The Privacy-First, Multi-Agent OS for Biological & Pharma Labs
Building the local super-brain for wet & dry labs, beyond monolithic AI.
While existings scientific agents dump fragile tools into a single context window and rely on Cloud LLMs, BioClaw approaches AI-for-Science differently:
- 🔒 Zero-Trust Privacy: Compute molecular docking and genomics entirely locally. Your SME data never leaves the hardware.
- 🧫 Wet-Lab Safety Sandbox: Uses AST parsing to strictly prevent AI-generated destructive code from crashing your physical automated liquid handlers.
- 🧩 Decentralized Micro-Agents: Mathematical deterministic evaluations (like Lipinski Rules via
rdkit) handle data, while the LLM purely routes intent.
The easiest way to run BioClaw is via standard Python or Docker:
git clone https://github.com/qinheming/BloClaw.git
cd BloClaw
poetry install