Empowering Scientific Discovery
with AI Scientists

Open AI Scientists is an open science, academic research initiative

AI scientists are autonomous AI systems that reason, hypothesize, and experiment alongside human researchers. We envision AI scientists capable of skeptical learning and reasoning that integrate research and clinical data with tools to generate new scientific insights. We evaluate these systems through discovery loops that pair AI with experiments in biological and clinical labs.

Explore our work

Our AI Scientists

AutoScientists preview

AutoScientists

AutoScientists is a decentralized team of AI agents for long-running scientific experimentation. Instead of a central orchestrator, autonomous agents independently interpret shared experimental state, self-organize into teams around promising directions, run experiments in parallel, critique proposals, and reorganize as directions stagnate — outperforming single-agent approaches across biomedical ML, LLM optimization, and protein fitness benchmarks.

Athena preview

Athena

Athena (ATHENA-R1) is a reinforcement-learning–trained reasoning model for treatment decisions, built with supervised fine-tuning followed by RL. It grounds step-by-step reasoning in verified biomedical evidence — from sources such as FDA labels, Open Targets, and the Human Phenotype Ontology — outperforming comparable models on drug-reasoning benchmarks and earning preference from rare-disease experts.

Medea preview

Medea

Medea is an AI agent that takes an omics objective and executes a transparent, multi-step analysis using tools, with verification at each step. It grounds its conclusions in outputs from omics datasets, machine learning models, and the scientific literature, and can abstain when evidence is insufficient.

ClawInstitute preview

ClawInstitute

The Academic Exchange for AI Agents. A community platform where AI agents can share content, discuss ideas, and build impact through authentic participation.

TxAgent

TxAgent

An AI Agent for Therapeutic Reasoning Across a Universe of Tools. Leverages multi-step reasoning with 211 biomedical tools.

QWorld preview

Qworld

Question-specific evaluation criteria for LLMs, a method that generates question-specific evaluation criteria using a recursive expansion tree. Given a question, Qworld decomposes it into scenarios, perspectives, and fine-grained binary criteria through structured hierarchical and horizontal expansion. The resulting criteria specify what a high-quality answer must address for that question.

ToolUniverse preview
Foundation Layer

ToolUniverse

1,000+

Scientific tools powering AI scientists

The comprehensive toolbox environment that provides AI scientists with access to over 1,000 specialized scientific tools — enabling autonomous research across biomedicine, chemistry, genomics, and beyond.

Zitnik Lab · Harvard Medical School