OpenForgeRL: Train Harness-native Agents in Any Environment

2026-07-23

AI Summary

This work addresses the lack of open-source infrastructure for end-to-end training of stateful, multi-process AI agents, which currently rely on complex reasoning harnesses. The authors propose the first general-purpose training framework decoupled from specific harnesses and environments: it employs lightweight proxies to translate harness invocations into standard reinforcement learning (RL) data and leverages Kubernetes for remote, containerized rollouts. The framework seamlessly integrates with mainstream RL stacks such as veRL and supports multimodal GUI and tool-use environments. Evaluated on benchmarks including ClawEval, QwenClawBench, and OSWorld-Verified, the approach substantially outperforms existing open-source models—matching or exceeding the performance of closed-source systems several times larger—and provides the first empirical evidence of how harness design and RL training jointly influence agent reliability.

Abstract

Modern AI agents rely on elaborate inference harnesses such as Claude Code, Codex, and OpenClaw to drive multi-turn reasoning, tool use, and access to external systems. While powerful, these complex harnesses also make agents hard to train end-to-end with open infrastructure, whose SFT/RL stacks cannot natively express stateful, multi-process harness inference. To address this, we present OpenForgeRL, an open-source framework for training harness-based agents end-to-end in diverse environments. OpenForgeRL achieves this with a lightweight proxy that serves the harness's model calls while recording them as training data for a standard RL codebase (e.g., veRL), and a Kubernetes orchestrator that runs each rollout in its own remote container, together enabling training on any harness in any environment at scale. By decoupling training and inference, OpenForgeRL allows researchers to easily train, study, and improve agents directly in the real harnesses and environments they are deployed with. We validate our framework across diverse, complex harnesses and environments, spanning tool/claw-based agents and multimodal GUI browser- and computer-use agents. Using only hundreds to a few thousand tasks, OpenForgeClaw reaches 31.7 pass^3 and 55.9 pass@3 on ClawEval and 33.7 on QwenClawBench. OpenForgeGUI reaches 37.7 on OSWorld-Verified, 63.0 on Online-Mind2Web, and 72.3 on WebVoyager. Both outperform open baselines of similar size on nearly all benchmarks, and in the GUI setting match or surpass models several times larger. Beyond benchmarks, we analyze how harness choice (e.g., ZeroClaw, OpenClaw, Codex) and RL shape agent behavior. We find that some harnesses are substantially harder to learn than others, and that RL improves agentic reliability, such as self-verification, tool coverage, and completing multi-step plans, though critical abilities such as error recovery remain weak.

Problem

• harness-based agents
• end-to-end training
• reinforcement learning
• open infrastructure
• stateful inference

Innovation

• harness-native agents
• end-to-end RL training
• Kubernetes orchestration
• proxy-based data collection
• agentic reasoning