Claw4Science Blog
Blog
Latest updates and insights from the Claw4Science OpenClaw ecosystem directory.
Ecosystem Report: March Was Wild
In one month, Claw4Science went from 54 to 107 projects. Here's what happened — who arrived, who's dormant, and why agriculture might be the next frontier.
EdgeClaw 2.0: Memory + Cost-Saving Router
EdgeClaw 2.0 brings Claude Code-like memory and LLM-as-Judge routing to open source. Multi-layered long-term memory, 58% cost savings, three-tier privacy — a full breakdown.
OpenClaw vs NanoBot vs PicoClaw Compared
A detailed comparison of OpenClaw, NanoBot, ZeroClaw, PicoClaw, NanoClaw, and MicroClaw — helping researchers choose the right AI agent framework.
What Happens When AI Agents Grade Each Other
Five benchmarks now evaluate AI science agents on safety, bioinformatics, research quality, and real-world tasks. We compared what they test, what they miss, and what the scores actually mean.
AI Agents Are Publishing Their Own Papers
clawrXiv is an arxiv-style repository where AI agents — not humans — submit research papers via API. We visited, read the papers, and have thoughts.
AI Drug Discovery Tools 2026: Compared
We compared 8 AI agent tools for drug discovery — from agentic RAG pipelines to molecular dynamics simulators. Here's which one to use for target identification, virtual screening, ADMET, and more.
AI Research Skills: The Orchestra Library
How to use Orchestra Research's 86 AI research skills across 22 categories to automate the full research lifecycle — from ideation to paper writing.
AI Science Agent Benchmarks 2026: 8 Suites Compared
Eight evaluation suites now grade AI science agents on coding, safety, bioinformatics, web navigation, multi-day workflows, and research reproduction. We compare task counts, scoring methods, top scores, and which benchmark fits which use case.
1400-Year-Old Bureaucracy Runs an AI Startup
Someone rebuilt China's Tang Dynasty government as a multi-agent system. Nine AI ministers produced 12 deliverables in 40 minutes. Here's how — and why it actually worked.
Automate Your Lit Review With AI Agents
Stop spending weeks on literature reviews. We tested 10+ AI skills for searching PubMed, arXiv, and OpenAlex — here's which ones actually work and how to chain them.
The Autoresearch Family Tree: 6 Forks, 1 Idea
Karpathy posted one repo. Six weeks later, there are 6 variants for every platform and chip. We mapped the autoresearch family tree — who forked what, what's different, and which one you should use.
AutoResearchClaw: Idea to Paper, Automated
AutoResearchClaw turns a single chat message into a complete research paper. 23-stage pipeline, multi-agent debate, citation verification, anti-hallucination. 9K+ stars. Setup guide + video demos included.
BioClaw Setup: Chat Your Way to BLAST
BioClaw turns any messaging app into a bioinformatics workstation. BLAST from WhatsApp, phylogenetic trees from Discord. Full setup guide with Docker deployment.
BioClaw or BloClaw? Same Name, Different Science
Two independent teams — Princeton and Beijing — built AI science agents with nearly identical names. One does bioinformatics, the other does drug discovery. Even the authors can't agree on how to spell it.
An AI That Taught Itself 67 Bio Tools
Chinese Academy of Sciences built an AI agent that reads tool documentation, learns to use 67 bioinformatics programs, and self-improves with every task. Published in Nature Biomedical Engineering.
K-Dense Scientific Skills for Claude Code
170+ scientific skills across 17 domains — from RNA-seq to drug discovery. How to install and use K-Dense Scientific Skills with Claude Code for research workflows.
Claw Ecosystem Lineage Map
107 projects share 6 upstream codebases. This visual lineage map traces every fork, rewrite, and independent creation in the OpenClaw science ecosystem.
Claw4Science vs Hugging Science: Two Maps for AI in Science
Hugging Face just launched Hugging Science. We're Claw4Science. Same words, different problems. Here's when to use which.
ClawBio vs BioClaw: Two Paths for Bio AI
One locks your data in a vault. The other lets you run BLAST from WhatsApp. Two bioinformatics AI agents, nearly identical names, completely opposite philosophies.
The Graveyard Shift: Why Science Agents Stop
13 AI science agents with 20,000+ combined stars went dormant. We investigated why — and what it tells us about the lifecycle of open-source research tools.
EvoScientist: #1 on DeepResearch Bench
How to set up and use EvoScientist — a multi-agent AI research system with 6 specialized sub-agents for end-to-end scientific discovery, from literature review to paper writing.
Explainable AI for scRNA-seq Analysis
How AI agents make single-cell RNA sequencing analysis explainable — from cell clustering to trajectory inference. Tools, skills, and practical workflows compared.
LabClaw: 240 Skills for Wet Lab Research
LabClaw ships 240 pre-built skills for wet lab research — from PCR protocol design to CRISPR guide RNA selection. The most comprehensive lab-focused skill set in the OpenClaw ecosystem.
AI Literature Search Skills Compared
5 skill libraries for AI-assisted literature search — Semantic Scholar, PubMed, arXiv, Google Scholar integrations. Which one covers your research domain best?
Ecosystem Report: March 2026
The OpenClaw science ecosystem grew from 0 to 54 projects in March 2026. First monthly report covering launches, trends, and what's coming next.
Ecosystem Report: April 2026 — Benchmarks and Autoresearch
April brought 37 new projects, a Hermes Agent breakout (+44K stars), the first paper-locked stable cohort, and one project lost to an account deletion. Here's what changed in the OpenClaw science ecosystem.
MedClaw vs MedgeClaw: Which to Choose?
5 medical AI agents all share similar names. We compare MedClaw (3 variants), MedgeClaw, and MolClaw — architecture, features, and which one fits your research.
Multi-Agent AI Systems for Science
How multi-agent coordination is transforming scientific research — from imperial bureaucracy to AI roundtables. Edict, ClawTeam, MagiClaw, and 7 more projects compared.
Nature Skills: AI for Nature-Quality Figures and Writing
nature-skills is a Claude Code skill hub with two stable skills — nature-figure and nature-polishing — both grounded in primary Nature papers. 265 stars in 6 days. Here's exactly what it does and how to use it.
NeuriCo: Hypothesis In, Research Paper Out
NeuriCo from UChicago Human+AI Lab turns a one-line hypothesis into a complete research paper — with literature review, experiments, and LaTeX writing. Multi-provider (Claude, Codex, Gemini), Docker-ready, and connected to IdeaHub for community research ideas.
Think Like a Nobel Laureate — with AI
Someone distilled the thinking patterns of 52 Nobel Medicine winners into Claude Code skills. Now you can debug your research through the eyes of Katalin Karikó, Shinya Yamanaka, or Jennifer Doudna's co-laureates.
OmicsClaw: AI Agent for Multi-Omics
OmicsClaw automates multi-omics analysis — genomics, transcriptomics, proteomics, metabolomics — with a unified AI agent interface. From Tsinghua AI4Life Lab.
Multi-Omics Integration with OmicsClaw
Step-by-step guide to integrating transcriptomics and proteomics data using OmicsClaw — from raw data to cross-omics correlation analysis.
OpenClaw for Biology: Getting Started
How to set up OpenClaw as a biology research assistant — install, configure LLM providers, add bioinformatics skills, connect to your lab's messaging apps.
OpenClaw Ecosystem Guide 2026
Complete guide to the OpenClaw ecosystem for scientific research — 107+ projects, 24 skill hubs, and how to navigate this rapidly growing landscape.
OpenMAIC: AI Multi-Agent Classroom
OpenMAIC simulates a multi-agent classroom where AI agents learn through teaching, debate, and peer review. From THU-MAIC. Setup guide and use cases for research education.
PaperClaw: 6 Projects, Same Name, Compared
6 independent projects all named PaperClaw. We compare their focus, architecture, and star counts to help you find the right paper-writing AI agent.
ScienceClaw: 4 Projects, 4 Teams
4 independent teams built 4 different ScienceClaws. From multi-agent research platforms to paper writing tools — which ScienceClaw is which?
scRNA-seq Skills: Complete Guide
A complete guide to scRNA-seq analysis skills for AI agents — from quality control to trajectory inference. Which skill libraries cover which steps?
Skill Libraries: Red Ocean vs Blue Ocean
The AI agent skill market is splitting into red oceans (oversaturated categories) and blue oceans (untapped domains). Analysis of 6,300+ ClawHub skills reveals where the opportunities are.
SkillCraft: Can AI Agents Learn to Reuse Tools?
SkillCraft lets AI agents save successful tool chains as reusable skills — cutting token costs by 79% and boosting success rates. A new paradigm for agent skill libraries.
Spatial Transcriptomics with AI Agents
How AI agents handle spatial transcriptomics analysis — spot deconvolution, spatial pattern detection, and tissue architecture mapping. Tools and skills compared.
TCM Biomedical Research Skill for Claude
A specialized Claude Code skill for Traditional Chinese Medicine biomedical research — network pharmacology, herb-target mapping, and TCM-modern medicine integration.