I’ve noticed today that when building a product with Claude Code (understand that this is a competitor). It gets a 403 Unauthorized error when attempting to fetch information from the OpenAI documentation pages.
I can totally understand OpenAI blocking competitors from calling the API and training on responses but I think this is not in OpenAI’s interest when we are using competing coding tools to help us build a product that uses OpenAI’s API.
Trying to build a product that will use OpenAI’s api because it’s best for my use case and this will make OpenAI money.
I can get the page with a http/1.1 script call - to see nothing but js needing descrambling to kick off loading 5MB+ of nonsense .
So:
blocked by origin or request method?
useless unless a page renderer.
The actual OpenAPI yaml documentation full of a maze of refs is 2MB, if you want to send a link into the AI and blast the context full. However, you’ll have to re-host it and serve it as text:
I was thinking of making a chunker that would pull out a tree of semantic items with the path, so they are searchable. There’s over 100 schemas alone there to mix and match, beyond the endpoints…
