<p>Inspired by information sensing, storage, and processing capabilities of natural systems, mechanical computing built upon intelligent matters is emerging toward directly perceiving environmental changes and making autonomous decisions. However, challenges remain in achieving general-purpose mechanical computing architecture due to the trade-off between programmability and scalability. Here, we present a mechanical programmable gate array, a scalable architecture integrating dynamic activation mechanisms for general-purpose computing. By mapping local factored Quine-McCluskey logic functions onto bistable origami switch-based logic units and embedding conductive networks, we create low-redundancy, high-stability logic modules. A robotic activation mechanism, guided by magnetic instructions, dynamically configures the logic array for all input combinations, ensuring complete programmability and scalability. The architecture also interfaces with storage units for iterative processes like function reuse and neural network weight updates. With programmability, scalability, and adaptability, this approach lays the foundation for decision-making materials with applications spanning distributed edge computing and embodied intelligent robotics.</p>