GitHub - aiming-lab/AutoResearchClaw: Fully autonomous research from idea to paper. Chat an Idea. Get a Paper. Fully Autonomous. 🦞

AutoResearchClaw is an open-source Python framework that autonomously generates conference-ready academic papers from a single research idea. The 23-stage pipeline covers literature discovery (OpenAlex, Semantic Scholar, arXiv), hypothesis generation via multi-agent debate, hardware-aware sandbox experiment execution, self-healing code repair, multi-agent peer review, and LaTeX export targeting NeurIPS/ICML/ICLR templates. Key capabilities include 4-layer citation verification to eliminate hallucinated references, a PIVOT/REFINE decision loop at Stage 15, and optional MetaClaw integration for cross-run learning that improved robustness by 18.3% in controlled experiments. It supports ACP-compatible coding agents (Claude Code, Codex CLI, Gemini CLI) as LLM backends and integrates with OpenClaw for chat-driven invocation.