole-body force control in legged manipulators, paving the way for more versatile and adaptable legged robots.