Open-source infrastructure for AI-native scientific labs.
LabClaw provides a standard framework for running scientific workflows with AI agents — from real-time data capture to hypothesis generation to self-improving experimental design.
| Traditional Tools | LabClaw | |
|---|---|---|
| Memory | Manual notes, lost when people leave | Persistent 3-tier memory that grows with your lab |
| Instruments | Disconnected from analysis | Real-time edge monitoring with automated quality checks |
| Learning | Redundant experiments, no institutional knowledge | Self-evolving strategies that improve with every experiment |
Layer 5 PERSONA Digital lab staff with training and promotion pipeline
Layer 4 MEMORY Markdown + Knowledge Graph + Shared Blocks
Layer 3 ENGINE OBSERVE → HYPOTHESIZE → EXPERIMENT → VALIDATE → EVOLVE
Layer 2 INFRA FastAPI Gateway, Event Bus, Dashboard, Edge Nodes
Layer 1 HARDWARE Device Registry, Safety Checker, Protocol Adapters
| Project | Description |
|---|---|
| labclaw | Core platform — Python 3.11+, Apache 2.0 |
| awesome-physical-ai-for-science | Curated list of AI systems for scientific laboratories |
| Website | labclaw.org |
| Docs | docs.labclaw.org |
| License | Apache 2.0 |
Open source, expanding to all experimental sciences.
