from jenkins docs: When applied at the top-level of the pipeline block no global agent will be allocated for the entire Pipeline run and each stage section will need to contain its own agent section.

Use runtime discovery of cpu core count