Harnessing skeletal muscle for circulatory support would improve on current blood pump technologies by eliminating infection-prone drivelines and expensive transcutaneous transmission systems. Here we describe an implantable muscle energy converter (MEC) designed to transmit the contractile energy of the latissimus doris muscle in hydraulic form. The MEC weighs just 290 grams and comprises a metallic bellows actuated by a rotary arm fixed to the humeral insertion of the muscle via a looped artificial tendon. The housing is anchored to the ribcage using a perforated mounting ring (83 mm diameter). Lessons learned through six design iterations have produced a pump with excellent durability, energy transfer efficiency, anatomic fit, and tissue interface characteristics. This report describes recent improvements in MEC design and summarizes results from in silico, in vitro, and in vivo testing. The components most subject to wear in this device are the stainless-steel bellows, spring-loaded lip seals, and load-bearing surfaces (bearings, cams and shafts). Roller bearings supporting the camshaft and cam follower were replaced with needle bearings for better stress distribution and longer cycle life. Camshaft bearings were improved still further by changing to a full-complement configuration to lower stress concentration and reduce lateral (off-axis) shaft movement that could reduce lipseal life. Bellows cycle life was estimated using ANSYS V11 finite element analysis (FEA) software with a mesh size of 0.002”. In this simulation a pressure of 22 psi was applied to the internal surface of the bellows and compression length was set to the longest possible stroke (0.177”). All load-bearing surfaces were analyzed for fatigue stress and cycle life under these same loading conditions following closed form equations. Results show that the overall durability of the MEC device can be expected to exceed 450 million cycles, resulting in a minimum working life of 14.5 years given a 1 Hz cycle rate. Lipseal durability was tested empirically in a saline bath using a cycling apparatus designed specifically for that purpose. After 55 days (12.3 million cycles) the test was stopped and the unit disassembled and inspected. The shaft and seals showed evidence of contamination buildup in front of the lip seal but not behind it, indicating that the seal had functioned properly throughout the test period. Importantly, implant studies in 30–35 Kg dogs confirm excellent anatomic fit, patient comfort, and device functionality to one month. These results suggest that muscle-powered cardiac assist devices are feasible and that efforts to further develop this technology are warranted.