If a monad or comonad on a topos is sufficiently well behaved, then the category of (co)algebras over the (co)monad is itself an (elementary) topos.
Let be a topos. Then
if a comonad is left exact, then the category of coalgebras is itself an (elementary) topos.
Moreover,
the cofree/forgetful adjunction
is a geometric morphism.
If is furthermore accessible and is a sheaf topos, then also is a sheaf topos.
Even if is merely pullback-preserving, the category of coalgebras is a topos.
Therefore, if a monad has a right adjoint, then the category of algebras is itself an (elementary) topos. (Because the right adjoint of a monad carries a comonad structure, evidently a left exact comonad, and there is a canonical equivalence between the category of algebras over the monad and the category of coalgebras over the comonad.)
If a monad on a topos is pullback-preserving and idempotent, then the category of algebras is a subtopos (hence the category of sheaves for some Lawvere-Tierney topology).
The result for left exact comonads appears for instance as (MacLaneMoerdijk, V 8. theorem 4); the result for monads possessing a right adjoint appears in op. cit. as corollary 7. The statement on pullback-preserving comonads is given in The Elephant, A.4.2.3.
The geometric morphisms of the form from prop. 1 are precisely, up to equivalence, the geometric surjections.
This appears as (MacLaneMoerdijk, VII 4. prop. 4).
This way the geometric surjection/embedding factorization in Topos is constructed. See there for more.
For any geometric morphism, the induced comonad
is evidently left exact, hence is a topos of coalgebras.
The so-called “fundamental theorem of topos theory”, that an overcategory of a topos is a topos, is a corollary of the result that the category of coalgebras of a pullback-preserving comonad on a topos is a topos (the slice being the category of coalgebras of the comonad ).
Section V 8. of