This curated list compiles open-source projects designed to automate various stages of the scientific research lifecycle, from literature review and idea generation to experiment execution and paper writing. It targets researchers and engineers seeking to accelerate their workflows by leveraging AI agents and specialized tools, offering a focused alternative to general-purpose AI agent lists.

How It Works

The collection categorizes projects automating specific research tasks or the entire workflow. It includes end-to-end autonomous systems, deep research and literature synthesis tools, automated experiment and code agents, and reusable skill/plugin collections. Core technologies often involve Large Language Models (LLMs) accessed via APIs (OpenAI, Anthropic, Gemini, etc.), frameworks like LangChain and LangGraph, and specialized components for information retrieval, code execution, and document generation (e.g., LaTeX). Approaches range from multi-agent systems and agentic tree search to Retrieval-Augmented Generation (RAG).

Quick Start & Requirements

This repository is a curated list and does not provide a direct installation or execution command. Users must refer to individual project repositories for setup instructions. Listed projects commonly require Python environments, specific LLM API keys, and may have dependencies on Docker, CUDA, or particular hardware for advanced functionalities. Links to specific project documentation, demos, or quick-start guides are typically found within each project's respective repository.

Highlighted Details

• End-to-End Systems: Features projects like Microsoft's RD-Agent (top MLE-bench agent), AI-Scientist (fully automated open-ended scientific discovery), AutoResearchClaw (idea-to-paper automation), AI-Scientist-v2 (produced first AI-written workshop paper accepted via peer review), and Shanghai AI Lab's InternAgent (unified framework for long-horizon discovery).
• Literature Synthesis: Includes Stanford's STORM (knowledge curation), GPT Researcher (generates cited factual reports), and OpenScholar (published in Nature, retrieval-augmented LM outperforming competitors).
• Code/Experiment Agents: Highlights AutoGPT (early autonomous agent framework), Aider (widely used coding backbone), Princeton's SWE-agent (fixes GitHub issues, pioneered SWE-Bench), and AIDE (achieved 4x Kaggle medals compared to linear agents).
• Skills & Plugins: Lists scientific-agent-skills (133 scientific skills) and AI-Research-SKILLs (86 AI research skills).
• Notable Claims: Projects boast NeurIPS spotlights, ICLR/Nature publications, SOTA performance on benchmarks like SWE-Bench and BrowseComp-Plus, and significant Kaggle success.

Maintenance & Community

The list encourages contributions via pull requests for projects meeting specific criteria (500+ stars or notable publications, directly related to automating scientific research, open-source with active repositories). Projects are sorted by star count. No specific community channels (e.g., Discord, Slack) or notable maintainers/sponsors for the list itself are detailed.

Licensing & Compatibility

The list itself is licensed under CC0 1.0 Universal, dedicating it to the public domain. This permissive license allows for unrestricted use, modification, and distribution, including for commercial purposes, without requiring attribution or imposing copyleft restrictions. Individual projects within the list will have their own licenses.

Limitations & Caveats

This list focuses exclusively on projects automating scientific research and requires a minimum star count or notability, potentially excluding smaller or emerging tools. The maturity, specific requirements (e.g., LLM APIs, hardware), and performance claims vary significantly across the curated projects, necessitating individual evaluation. Star counts were last verified in April 2026, and current project popularity may differ.