Classical rocket guidance is generalized to the case in which the “rocket” is a massive, natural celestial body. The main qualitative difference with the classical case lies in the gravitational interaction between such a body and the ejected propellant. The achieved thrust for example is zero for ejection speeds less than the escape speed from the celestial body. The theory presented is valid for arbitrary ejection speeds of the propellant. The effective thrust for guiding such a body is worked out thrice. First, for propellant ejection at classical speeds, next, for propellant ejection at the speed of light, and last, for propellant ejection at an arbitrary speed. All three cases result in the same thrust expression. The most general last case examines both gravitational interaction effects and mass loss effects due to conversion into energy using the unifying framework of the Schwarzschild solution of general relativity. Gravitational interaction effects dominate when propellant is ejected at nearly the escape speed, while effects of mass loss due to conversion into energy dominate when propellant is ejected at nearly the speed of light. Next, it is shown that to impart a fixed impulsive velocity change to a celestial body by ejecting part of it—to be optimized—as propellant one needs a, nonzero, minimum amount of energy, if the body is massive enough. This constitutes the most important qualitative difference with classical rocket guidance, in which case the corresponding minimum energy requirement, for the same problem, is a trivial zero. The minimum energy turns out to be proportional to the fourth power of the characteristic length of the celestial body. Thus, for example, delivering a given velocity impulse requires 10,000 times more energy for a 20 km asteroid than for a 2 km asteroid. The symmetric case of minimizing the total mass loss of a celestial body to achieve a fixed impulsive velocity change corresponds to propellant ejection at the speed of light. The reverse question of intercepting a celestial body with “propellant,” is also briefly examined. This refers to colliding an external chunk of mass with a celestial body for the purpose of diverting the latter’s path. The results, applied to two archetypal asteroids with diameters 20 and 2 km, and densities the same as the Moon’s, suggest that the aforementioned gravitational interaction is not negligible and that, given enough warning time, the minimum energy requirements for guiding such objects are achievable.