OpenForgeRL is an open-source framework that allows researchers to train AI agents inside their actual deployment harnesses (e.g., Claude Code, Codex, OpenClaw) across diverse environments like tool use, web browsing, and desktop control. It works by using a lightweight proxy to record model calls from the harness as training data for standard RL stacks, and a Kubernetes orchestrator to run each agent rollout in its own remote container. The authors show that agents trained with OpenForgeRL outperform similar-sized open baselines on multiple benchmarks and, for GUI tasks, match or surpass models several times larger, using only hundreds to a few thousand tasks.
New framework lets you train AI agents inside the same harness systems they use at inference
OpenForgeRL decouples training from inference to make existing RL stacks compatible with complex, stateful agent harnesses like OpenClaw and Codex.
Big Tech
Xiao Yu · Baolin Peng · Ruize Xu · Hao Zou · Qianhui Wu · Hao Cheng · +4 more
Columbia University · Dartmouth College · Microsoft Research
Research Digest··2 min read
Thread:Agent Harness Optimization
In one line
OpenForgeRL enables end-to-end reinforcement learning of harness-based AI agents in any environment using a proxy and Kubernetes orchestrator.
Before you read this
Three ideas this paper builds onagent harnessA system that structures, verifies, monitors, or controls how an AI agent interacts with external tools, environments, or execution runs.150 papers →agentic reinforcement learningTraining approach that lets LLM-based agents learn multi-step tool use and planning via reinforcement learning, often with dense or per-tool rewards.102 papers →vision-language modelA model that handles tasks combining visual and textual information, such as editing images from conversation or reasoning across images and text.216 papers →
Read it as
The authors present OpenForgeRL, an open-source framework that trains stateful, harness-based AI agents end-to-end using standard reinforcement learning libraries.
The authors designed a framework that decouples the training and inference harnesses of modern AI agents.
The Daily Edition
The front page in your inbox each weekday morning: the lead, the picks, and why each one earned its place.