HarnessX: A Composable, Adaptive, and Evolvable Agent Harness Foundry

2026-06-12

AI Summary

Current AI agent runtime frameworks are predominantly handcrafted and statically fixed, limiting their ability to adaptively optimize in response to varying tasks or models, and failing to effectively leverage execution trajectories. This work proposes a composable, adaptive, and evolvable agent runtime generation platform that constructs a modular framework using typed primitives and substitutable abstractions. It introduces AEGIS, a multi-agent evolutionary engine that continuously refines the runtime through execution trajectories, establishing a closed-loop feedback mechanism that jointly enhances both the runtime framework and model training. Evaluated across five benchmarks—including ALFWorld and GAIA—the approach achieves an average performance gain of 14.5%, with improvements as high as 44.0%, particularly benefiting tasks where baseline performance is weakest.

Abstract

AI agent performance depends critically on the runtime harness, comprising the prompts, tools, memory, and control flow that mediate how a model observes, reasons, and acts. Yet today's harnesses remain largely hand-crafted and static: each new model or task still demands bespoke scaffolding, and the rich traces produced during execution are rarely distilled back into systematic improvement. We introduce HarnessX, a foundry for composable, adaptive, and evolvable agent harnesses. HarnessX assembles typed harness primitives via a substitution algebra, adapts them through AEGIS, a trace-driven multi-agent evolution engine grounded in an operational mirror between symbolic adaptation and reinforcement learning, and closes the harness-model loop by turning trajectories into both harness updates and model training signal. Across five benchmarks (ALFWorld, GAIA, WebShop, tau^3-Bench, and SWE-bench Verified), HarnessX yields an average gain of +14.5% (up to +44.0%), with gains largest where baselines are lowest. These results suggest that agent progress need not come from model scaling alone: composing and evolving runtime interfaces from execution feedback is an actionable and complementary lever. The complete codebase will be open-sourced in a future release.

Problem

• agent harness
• composable
• adaptive
• evolvable
• execution feedback

Innovation

• composable harness
• adaptive evolution
• trace-driven learning
• multi-agent optimization
• runtime interface co-evolution