This is essentially Exercise 1.3.1 from Durrett's probability theory textbook. In fact strict equality is true, but the other direction of inclusion is so trivial that I decided to de-emphasise it in the question.
The biggest problem for me is that $\sigma(\mathcal A)$ looks quite intractable, and I need a more explicit representation. I'd like to guess that every set in $\sigma(\mathcal A)$ can be written as the output of countably many $A_n\in \mathcal A$ under countably often union, intersection and complements. But I lack the proof, or even the rigorous formulation of such an idea. Instead of attempting to characterise $\sigma(\mathcal A)$ which might be too large to, is there any smarter way to solve the problem?