IOWarp
IOWarp is GRC's research and engineering program behind CLIO, the Context Layer for I/O. It connects AI-assisted science to the data, software, memory, and computing systems that research depends on, from a workstation to a national facility.
This page records the research program, its relationships, and its publications. Visit iowarp.ai (opens in a new tab) for the platform, applications, demonstrations, and current product information.
Research focus
- Context: keep scientific data connected to its metadata, provenance, tools, and system state so people and agents can interpret what they use.
- Memory: retain useful context across sessions and connect it to the evidence that produced it.
- Scale: connect scientific workflows to memory and storage tiers, cluster schedulers, and computing facilities as their needs grow.
The program builds on GRC research in Hermes, LABIOS, ChronoLog, and tiered memory systems. The platform overview (opens in a new tab) explains how those research lines inform CLIO.
The CLIO suite
CLIO Agent serves scientists working with data; CLIO Coder serves researchers developing software. CLIO Kit supplies scientific tools and workflow guidance, while CLIO Core manages the underlying context and data. The products can be adopted independently; their own sites document installation, supported environments, and integration requirements.
CLIO Agent (opens in a new tab)
A desktop and browser workspace for scientific data, analysis, and research.
CLIO Coder (opens in a new tab)
A coding agent for scientific software, with shared project guidance and verifiable checks across models.
CLIO Core (opens in a new tab)
A composable runtime for scientific data and context across memory and storage tiers.
CLIO Kit (opens in a new tab)
Scientific MCP servers, skills, and workflow plugins for the agent you use.
Research and applications
IOWarp explores how agents can use scientific data and systems while leaving results that researchers can inspect. Applications span materials science, simulation workflows, scientific analysis, and data discovery. See the application stories (opens in a new tab) for current examples and the research record (opens in a new tab) for their supporting work.
Related GRC publications include Acropolis, on adaptive content indexing for scientific data; AgentStage, on staging data during an agent's reasoning; and MegaMmap (opens in a new tab), on tiered distributed shared memory. These are individual research results, with their own workloads and evaluation conditions.
Browse all IOWarp publications or visit the project's results and deployment record (opens in a new tab).
Laboratory collaborations
GRC contributes CLIO capabilities to PORTUS, led by Pacific Northwest National Laboratory, and SPOTTER-AI, led by Argonne National Laboratory, through DOE Genesis Mission work. The individual project records describe their research scope, teams, and support.
The IOWarp program's team, funding, research lineage, and related publications are recorded alongside this overview. Product architecture, APIs, installation instructions, and deployment tutorials live on the IOWarp documentation hub (opens in a new tab) and the four product sites above.