As large‑language‑model (LLM) based agents advance, their evaluation becomes increasingly multi‑dimensional: a capable agent must excel not only in task‑completion accuracy but also in interaction quality, safety, and efficiency. This raises a central question—can these objectives be optimized simultaneously? Existing approaches often collapse heterogeneous metrics into a fixed scalar score. Scalarization depends on metric normalization and preference weights, which can discard candidates that represent useful deployment trade‑offs.\ \ We introduce Multi‑Objective Agent Evolution (MOAE), which frames iterative in‑context refinement as a Pareto‑preserving evolutionary search over complete agent rollouts. Given a limited rollout budget, MOAE maintains an empirical archive of non‑dominated candidates, uses objective‑specific diagnostics to guide offspring generation, and applies constraint‑aware selection only at deployment. This separates candidate preservation during search from the preference used to return the final solution.\ \ MOAE requires no parameter updates, and each objective can be replaced by any measurable property. In this work we instantiate them as task performance, trajectory quality, and safety. Experiments on the TravelPlanner and AgentDojo benchmarks show that, under matched rollout budgets, MOAE consistently improves task performance and trajectory quality while maintaining strong safety. A search‑behavior analysis further reveals that Pareto preservation expands the attainable objective region and increases the frequency of joint improvement.\ \ These results demonstrate the potential of Pareto‑preserving in‑context evolution for optimizing multiple agent properties without committing to a fixed scalarization during search.\ \ Review: MOAE demonstrates that multi‑objective agent optimization can be achieved through Pareto preservation, avoiding the need for predefined scalar weights and yielding synergistic gains across performance, quality, and safety.