43 projects · 8 categories · End-to-end AI scientists · Experiment loops · Research co-pilots · Skill packs · Paper tools · Benchmarks
개요
43 projects · 8 categories · End-to-end AI scientists · Experiment loops · Research co-pilots · Skill packs · Paper tools · Benchmarks
README
What Is This?
AI can now carry out research autonomously — from generating ideas and running experiments to writing full papers. This repository tracks the open-source ecosystem making that happen.
| Without these tools | With these tools |
|---|---|
| You spend weeks reading literature | AI reads 500 papers in 10 minutes |
| You manually iterate on experiments | AI runs overnight loops: tweak → run → score → keep or revert |
| You write papers from scratch | AI generates camera-ready LaTeX drafts with figures |
| You review your own blind spots | AI self-reviews, cross-checks, and iterates |
43 projects tracked as of 2026-04-03.
Quick Guide
| If you want to… | Start here |
|---|---|
| Fully automate idea → paper | End-to-End AI Scientists |
| Optimize code metrics overnight | Experiment Optimization Loops |
| Get an AI research co-pilot | Research Co-pilots & Interactive Agents |
| Set up a research workspace | Research Workspaces |
| Add research skills to your agent | Skill & Workflow Packs |
| Search papers or make figures | Literature Search, Citation & Paper Tools |
| Go deep in bio/med | Vertical Domain Agents |
| Evaluate AI research quality | Benchmarks & Evaluation |
Contents
- People To Follow
- Project Type Overview
- Category Tables
- Events & Conferences
- X News & Trends
- Related Lists
- Contributing
People To Follow
Project Type Overview
| Category | Count | Representative Repos |
|---|---|---|
| End-to-End AI Scientists | 9 | AI-Scientist, AutoResearchClaw, InternAgent, NeuriCo |
| Experiment Optimization Loops | 5 | autoresearch, pi-autoresearch, codex-autoresearch |
| Research Co-pilots & Interactive Agents | 4 | EvoScientist, MagiClaw, ScienceClaw |
| Research Workspaces | 3 | dr-claw, Research-Claw, DrClaw |
| Skill & Workflow Packs | 8 | claude-scientific-skills, AI-Research-SKILLs, ARIS, uditgoenka/autoresearch |
| Literature Search, Citation & Paper Tools | 6 | paper-search-mcp, PaperBanana, CitationClaw |
| Vertical Domain Agents | 2 | MedgeClaw, BioClaw |
| Benchmarks & Evaluation | 2 | ResearchClawBench, paperreview.ai |
Category Tables
Interface: terminal tool · agent plugin / skill pack · browser · messaging · 
End-to-End AI Scientists
Fully autonomous pipelines that carry the research loop from idea or literature all the way to paper-like output without requiring human steering at each stage.
| Project | Stars | Description | Interface | LLM | Update Recency |
|---|---|---|---|---|---|
| AI-Scientist | 12.5K | First open-source system to autonomously generate ideas, run experiments, and write research papers |      |
last 1 year | |
| AutoResearchClaw | 7.5K | Give it a research idea, get back a complete paper with figures and peer review | |
last 1 week | |
| AI-Scientist-v2 | 2.3K | Next-gen AI Scientist that explores research directions via tree search without needing templates | |
last 1 year | |
| InternAgent | 1.2K | Unified long-horizon scientist framework that links deep research, executable verification, and memory-driven evolution across algorithm and empirical discovery |  |
last 1 week | |
| SibylSystem | 185 | Give it a topic, it runs 19 stages to produce a conference-style paper and self-improves across runs | |
last 1 week | |
| NanoResearch | 160 | 9-stage pipeline that runs real GPU/SLURM experiments, analyzes results, and writes LaTeX papers with grounded evidence | |
last 1 week | |
| NeuriCo | 40 | YAML-in autonomous research framework that reviews literature, runs experiments, writes LaTeX papers, and auto-pushes complete results to GitHub repos |  |
last 1 week | |
| FARS | — | Public Analemma deployment of a multi-agent system that runs ideation, planning, experiments, and short-paper writing at scale |  |
last 1 month | |
| Kosmos | 471 | Autonomous discovery engine that tests hypotheses in sandboxed containers and tracks findings in a knowledge graph |  |
last 3 month |
Experiment Optimization Loops
Tight modify-run-score-keep/discard loops. The goal is repeated metric improvement on a runnable target, not broad research coverage.
| Project | Stars | Description | Interface | LLM | Update Recency |
|---|---|---|---|---|---|
| autoresearch | 48.8K | The original overnight loop: AI tweaks code, runs experiments, keeps what works, reverts what doesn’t | |
last 1 week | |
| pi-autoresearch | 2.7K | Generic version of the autoresearch loop that works on any measurable optimization target | |
last 1 week | |
| codex-autoresearch | 473 | Autoresearch loop built for OpenAI Codex with smart recovery when stuck | |
last 1 week | |
| autoresearch-mlx | 941 | Karpathy’s autoresearch loop adapted for Apple Silicon Macs using MLX | |
last 1 month | |
| autoresearch-claude-code | 182 | Autoresearch loop ported to Claude Code as a drop-in skill | |
last 1 month |
Research Co-pilots & Interactive Agents
Capable research agents that work alongside humans rather than running fully unattended. Includes interactive frameworks, discovery engines, and multi-purpose research assistants.
| Project | Stars | Description | Interface | LLM | Update Recency |
|---|---|---|---|---|---|
| EvoScientist | 1.4K | Interactive research assistant you chat with via terminal, Telegram, Slack, or WeChat |  |
last 1 week | |
| MagiClaw | 81 | Feishu/Lark command center that orchestrates specialized scientific agents and can bootstrap new agents on EvoMaster | |
last 1 week | |
| Amadeus | 49 | Personal research assistant that reads, summarizes, and organizes papers with multi-pass AI analysis and ARIS workflows | |
|
last 1 week |
| ScienceClaw | 301 | Long-running research coworker that generates new skills at runtime and retains context across sessions | |
last 1 month |
Research Workspaces
Persistent research operating surfaces — IDEs, dashboards, and multi-agent team environments where human and agent researchers coordinate work over time.
| Project | Stars | Description | Interface | LLM | Update Recency |
|---|---|---|---|---|---|
| dr-claw | 598 | Browser-based research IDE with chat, file explorer, terminal, and research dashboard in one window | |
last 1 week | |
| Research-Claw | 369 | Academic workspace with built-in literature manager, task board, and arXiv monitoring | |
last 1 week | |
| DrClaw | 127 | 24/7 AI research team you manage from browser, desktop app, or chat |   |
last 1 week |
Skill & Workflow Packs
Reusable skill bundles, workflow templates, agent plugins, and scaffolding that research agents can directly consume. These are not standalone applications — they extend host agents.
| Project | Stars | Description | Interface | LLM | Update Recency |
|---|---|---|---|---|---|
| claude-scientific-skills | 15.8K | 170+ plug-and-play research skills for Claude Code, Cursor, Codex, and other agents | |
last 1 week | |
| AI-Research-SKILLs | 5.4K | Reusable skills covering the full ML research lifecycle from literature survey to paper writing | |
last 1 week | |
| ARIS | 3K | 38 first-level skills that chain into a full overnight workflow: find ideas, run experiments, review, write paper | |
last 1 week | |
| uditgoenka/autoresearch | 1.7K | Multi-purpose Claude Code plugin: optimization loop, debug, fix, security audit, ship, and predict in one skill | |
last 1 week | |
| OpenClaw-Medical-Skills | 1.5K | Biomedical skill pack for OpenClaw agents covering clinical, pharma, and precision medicine | |
last 1 week | |
| LabClaw | 835 | 240 lab-oriented skills spanning biology, pharmacy, medicine, literature, and visualization | |
last 1 week | |
| autoresearch-skill | 429 | A skill that optimizes other skills: mutates prompts and keeps versions that score higher | |
last 1 week | |
| PaperClaw | 196 | Scaffolding tool that generates domain-specific paper-reading agents from templates | |
last 1 month |
Literature Search, Citation & Paper Tools
Paper retrieval, citation analysis, academic visualization, and publishing layers. These tools focus on the literature-facing and output-facing sides of research.
| Project | Stars | Description | Interface | LLM | Update Recency |
|---|---|---|---|---|---|
| PaperBanana | 5.3K | Turns paper content into publication-quality academic illustrations and diagrams | |
last 1 month | |
| paper-search-mcp | 857 | Searches 20+ academic databases (arXiv, PubMed, Semantic Scholar, etc.) at once via MCP | |
 |
last 1 week |
| meta-knowledge-graph | 7 | LLM-powered academic knowledge graph engine that extracts hierarchical concepts from PDFs and turns them into interactive discovery maps |  |
last 1 week | |
| CitationClaw | 200 | Analyzes who cites your papers, why, and generates visual impact dashboards |  |
last 1 week | |
| ClawPhD | 136 | Converts papers into posters, diagrams, websites, and other shareable assets | |
last 1 week | |
| autoresearcher | 423 | Takes a research question and produces an automated literature review from Semantic Scholar | |
over 1 year |
Vertical Domain Agents
Deep single-domain agents that win by specializing in one scientific vertical rather than trying to be universal.
| Project | Stars | Description | Interface | LLM | Update Recency |
|---|---|---|---|---|---|
| BioClaw | 243 | Bioinformatics chatbot with BLAST, SAMtools, FastQC and other bio tools in isolated containers | |
last 1 week | |
| MedgeClaw | 925 | Biomedical research assistant with 140 specialized skills and multi-platform messaging | |
last 1 month |
Benchmarks & Evaluation
Frameworks and services that measure how well AI agents do research — scoring papers, comparing against human baselines, and grading agent outputs.
| Project | Stars | Description | Interface | LLM | Update Recency |
|---|---|---|---|---|---|
| ResearchClawBench | 21 | Benchmark with 40 real-science tasks across 10 domains that evaluates whether AI agents can produce publication-quality research | |
last 1 week | |
| paperreview.ai | — | Agentic paper reviewer by Stanford that scores research papers on novelty, rigor, and clarity | — | last 1 week |
Events & Conferences
Last updated: 2026-03-22
| Type | Event | Date | Why it matters |
|---|---|---|---|
| Conference | CAISc 2026 | submissions open 2026-04-15 |
AI authors + AI reviewers — directly tests the limits of agent-driven research |
| Conference | Claw4S Conference 2026 | deadline 2026-04-05 |
Submit an executable SKILL.md, not a PDF — $50K prizes, up to 364 winners |
| Hackathon | SciMaster 上海西岸科研黑客松 | 2026-03-27 |
SciMaster + BohrClaw + OpenClaw, 48 hours to produce a paper |
| Hackathon | Stanford AI × Scientific Replication | 2026 |
Reproduce key results from papers using AI agents — bridging KIPAC, SLAC, HAI, and CS |
X News & Trends
Related Lists
We refer to and recommend several curated paper lists and repositories:
Contributing
If you have projects, people, or lists worth recommending, feel free to open a PR.
추천 도구
다른 키워드를 입력하거나 필터를 제거해 보세요.
설치
npx skillfish add minhaoxiong/awesome-automated-research