Welcome to IOWarp Documentation
IOWarp is a context orchestration platform for agentic AI in scientific computing. It provides a high-performance runtime, SDK, and agent toolkit that bridges AI agents with HPC infrastructure, scientific data formats, and research workflows.
Who is this for?
| Audience | Start here |
|---|---|
| Researchers | Installation Guide → Quick Start |
| HPC Practitioners | Deployment Guide → Configuration |
| Developers | SDK Reference → Python API |
| AI Researchers | CLIO Kit MCP Servers → Platform Overview |
Architecture Overview
IOWarp consists of three layers:
- CLIO Agent — The orchestration intelligence. Coordinates AI agents with scientific workflows through context engineering.
- CLIO Kit — 16 MCP servers with 150+ tools for scientific computing (HDF5, Slurm, ParaView, Pandas, ArXiv, and more).
- IOWarp Runtime (Chimaera) — High-performance distributed task execution framework with the Context Transfer Engine (CTE) and Context Assimilation Engine (CAE).
Key Capabilities
- Context Transfer Engine (CTE) — Intelligent data buffering with multi-tier storage, automatic data placement, and zero-copy transfers
- Context Assimilation Engine (CAE) — Transform and adapt data across formats and domains
- MCP Integration — Model Context Protocol servers for seamless AI agent connectivity
- HPC-Native — Built for Slurm clusters, POSIX/MPI-IO, and distributed scientific workflows
Quick Links
- GitHub Organization
- Zulip Chat
- Research Publications
- CLIO Kit (toolkit.iowarp.ai)
- IOWarp Website (iowarp.ai)
Need Help?
- Community — Zulip Chat
- Bug Reports — GitHub Issues
- Feature Requests — GitHub Discussions
- Email — grc@illinoistech.edu