I have a recursive definition which is similar to the Fibonacci-numbers:
$n_1 = 1$ (I)
$n_2 = 2$ (II)
$n_k = n_{k-2} + n_{k-1} + 1$ (III)
I am looking for a closed formula for it.
If that $+1$ weren't on the end of (III), everything would be very easily linearizable, because I could apply matrices and solve the problem in its own system of eigenvectors. This is how the closed formula of the Fibonacci-numbers is derived.
But now, with this $+1$, this solution doesn't work any more.
What to do?
You could try to use recurrence annihilators, or equivalently, the shift operator, as follows:
$$F(n)=F(n-2)+F(n-1)+1$$ $$F(n+2)−F(n+1)−F(n)=1$$
Let $\partial$ denote $\frac{d}{dn}$. Than $$e^{h\partial}F(n)=F(n+h)$$ which can be used to express the above as
$$(e^{2\partial}-e^\partial -1)F(n)=1$$ which can be factored $$(e^\partial-\frac{1-\sqrt{5}}{2})(e^\partial-\frac{1+\sqrt{5}}{2})F(n)=1$$ what is left to do, is to notice that $(e^\partial-1)$ annihilates a constant. This is why the first differences of the sequence you computed are the Fibonacci numbers($(e^\partial-1)$ is the first difference). As such $$(e^\partial-1)(e^\partial-\frac{1-\sqrt{5}}{2})(e^\partial-\frac{1+\sqrt{5}}{2})F(n)=0$$ which leads to a general solution
$$F(n)=c_0(\frac{1-\sqrt{5}}{2})^n+c_1(\frac{1+\sqrt{5}}{2})^n+c_2$$
The constants $c_i$ can be computed from initial conditions.
If you want to learn more about such computations, and proofs of the used theorems, you can read this post on Recurrence annihilators and Algebraic encodings. The post also contains a recurrence solver, that will show you these calculations, step by step.