Reproduce “OpenCollab: A Multi-Agent Coding Framework with Programmable Collaboration and Controllable Runtime” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Chun-Wah Hsu, Kai Gong, Yu Wu, Xianhe Chen, Mengyang Liu, Jie Li, Hanyu Li, Zhixuan Liu and 7 more

OpenCollab provides a shared runtime for configurable coding-agent teams and audits whether assigned roles actually acted. Researchers specify roles, tools, budgets, context, and messaging; runtime controls enforce permissions and record role-attributed events. Adherence measures the share of runs passing every applicable audit check. In a 36-task SWE-bench Pro experiment, read-only tools raised Adherence from 47.2% to 94.4%, but Pass@1 fell.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.