Cache awareness — why token count is not the bill · WCO.ai
Provider prompt caching discounts repeated prefixes by roughly an order of magnitude, so a transformation that rewrites a cached prefix can cut your token count and raise your cost. How WCO detects a stable prefix, prices read and write multipliers, and when it refuses to optimize.
The interactive playground and tools require JavaScript because they run the optimization engine locally in your browser rather than sending your data to a server.