The secant method is a root-finding algorithm that uses successive secant lines to converge on a root of an equation. It begins with two initial points and finds where the secant line between those points intersects the x-axis. It then uses the intersection point as the next estimate and draws a new secant line. This process repeats until the estimate converges within a specified tolerance of the root. The secant method requires only function evaluations, unlike other methods that also require derivative evaluations. However, it may not always converge and provides no error bounds for the estimates.