Why the same compute sells at two prices, and what to build instead. The docs are next door.
What it actually takes to run an agent on compute your user already pays for: the routing, the permissions, the metering, and why the call site should never know which side it landed on.
Read post →Most AI pricing pages are a guess with a dollar sign in front of it. Here is the arithmetic that decides whether a subscription survives its power users, and the three levers that actually move it.
A Claude Pro plan and an Anthropic API key buy inference from the same lab, on the same GPUs, at prices that are not close to each other. Almost every AI product is built on the expensive one.
Point your OpenAI client at one base URL. Soba prices the plan, routes the run, and bills the user at your rate.
Let your users run your AI features on the Claude or ChatGPT plan they already pay for.