pxpipe

pxpipe An ingenious local proxy called pxpipe is cutting AI API costs by up to seventy percent by turning bulky text into images before it leaves your machine. While large language models charge a premium to read massive blocks of text like system prompts, tool documentation, and long chat histories, their price for reading an image is fixed by its pixel dimensions. By pack-rendering dense text into compact PNGs, the model reads the visual data at a fraction of the token cost. It automatically evaluates each request only converting when the math saves you money and leaves your actual output untouched. Text needles read fine on both arms; the imaged phrase-count doesn't read on Opus, and pxpipe says so instead of fabricating a number. That misread rate is why Opus is opt-in. Try it (30 seconds)