Summary

This repository serves as a comprehensive survey and framework for understanding "Agentic Science" – the paradigm shift where AI acts as an autonomous scientific researcher. It targets researchers and engineers aiming to leverage AI for accelerating discovery in natural sciences, providing a structured overview of capabilities, workflows, and recent advancements.

How It Works

The project synthesizes research by defining a framework for "Agentic Science," detailing five core capabilities of scientific agents (Reasoning/Planning, Tool Use, Memory, Multi-Agent Collaboration, Optimization/Evolution) and a four-stage dynamic workflow (Observation/Hypothesis, Experimentation, Interpretation, Synthesis). It provides a domain-oriented review across Life Sciences, Chemistry, Materials Science, and Physics, cataloging key advancements and discoveries.

Quick Start & Requirements

This repository is a curated list of research papers and does not provide direct installation or execution instructions.

• Project Homepage: https://agenticscience.github.io/
• Related Paper: https://arxiv.org/abs/2508.14111

Highlighted Details

• Comprehensive survey of agentic systems across Life Sciences, Chemistry, Materials Science, and Physics.
• Defines core capabilities and workflows for autonomous scientific agents.
• Extensive catalog of research papers, serving as a knowledge base for the field.

Maintenance & Community

• Contributions are welcomed via pull requests or direct contact with lead authors at {weijiaqi, yangyuejin}@pjlab.org.cn and xzhang23@ualberta.ca.
• No specific community channels (e.g., Discord, Slack) or roadmap links are provided in the README.

Licensing & Compatibility

• License: MIT. This license is permissive and generally compatible with commercial use and closed-source integration.

Limitations & Caveats

As a curated survey and paper list, this repository does not offer executable code or a direct implementation. Its value lies in synthesizing existing research rather than providing a deployable tool. The field itself is described as lacking a unified framework, though this repository aims to provide one for understanding.