The crawl token starts the moment you get it. The AI worker fetches every chapter, module spec and file straight from the server, so it reads the kit once and works from what it pulled. Nothing sits idle in a queue.
I don't publish a per-hour or per-fetch limit on the token, and I won't guess at one. What I can say plainly: the token fetches the kit. It isn't metered by the number of questions your worker asks afterward.
If your worker burns through it too fast, that usually means it's crawling the whole kit again instead of reading its notes. Do this:
- Have the worker save what it pulls into your project files.
- Point it at those saved files on later passes.
- Direct it by module, not by whole kit.
- Keep the 206-page PDF as the fallback. That one is free and always yours.
The kit is free, the PDF is free, the crawl token is free. What costs money is your time and the AI tool you run.
The cost of waiting is the hours your worker spends re-fetching what it already has. Steer it, don't let it wander.
Mouldi
