Autonomous Science Agents — Researcher Index

A filterable index of AI systems that act as scientific agents: literature, hypothesis, experiment planning, code/simulation execution, lab automation, analysis, and reporting across STEM. Answer fast: what domain? · in → out? · accessible today? · how autonomous? Catalog version 2.0.1 modified 2026-07-14; all records carry explicit evidence-verification dates, and the release audit records remaining limitations without treating a rebuild as re-verification. Access tags are best-effort — verify before relying on any system. Rows tagged Paper-only are leads from papers with no confirmed public code. Click a row for detail.

372systems & benchmarks

219open source

41physical-lab (A5)

85benchmarks

5domains

Domain

Access

Autonomy

Only

⚡ Best starting points — by need

	Biomedical agent you can run	Biomni, PaperQA2, TxAgent
	Literature-grounded QA	PaperQA2, Asta, Elicit
	End-to-end ML research	The AI Scientist, AI Scientist-v2, Agent Laboratory
	Autonomous discovery platform	Kosmos, Google AI Co-Scientist, Robin
	Single-cell / omics analysis	CellVoyager, SpatialAgent, CellAgent
	Chemistry tool-use agent	ChemCrow, CACTUS, ChemToolAgent
	Self-driving chemistry lab	Coscientist, Chemist-X, A-Lab
	Materials / DFT automation	AtomAgents, VASPilot, MatClaw, ChemGraph
	Physics / simulation	OpenFOAMGPT, MechAgents, cmbagent
	Protein / antibody design	ProtAgents, The Virtual Lab, OriGene
	Evaluate an agent (benchmark)	ScienceAgentBench, AstaBench, MLE-bench

					Inputs → Outputs	Links