Getting Started

Welcome to IOWarp Agents - specialized AI subagents for scientific computing workflows!

What are IOWarp Agents?

IOWarp Agents are task-specific AI assistants designed to enhance your scientific computing workflow. Each agent specializes in a particular domain of scientific computing, providing expert guidance and leveraging the IOWarp MCP ecosystem when available.

Quick Installation

Project-Specific Installation (Recommended)

1. Clone the repository:

git clone https://github.com/iowarp/iowarp-agents.git

2. Copy agents to your project:

mkdir -p .claude/agents
cp iowarp-agents/agents/*.md .claude/agents/

Global Installation

For use across all your projects:

# Copy to user-level directory
mkdir -p ~/.claude/agents
cp iowarp-agents/agents/*.md ~/.claude/agents/

Usage in Claude Code

Once installed, agents are automatically available in Claude Code. You can invoke them in two ways:

Using the /agents Command

/agents

This opens the agent selection interface where you can choose the appropriate agent for your task.

Using Natural Language

Simply mention the agent in your request:

• "Use the data-io-expert to help me convert this HDF5 file to Parquet"
• "I need the hpc-performance-expert to optimize my SLURM job"
• "Ask the analysis-viz-expert to create a publication-quality plot"

Available Agents

Agent	Specialization	When to Use
data-io-expert	Scientific data formats & I/O	Converting between HDF5, ADIOS, Parquet, handling compression
analysis-viz-expert	Data analysis & visualization	Statistical analysis, creating plots, pandas operations
hpc-performance-expert	HPC & performance optimization	SLURM jobs, performance profiling, resource optimization
research-doc-expert	Research & documentation	Literature search, experiment tracking, technical writing
workflow-orchestrator	Workflow & environment management	Pipeline creation, reproducibility, automation

Integration with IOWarp MCPs

These agents work seamlessly with IOWarp MCPs when available:

# Install IOWarp MCPs
pip install iowarp-mcps

# Agents will automatically utilize available MCPs
uvx iowarp-mcps pandas  # Used by analysis-viz-expert
uvx iowarp-mcps slurm   # Used by hpc-performance-expert
uvx iowarp-mcps hdf5    # Used by data-io-expert

Example Workflow

Here's a typical scientific computing workflow using IOWarp Agents:

1. Data Preparation: Use data-io-expert to convert and compress your datasets
2. Analysis: Employ analysis-viz-expert for statistical analysis and visualization
3. High-Performance Computing: Leverage hpc-performance-expert for job optimization
4. Documentation: Use research-doc-expert to document your methods and findings
5. Workflow Automation: Apply workflow-orchestrator to create reproducible pipelines

Next Steps

• Browse all available agents on the home page
• Learn about creating custom agents
• Explore advanced usage patterns
• Check out example workflows

---

Need help? Check out our GitHub Issues or join the community discussion.