alphaXiv open-sources OpenResearch, turning coding agents into reproducible research agents
aigclink · x · 2026-09-14
alphaXiv has open-sourced OpenResearch, a local-first workspace for research agents that turns Claude Code, Codex, OpenCode, or Cursor into agents able to review literature, form hypotheses, run experiments, and produce research artifacts. The repo hit 2.3k stars quickly.
It targets the reproducibility mess of agent-driven research: dozens of experiments, endless code changes, and no trace of which version produced which result. The approach ports software engineering tools to science:
- Git-native experiment tree + immutable run archives: every result links back to the commit snapshot that produced it, with logs, diffs, and artifacts in one ledger.
- Isolated parallelism: per-direction agent sessions with isolated git worktrees and lineage tracking, so 20 parallel research threads stay untangled.
- Local-first: data lives in local SQLite by default, with jobs dispatchable to Slurm/K8s/HPC.
Install with curl -LsS https://openresearch.sh/install.sh | sh, run orx up for a local dashboard, and connect local models via LM Studio, Ollama, or custom endpoints. macOS 11+ and Windows beta supported.
More from coding & agent
- Can Finished Agents Be Reused Across Projects? A Reddit User's Multi-Agent Experiment — Coolerwookie · 2026-09-14
- Sole-Member LLC Owner Weighs Replacing QuickBooks With an AI Bookkeeping Agent — DefraudedCitizen · 2026-09-14
- Astra tip: shift reasoning levels mid-conversation without invalidating the cache — intellectronica · 2026-09-14
- Selling n8n + AI Voice Agents to Local Businesses: Hosting Setup and Retainer Pricing — Guilty_Assignment987 · 2026-09-14
- BountyScout: open-source hourly GitHub bounty scanner with alerts — gleech · 2026-09-14
- AI agent scans Amazon and eBay to run automated arbitrage listings — gleech · 2026-09-14