IOWarp is an NSF-funded ($5M, Award #2411318) open-source context engineering platform enabling autonomous AI agents to orchestrate petabyte-scale scientific discovery.

CLIO (Context Layer for IO) is the software platform — transforming scattered data, metadata, telemetry, and computational state into unified, actionable intelligence for AI agents.

graph TB
    subgraph Intelligence["Intelligence Layer"]
        direction LR
        subgraph ExtAgents["External Agents"]
            CC[Claude Code]
            GC[Gemini CLI]
            OC[OpenCode]
        end
        subgraph CLIOAgent["CLIO Agent"]
            CA["Agent SDK"]
        end
    end

    subgraph Tooling["Tooling Layer"]
        direction LR
        subgraph CLIOKit["CLIO Kit"]
            MCP[MCP Servers]
            SK[Skills]
            PL[Plugins]
        end
    end

    subgraph Storage["Storage Layer — CLIO Core"]
        subgraph Engines["Four Context Engines"]
            direction LR
            Ingest["CLIO Ingest"]
            Transfer["CLIO Transfer"]
            Explore["CLIO Explore"]
            Plugin["CLIO Plugin"]
        end
        subgraph Infra["Infrastructure"]
            direction LR
            RT["Chimaera Runtime"]
            CTP["Hermes SHM"]
        end
    end

    ExtAgents <--> CLIOKit
    CLIOAgent <--> CLIOKit
    CLIOAgent -.->|direct| Engines
    CLIOKit <--> Engines
    Engines <--> Infra

• CLIO Ingest — "The Way In" — ingests 14+ scientific formats (HDF5, NetCDF, Zarr, ROOT, FITS)
• CLIO Transfer — "The Mover" — ML-based predictive data movement across storage tiers
• CLIO Explore — "The Finder" — natural language queries at petabyte scale
• CLIO Plugin — "The Connector" — HPC scheduler, storage, and instrument integration

• 7.5x faster scientific workflows
• 2.6x DRAM reduction
• 150+ agent tools via MCP
• 25+ peer-reviewed publications
• 2 Best Paper Awards (SC, HPDC)
• Deployed at 3+ national labs (LLNL, ANL, NERSC)

We welcome contributions across all repositories. See individual CONTRIBUTING.md files for guidelines.

Development workflow:

1. Fork the relevant repository
2. Create a feature branch
3. Follow coding standards (Google C++ Style Guide for C++, PEP 8 for Python)
4. Submit a pull request with tests

Built by Gnosis Research Center at Illinois Institute of Technology, with University of Utah and The HDF Group.

All IOWarp repositories are licensed under the BSD 3-Clause License.

Copyright (c) 2024-2026, Gnosis Research Center, Illinois Institute of Technology

Gnosis Research Center · iowarp.ai · docs.iowarp.ai · toolkit.iowarp.ai · GitHub