Despite widespread acceptance of fresh autologous bone marrow (BM) for use in clinical practice, limited information exists to analyze if tendon-to-bone healing could be accelerated with local use of fresh autologous BM. To investigate the effect of fresh autologous BM on tendon-to-bone healing with a novel rat model. Controlled laboratory study. An extra-articular bone tunnel was created and filled with an autologous tendon graft in skeletally mature Sprague-Dawley rats (N = 60). They were then randomly divided into 3 groups: BM group (injection of fresh autologous BM into the tendon-bone interface, n = 20), BM-derived mesenchymal stem cell (BMSC) group (injection of allogenic cultured BMSCs, n = 20), and the control group (tendon-bone interface without injection of BM or BMSCs, n = 20). Biomechanical, histological, and immunohistochemical analyses were performed at 2 and 6 weeks after surgery. The BM group showed a relatively well-organized and dense connective tissue interface with better orientation of collagen fibers as compared with the BMSC group. At 2 weeks, the tendon-bone interface tissue thickness of the BMSC group was 140 ± 25 μm (mean ± SEM), which was significantly greater than the BM group (58 ± 15 μm). The BM group showed fewer M1 macrophages at the tendon-bone interface at 2 and 6 weeks (P < .001). In contrast, there were more M2 macrophages at the interface in the BM group 2 and 6 weeks postoperatively when compared with controls and the BMSC group (P < .001). Biomechanical tests revealed significantly higher stiffness in the BM group versus the control and BMSC groups at 2 and 6 weeks after surgery (P < .05). Load to failure showed similar trends to stiffness. These findings indicate that local delivery of fresh autologous BM enhances tendon-to-bone healing better than the alternative treatments in this study. This effect may be partially due to the observed modulation of inflammatory processes, especially in M2 macrophage polarization. Fresh autologous BM could be a treatment option for this disorder.