Efficient way of feeding documentation into LLM?

Isn't feeding it urls to documentation pages wasteful? any tool for this which cleanly presents documentation with minimum tokens?