To specify a variable's PMF, you need two things: (a) a list of what values it can be, and a corresponding list of probabilities of it being those values.
You already have the PMF for $X$, but let's make a table out of it to make it a bit more concrete.
$$\begin{array}{r | l}
x & p(x) \\ \hline
0 & 1/3 \\
1 & \frac 1 3 \cdot \frac 2 3 = \frac 2 9 \\
2 & \frac 1 3 \cdot (\frac 2 3)^2 = \frac 4 {27} \\
\vdots & \vdots
\end{array}$$
I obtained the probabilities, denote above as $p(x)$, by plugging each value of $x$ into the given PMF.
Once you have these, you now have a complete listing of what $Y$ can be, along with all its probabilities. For instance, if $X = 0$, then $Y = \frac{X}{X+1} = \frac{0}{1} = 0$ as well; this will occur with probability $1/3$. If $X = 1$, then $Y = \frac{1}{2}$, which will occur with probability $2/9$, and so forth. Using this idea, you could make a table that looks just like the above, to get the the PMF for $Y$.
Of course, it may be more desirable to put this into a formula; you may want something compact like the expression you gave for $f(x)$. I claim that the clearest way to see what such a formula should be is to write out the first 3 or 4 lines of a table like the above, though.