Refua enables AI agents to perform complex biomolecular folding, binding affinity estimation, and ADMET profiling within the Openclaw Skills ecosystem.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install refua
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install refua using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Refua?
Refua is a specialized computational biology tool designed to bridge the gap between AI agents and drug discovery workflows. By integrating with Openclaw Skills, it provides a unified interface for folding and scoring biomolecular complexes, such as protein-ligand and protein-protein interactions. It leverages high-performance models like Boltz2 and BoltzGen to provide structural insights that are critical for prioritizing molecular candidates in research pipelines.
This skill acts as a bridge to the refua-mcp server, allowing agents to execute GPU-accelerated tasks for scientific discovery. Whether you are predicting the 3D structure of a novel binder or evaluating the pharmacological properties of a lead compound, Refua offers a robust, developer-friendly framework for automated molecular analysis.
Refua Use Cases
- Folding high-resolution structures of protein-ligand and protein-protein complexes.
- Predicting structural orientations for DNA and RNA complexes.
- Estimating the binding affinity of specific binders to identify high-potential drug candidates.
- Generating ADMET (Absorption, Distribution, Metabolism, Excretion, and Toxicity) profiles for SMILES-based ligands.
- Automating complex molecular design workflows through generative AI tools within the Openclaw Skills framework.
How Refua Works
- The AI agent receives a molecular research request and identifies the need for structural analysis using Openclaw Skills.
- The skill initiates a connection to the refua-mcp server, which exposes the Refua Unified Complex API.
- Input data, such as protein sequences and ligand SMILES strings, are passed to the Boltz2 or BoltzGen engine.
- The system performs GPU or CPU-based computational folding and evaluates binding metrics or pharmacological properties.
- The refined structural data and scores are returned as structured metadata, allowing the agent to provide actionable insights or move to the next stage of the discovery pipeline.
Refua Setup
To integrate Refua into your environment as part of the Openclaw Skills library, follow these installation steps:
# Install Refua with CUDA support (recommended for performance)
pip install refua[cuda]
# Install the MCP server interface
pip install refua-mcp
# Optional: Install ADMET profiling dependencies
pip install refua[admet]
# Download necessary model weights and molecular assets
python -c "from refua import download_assets; download_assets()"
# Launch the MCP server to enable tool access
python3 -m refua_mcp.server
Refua Data Schema & Taxonomy
Refua manages scientific data through a structured taxonomy of model assets and tool outputs:
| Component | Type | Description |
|---|---|---|
| Boltz2 Cache | Filesystem | Stores model weights in ~/.boltz by default. |
| BoltzGen Assets | HF Artifacts | Bundled molecular design data for generative workflows. |
| SMILES | String | Standard input format for chemical ligands. |
| FASTA | String | Standard input format for protein and nucleic acid sequences. |
| MCP Tool Results | JSON | Structured output containing folding coordinates and affinity scores. |
Refua Advanced Features
- GPU-accelerated folding using CUDA-optimized Refua builds for rapid structural prediction.
- Native support for BoltzGen, enabling complex molecular design and optimization workflows.
- Configurable asset management allowing users to override default cache directories for Boltz2 and BoltzGen.
- Seamless multi-agent interoperability through the standardized Model Context Protocol (MCP).
- Extensible ADMET profiling module for comprehensive pharmacological evaluation of small molecules.
SKILL.md
Loading
Related Openclaw Skills
Chaos Lab
A multi-agent research framework designed to study AI alignment through conflicting optimization targets and emergent behaviors.
Fliz AI Video Generator
Programmatically transform text content into professional AI-generated videos with voiceovers and automated subtitles.
Humanizer
A sophisticated writing editor that identifies and removes common AI-generated signatures to create natural, human-sounding text.
Smart Memory System
A Retrieval-Augmented Generation (RAG) based intelligence system providing semantic search and memory optimization for Openclaw Skills.
docstream
docstream is a high-performance document processing skill that enables automated text extraction, AI summarization, and format conversion.
InvoiceForge API
A high-performance API for creating, managing, and downloading professional PDF invoices with automated calculations.
Featured*
