What is zloop?

zloop is a speed upgrade for Python's built-in system that handles asynchronous operations, meaning the behind-the-scenes machinery that lets a program juggle many tasks at once without waiting for each to finish before starting the next. It is designed as a drop-in replacement: you change a line or two of code and your existing application starts using it. The catch is that it is explicitly experimental and not yet ready for real-world production use. The interesting part is that the performance-critical pieces of this system are written in Zig, a lower-level programming language, rather than in Python itself. Python is great for quickly building things but can be slow for repetitive background work. By rewriting the core engine in Zig, the project speeds up scheduling, timing, and network communication. It still uses Python for the parts where relying on the standard library makes sense, such as managing secure connections. This tool is aimed at people running Python web servers or applications that handle heavy network traffic. For example, if you run a FastAPI application serving lots of concurrent requests, you could swap in this loop to handle more requests per second. The benchmarks show it beating existing fast alternatives in standard single-loop setups. It also has an optional mode that takes advantage of a newer Linux feature to shine even brighter when running multiple parallel processes, like on servers using free-threaded Python. The project makes a notable tradeoff with that optional Linux mode. In standard single-server setups, the default mode is actually faster. The alternative mode adds some overhead that only pays off when running multiple independent loops in parallel on free-threaded Python. So users need to pick the right mode for their specific workload rather than blindly switching to the faster-sounding option.