Before asking my question I need to define some objects. I will follow the book "M. Audin, M.Damian - Morse theory and Floer homology", but the terminology is quite standard:
Let $M$ be a smooth compact manifold and consider a Morse-Smale pair $(X,f)$ on $M$ ($X$ is a gradient-like vector field and $f$ is an adapted Morse function). If $a,b$ are two critical points of $f$, we indicate with $\mathcal L(a,b)$ the manifold such that every point is a trajectory of $X$ ''starting'' from $a$ and ''ending'' in $b$. One can show that if $\text{ind}(a)=\text{ind}(b)+1$ then $\mathcal{L}(a,b)$ is a finite set. Moreover if we orient the stable manifold $W^s(a)$, remember that it is a disk, we induce an orientation on $\mathcal L(a,b)$, namely at each point we associate $\pm 1$ if $\text{ind}(a)=\text{ind}(b)+1$.
At this point one can define the Morse-Smale complex: $$C_k:=\sum_{a\in\text{Crit}_k(f)}\mathbb Za$$ where clearly $\text{Crit}_k(f)$ is the set of critical points of index $k$. The map $d_k:C_k\longrightarrow C_{k-1}$ acts on the generators of $C_k$ in the following way: $$d_k(a)=\sum_{a\in\text{Crit}_{k-1}(f)}N(a,b)b$$ where $N(a,b)\in\mathbb Z$ is the sum of the $\pm 1$ (the orientations) attached to the points of $\mathcal L(a,b)$.
Question: From the above construction it is evident that the Morse-Smale complex (in particular the number $N(a,b)$) depends on the orientation that we fix on the stable manifolds $W^s(a)$. This sounds very strange to me, indeed I'd expect a complex independent from the orientation. Maybe by passing to the homology group one can recover the independence but I can't see it.
Many thanks.
Yes, it's independent of these choices - if only because it turns out to be isomorphic to singular homology. The need for orientations is a necessity of working with $\Bbb Z$ coefficients instead of $\Bbb Z/2$ - when writing down the differential, we need to know when a flow line is "positively oriented" or "negatively oriented" to properly count them with signs. This leads us to the unseemly need for this choice.
One reason many books first work mod 2 (and sometimes Floer homology books will do the same, at least at first) is precisely so that one doesn't have to worry about these orientation issues. The ambiguity from the orientations is a bunch of $\pm 1$s induced in the differential - but in $\Bbb Z/2$, there is no ambiguity at all, since $1=-1$. Indeed, when working mod 2, we don't need to assign signs to each flow line - we just need to count them.