Characterized by their low moduli and high stretchability, soft composites have recently attracted great interest from researchers in related areas. The main objective of the present study is on the fracture property and toughening mechanisms of soft composites. Three types of soft composites will be studied: soft elastic foams, double-network (DN) composites and magnetic DN composites. A theoretical/numerical study is carried out over soft elastic foams. By using the analogy between the cellular structure of foams and the network of rubbery polymers, a scaling law for the fracture energy is proposed for soft elastic foams. A phase-field model for the fracture processes in soft elastic structures is further developed to study the crack propagation in an elastic foam, and results have all achieved good agreement with the scaling law. Simulations have shown that an effective fracture energy one order of magnitude higher than the base material can be reached by using the soft foam structure. To further enhance the fracture toughness, the second part of the thesis presents a combined experimental and theoretical study of a DN soft composite, which consists of stacked layers of fabric mesh and 3M VHB tapes. The composite exhibited a damage evolution process very similar to that in the well-known DN hydrogels. The testing results show that the strength and toughness of the DN composite was highly dependent on the composition, and in certain range, the DN composite exhibited much higher mechanical strength and toughness compared with the base materials. A 1D shear-lag model is developed to illustrate the damage-distribution toughening mechanism of the double network composite. The prediction of the model agrees well with the measured properties of the composite in various compositions. The DN composite may also be regarded as a macroscopic model of the DN gel for understanding its structure-property relation. By combing the properties of high-stretchablity, high-toughness and re-healing, the magnetic DN composite studied in the third part of the thesis consisted of a polymer matrix and permanent magnets. The initially connected permanent magnet chains were used as sacrificial and reconnecting components in the composite. The strength of the composite was limited by the magnetic attractions while the maximum stretch was limit by the stretchability of the matrix. The composite was significantly tougher than either of the constituents. A one-dimensional model is developed to examine the mechanical and the damage distribution and energy dissipation process of DN magnetic composite. To further study the property and fracture toughness of the DN magnetic composite, a quasi-static, two-dimensional phase-field model of fracture are developed. Simulations have shown that an effective fracture energy one order of magnitude higher than the base polymer material.