Rewriting logic can be used to prototype systems for automated deduction. In this paper, we illustrate how this approach allows experiments with deduction strategies in a flexible and conceptually satisfying way. This is achieved by exploiting the reflective property of rewriting logic. By specifying a theorem prover in this way one quickly obtains a readable, reliable and reasonably efficient system which can be used both as a platform for tactic experiments and as a basis for an optimized implementation. The approach is illustrated by specifying a calculus for the connection method in rewriting logic which clearly separates rules from tactics.