Most definitions I've seen [1] [2] are agnostic about what kind of polynomials the function must satisfy. PlanetMath, for example, says
A function of one variable is said to be algebraic if it satisfies a polynomial equation whose coefficients are polynomials in the same variable.
However, some references [3] [4] are more strict and require that the polynomial have integer coefficients. Wolfram Mathworld has
An algebraic function is a function $f(x)$ which satisfies $p(x,f(x))=0$, where $p(x,y)$ is a polynomial in $x$ and $y$ with integer coefficients.
The only difference (I think) is that allowing arbitrary real coefficients enables the use of transcendental constants. Since adding these constants doesn't add much complexity (it's the functions that are the interesting part), I'm inclined to allow them. What's the consensus here? Can an algebraic function contain transcendental constants?