ClawGym: A Scalable Framework for Building Effective Claw Agents

2026-04-29

AI Summary

This work addresses the lack of a systematic framework for developing Claw-style agents, which has hindered progress in verifiable data synthesis, training, and evaluation. To overcome these bottlenecks, we propose ClawGym—the first end-to-end scalable framework tailored for Claw-style environments—integrating role-driven task synthesis, a hybrid verification mechanism, and sandbox-parallel reinforcement learning. ClawGym supports the full agent development lifecycle, from intent-driven data generation and black-box trajectory-supervised fine-tuning to lightweight reinforcement learning. We release ClawGym-SynData (13.5K high-quality tasks), the ClawGym-Agents model series, and ClawGym-Bench (200 evaluation instances), collectively yielding significant improvements in agent performance and reliability on multi-step file manipulation tasks.

Abstract

Claw-style environments support multi-step workflows over local files, tools, and persistent workspace states. However, scalable development around these environments remains constrained by the absence of a systematic framework, especially one for synthesizing verifiable training data and integrating it with agent training and diagnostic evaluation. To address this challenge, we present ClawGym, a scalable framework that supports the full lifecycle of Claw-style personal agent development. Concretely, we construct ClawGym-SynData, a diverse dataset of 13.5K filtered tasks synthesized from persona-driven intents and skill-grounded operations, paired with realistic mock workspaces and hybrid verification mechanisms. We then train a family of capable Claw-style models, termed ClawGym-Agents, through supervised fine-tuning on black-box rollout trajectories, and further explore reinforcement learning via a lightweight pipeline that parallelizes rollouts across per-task sandboxes.To support reliable evaluation, we further construct ClawGym-Bench, a benchmark of 200 instances calibrated through automated filtering and human-LLM review. Relevant resources will be soon released at https://github.com/ClawGym.

Problem

• Claw-style agents
• scalable framework
• verifiable training data
• agent evaluation
• personal agent development

Innovation

• ClawGym
• synthetic data generation
• agent training framework
• hybrid verification
• scalable evaluation benchmark