The paper introduces the Configuration Space Distance Field (CDF), a novel representation that extends the concept of Signed Distance Fields (SDFs) to robot configuration spaces. CDFs provide a unified framework for manipulation planning and control by encoding the geometric relationships between the robot's joint angles and obstacles in configuration space. Unlike traditional SDFs, which are used in task space, CDFs offer efficient joint angle distance queries and derivatives, allowing for direct computation of zero-level-set configurations through gradient projection. This approach simplifies inverse kinematics problems and enables natural geodesic paths around obstacles. The paper presents an efficient algorithm for computing CDFs, including a fusion strategy to combine multiple CDFs for scene-agnostic applications. A neural CDF variant using multilayer perceptrons (MLPs) is also introduced, providing a compact and continuous representation that integrates well with learning, optimization, and control frameworks. Experiments on a planar robot and a 7-axis Franka robot demonstrate the effectiveness of CDFs in whole-body inverse kinematics and manipulation planning tasks, showing superior performance in terms of accuracy, efficiency, and scalability compared to traditional SDF-based methods.The paper introduces the Configuration Space Distance Field (CDF), a novel representation that extends the concept of Signed Distance Fields (SDFs) to robot configuration spaces. CDFs provide a unified framework for manipulation planning and control by encoding the geometric relationships between the robot's joint angles and obstacles in configuration space. Unlike traditional SDFs, which are used in task space, CDFs offer efficient joint angle distance queries and derivatives, allowing for direct computation of zero-level-set configurations through gradient projection. This approach simplifies inverse kinematics problems and enables natural geodesic paths around obstacles. The paper presents an efficient algorithm for computing CDFs, including a fusion strategy to combine multiple CDFs for scene-agnostic applications. A neural CDF variant using multilayer perceptrons (MLPs) is also introduced, providing a compact and continuous representation that integrates well with learning, optimization, and control frameworks. Experiments on a planar robot and a 7-axis Franka robot demonstrate the effectiveness of CDFs in whole-body inverse kinematics and manipulation planning tasks, showing superior performance in terms of accuracy, efficiency, and scalability compared to traditional SDF-based methods.