Claude -p and Agent SDK billing split — what the June 15 change means

Your bills are about to go up if you're not using my systems like I've been saying for months. But it's not as bad as everyone says it is. Everyone's saying, oh, it's killed programmatic things. It's no one has any idea what they're talking about. On June 15th, Claude Agen SDK and Claude P usage no longer counts in your plans usage limit. That sounds like a problem, but how do you know it is? What does that mean for you? Claude P and the Agen SDK are both programmatic commands you can send to Anthropic to call their API and run Claude. An example might be Claude-P find bug output format Jason. That opens up a chat window basically, says, hey, Claude, look at this code, do your thing, and output a Jason file. Open Claude, open code, all the third party harnesses, use that type of programmatic process to use the max plans to use Claude well. This is where agentic processes kind of start forming here. Now usually you have to pay per token use, but they've been linking it to your pro and max account, meaning you could rip through tokens using these third party harnesses and using this things while all putting it on the max and pro usage. The problem is, open Claude, all these things are terribly token inefficient and use so much compute and inference, it's costing Anthropic too much money. So to solve this, they're saying no to that, but instead of banning open Claude, what they're actually doing is giving that type of usage a separate limit where you have to pay per token or…