CLIO Kit

CLIO Kit is the IoWarp meta-marketplace for scientific computing. It brings together MCP servers, workflow skills, plugins, agent definitions, and community contributions—including plugins from external marketplaces—and enables AI agents to interact with HPC resources, scientific data formats, and research datasets.

Choose individual MCPs, skills, agents and hooks—or a plugin that bundles what your workflow needs. A plugin can include any of these components; it does not need all four.

Website | IOWarp

Documentation · For agents: read here.

Chat with us on Zulip or join us

Developed by <img src="https://grc.iit.edu/img/logo.png" alt="GRC Logo" width="18" height="18"> Gnosis Research Center

❌ Without CLIO Kit

Working with scientific data and HPC resources requires manual scripting and tool-specific knowledge:

• ❌ Write custom scripts for every HDF5/Parquet file exploration
• ❌ Manually craft Slurm job submission scripts
• ❌ Switch between multiple tools for data analysis
• ❌ No AI assistance for scientific workflows
• ❌ Repetitive coding for common research tasks

✅ With CLIO Kit

Install workflow bundles, use skills to guide your agent, and discover community tools:

• ✅ "Analyze the temperature dataset in this HDF5 file" - HDF5 MCP does it
• ✅ "Submit this simulation to Slurm with 32 cores" - Slurm MCP handles it
• ✅ "Find papers on neural networks from ArXiv" - ArXiv MCP searches
• ✅ "Plot the results from this CSV file" - Plot MCP visualizes
• ✅ "Optimize memory usage for this pandas DataFrame" - Pandas MCP optimizes
• ✅ Workflow guidance - Skills explain tool sequences; optional agents help plan and review results
• ✅ Community contributions - Discover external plugins and marketplace collections

MCP servers · Skills · Plugins · Agents · Hooks · Community marketplaces

CLIO Kit is part of the IoWarp platform's tooling ecosystem for AI agents. Its meta-marketplace brings tools and reusable workflows together for scientific computing—from exploring HDF5 datasets and managing Slurm jobs to analyzing results and finding research papers. Built by researchers, for researchers, at Illinois Institute of Technology with NSF support.

Part of IoWarp Platform: CLIO Kit is the platform's meta-marketplace and tooling layer, bringing together MCP servers, skills, workflow plugins, agent definitions, and community marketplace collections.

One launcher for MCP servers. Individual components or workflow bundles. BSD-3-Clause licensed, with automated checks and real installation, MCP, and HPC workflow tests. See the validation guide for tested coverage and environment requirements.

🚀 Quick Installation

git clone https://github.com/iowarp/clio-kit.git
cd clio-kit

Recommended: Ask your agent:

Read setup.md and set up CLIO Kit for me.

Manual setup: Install this checkout's launcher, then choose an option below. For the no-clone release route (2.11.0 or newer), see selective installation.

uv tool install --force --reinstall --editable ".[verification]"

Claude Code

Install a workflow plugin (bundle) with its MCP servers and skills:

claude plugin marketplace add "$PWD"
claude plugin install clio-hpc@clio-kit
claude mcp list

Restart Claude Code after installation and confirm the servers connect. First launches download dependencies; see startup recovery if a server exceeds the client's connection timeout.

Or install an individual MCP or a skill collection:

claude plugin install clio-hdf5@clio-kit                  # HDF5 server
claude plugin install clio-scientific-io-skills@clio-kit  # Scientific I/O skills

Claude Code uses plugin install for component packages too. CLIO Kit lists those separately from workflow plugins; see components and plugins.

Codex and Other Agents

Install a workflow's skills and MCP configuration together:

clio-kit plugin install clio-scientific-io --client codex --project /path/to/project

Also supports opencode, cursor, antigravity, vscode and claude-code. This sets up project components; native agents and hooks need client-specific adapters. See client installation and limits.

Install a bundle or one skill into your project:

# Bundle
clio-kit skill install --bundle clio-scientific-io --target /path/to/project/.agents/skills
# Individual skill
clio-kit skill install dataset-explore --target /path/to/project/.agents/skills

Configure the required MCP servers using setup.md. Use your agent’s supported skill directory.

Install Skills with npm

Requires Node.js 22.20.0+. From your working project:

# All skills
npx skills@1.5.25 add /path/to/clio-kit --skill '*' --agent codex --copy
# Individual skill
npx skills@1.5.25 add /path/to/clio-kit --skill dataset-explore --agent codex --copy

Replace codex with claude-code or antigravity. MCP servers require separate configuration.

Choose either the Python or npm installer for each skill. See the skills CLI guide for selecting skills, updates and removal.

One Command for Any Server

# From the CLIO Kit checkout, install the CLI in a persistent tool environment
uv tool install --force --reinstall --editable .
# If uv's executable directory is not on PATH:
uv tool update-shell

# List all 22 available MCP servers
clio-kit mcp-servers

# Run an individual server
clio-kit mcp-server hdf5
clio-kit mcp-server pandas
clio-kit mcp-server slurm

Once 2.11.0 is published, install it with uv tool install 'clio-kit>=2.11.0' (2.11.0 or newer; earlier releases lack the skill, plugin and doctor commands). uv tool install creates a persistent, isolated environment; use uvx --from 'clio-kit>=2.11.0' clio-kit ... for a temporary, one-shot invocation.

The release separates the small launcher from component downloads: one MCP fetches only that server, one skill fetches only its files, and a workflow selects its components. Server dependencies stay isolated and locked. See selective installation for release, checkout and native plugin routes. The --branch option is for development, not an immutable release installation.

The wheel also includes machine-readable tool contracts for JARVIS, Slurm, Spack and Scientific Catalog, with SHA-256 digests for downstream verification:

clio-kit mcp-contracts
clio-kit mcp-contract clio-kit-jarvis-user-v3.7.2
clio-kit mcp-contract clio-kit-slurm-user-v3
clio-kit mcp-contract clio-kit-spack-user-v2.3
clio-kit mcp-contract clio-kit-scientific-catalog-user-v1.1

Workflow Bundles

These are broad workflow plugins. Each combines its MCP servers and skills.

Some servers require additional system software. See setup and prerequisites. See plugin composition for primary bundles and task plugins that combine components across them.

For a complete dataset-report workflow—HDF5, Pandas, Plot, a skill, evidence review and a verification hook:

claude plugin install clio-dataset-report@clio-kit

See dataset reports for inputs and verification limits.

Skills and Agents

Skills provide workflow instructions and guidance for interpreting results. Claude Code users can install skills or agent definitions separately:

claude plugin install clio-skills@clio-kit
claude plugin install clio-agents@clio-kit

The shared installer supports Codex, Claude Code, OpenCode, Cursor, Antigravity and VS Code. Codex, Claude Code and OpenCode also have adapters for the scientific agents and Dataset Report hooks. See client support.

Add the optional Clio Coder coding and research skills, including Materio procedures:

clio-kit skill install --bundle clio-coder --target .agents/skills
# Or through Claude Code:
claude plugin install clio-coder-skills@clio-kit

Individual skills also work: clio-kit skill install clio-kit-scientific-debugging --target .agents/skills. Adapted Clio Coder skills use the clio-kit- prefix to distinguish them from upstream originals. See Clio Coder integration for upstream packages and native plugin support.

Contribute Servers, Skills or Plugins

Contribute a package folder under plugins/, skills/, agents/ or hooks/; website builds and CI update the catalogues automatically. See the folder contribution guide.

Add skills here, or list a plugin maintained in your own repository:

clio-kit plugin init my-plugin
clio-kit plugin validate my-plugin
clio-kit plugin submit my-plugin --repo owner/name

submit generates a community entry for a pull request. External servers can use any language; servers hosted inside CLIO currently support Python, Node.js and Go.

Add an optional Claude Code hook with clio-kit plugin init my-plugin --hook. See the hook guide.

See the contribution guide and community guide for requirements and submission details.

<a id="agent-integrations"></a>

Add to .cursor/mcp.json in your project, or ~/.cursor/mcp.json:

{
  "mcpServers": {
    "clio-hdf5": { "command": "clio-kit", "args": ["mcp-server", "hdf5"] },
    "clio-adios": { "command": "clio-kit", "args": ["mcp-server", "adios"] },
    "clio-parquet": { "command": "clio-kit", "args": ["mcp-server", "parquet"] },
    "clio-compression": { "command": "clio-kit", "args": ["mcp-server", "compression"] }
  }
}

Install the workflow skills from your project:

clio-kit skill install --bundle clio-scientific-io --target .cursor/skills

See Cursor MCP docs and skills docs.

From the CLIO Kit checkout, install the workflow plugin with its MCPs and skills:

claude plugin marketplace add "$PWD"
claude plugin install clio-scientific-io@clio-kit
claude mcp list

Skills-only and agent plugins:

claude plugin install clio-skills@clio-kit    # scientific workflow skills, no MCP servers
claude plugin install clio-agents@clio-kit    # workflow planner and evidence reviewer

See setup.md for details.

Add to .vscode/mcp.json for GitHub Copilot:

{
  "servers": {
    "clio-hdf5": { "type": "stdio", "command": "clio-kit", "args": ["mcp-server", "hdf5"] },
    "clio-adios": { "type": "stdio", "command": "clio-kit", "args": ["mcp-server", "adios"] },
    "clio-parquet": { "type": "stdio", "command": "clio-kit", "args": ["mcp-server", "parquet"] },
    "clio-compression": { "type": "stdio", "command": "clio-kit", "args": ["mcp-server", "compression"] }
  }
}

Install the workflow skills from your project:

clio-kit skill install --bundle clio-scientific-io --target .github/skills

Start the servers with MCP: List Servers. The Codex extension uses the Codex configuration below. See VS Code MCP docs and skills docs.

Edit claude_desktop_config.json through Settings → Developer → Edit Config:

{
  "mcpServers": {
    "clio-hdf5": { "command": "clio-kit", "args": ["mcp-server", "hdf5"] },
    "clio-adios": { "command": "clio-kit", "args": ["mcp-server", "adios"] },
    "clio-parquet": { "command": "clio-kit", "args": ["mcp-server", "parquet"] },
    "clio-compression": { "command": "clio-kit", "args": ["mcp-server", "compression"] }
  }
}

Restart Claude Desktop. This config adds MCP servers only; CLIO's native skill and agent plugins target Claude Code. See Claude Desktop MCP docs.

Add the MCP servers for Codex CLI or its IDE extension:

codex mcp add clio-hdf5 -- clio-kit mcp-server hdf5
codex mcp add clio-adios -- clio-kit mcp-server adios
codex mcp add clio-parquet -- clio-kit mcp-server parquet
codex mcp add clio-compression -- clio-kit mcp-server compression
codex mcp list

Install the workflow skills from your project:

clio-kit skill install --bundle clio-scientific-io --target .agents/skills

Check /mcp and /skills in Codex. To install a complete supported workflow, use clio-kit plugin install clio-dataset-report --client codex --project .; review its hooks with /hooks before running it. See Codex MCP docs and skills docs.

Open MCP Servers → Manage MCP Servers → View raw config and add:

{
  "mcpServers": {
    "clio-hdf5": { "command": "clio-kit", "args": ["mcp-server", "hdf5"] },
    "clio-adios": { "command": "clio-kit", "args": ["mcp-server", "adios"] },
    "clio-parquet": { "command": "clio-kit", "args": ["mcp-server", "parquet"] },
    "clio-compression": { "command": "clio-kit", "args": ["mcp-server", "compression"] }
  }
}

Install the workflow skills from your project:

clio-kit skill install --bundle clio-scientific-io --target .agents/skills

Reload the MCP configuration. For Antigravity CLI, start the first session from this project with agy --new-project so its local skills and MCP configuration are loaded. Reopen it with agy --project <project-name-or-id>. Reuse .agents/skills if already installed for Codex. CLIO does not yet ship a native Antigravity plugin. See Antigravity MCP docs and skills docs.

Available Packages

The version below identifies each MCP server's contract and runtime release, independently of the containing clio-kit wheel version. JARVIS 3.7 and SLURM 3.0 have contracts redesigned for agent use. Spack is at 2.3, while the other contracts retain their existing 2.x identities until a focused upgrade.

The Spack install contract makes concretization explicit: reuse=true passes spack install --reuse, while reuse=false passes spack install --fresh. Agents should discover first, install only when needed, then pass the exact spack_locate result to JARVIS for runtime loading. A find with no installed match is normal typed data (count=0, packages=[]); locate reports the distinct not_installed semantic, while real Spack failures remain errors.

📖 Usage Examples

HDF5: Scientific Data Analysis

"What datasets are in climate_simulation.h5? Show me the temperature field structure and read the first 100 timesteps."

Tools used: open_file, analyze_dataset_structure, read_partial_dataset, list_attributes

Slurm: HPC Job Management

"Submit simulation.py to Slurm with 32 cores, 64GB memory, 24-hour runtime. Monitor progress and retrieve output when complete."

Tools used: slurm_submit, slurm_list, slurm_describe

ArXiv: Research Discovery

"Find the latest papers on diffusion models from ArXiv, get details on the top 3, and export citations to BibTeX."

Tools used: search_arxiv, get_paper_details, export_to_bibtex, download_paper_pdf

Pandas: Data Processing

"Load sales_data.csv, clean missing values, compute statistics by region, and save as Parquet."

Tools used: load_data, handle_missing_data, groupby_operations, save_data

Plot: Data Visualization

"Create a line plot showing temperature trends over time from weather.csv with proper axis labels."

Tools used: line_plot, data_info

🚨 Troubleshooting

If clio-kit mcp-server <server-name> fails:

# Verify server name is correct
clio-kit mcp-servers

# Common names: hdf5, pandas, slurm, arxiv (not hdf5-mcp, pandas-mcp)

For development or local testing:

cd mcp-servers/hdf5
uv sync --all-extras --dev
uv run hdf5-mcp

Install uv package manager:

# Linux/macOS
curl -LsSf https://astral.sh/uv/install.sh | sh

# Windows
powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

# Or via pip
pip install uv

Then install CLIO Kit persistently and expose uv's tool directory:

uv tool install --force --reinstall --editable .
uv tool update-shell

Open a new shell after update-shell, then check it resolved:

clio-kit mcp-servers

Team

• Gnosis Research Center (GRC) - Illinois Institute of Technology | Lead
• HDF Group - Data format and library developers | Industry Partner
• University of Utah - Research collaboration | Domain Science Partner

Sponsored By

<img src="https://www.nsf.gov/themes/custom/nsf_theme/components/molecules/logo/logo-desktop.png" alt="NSF Logo" width="24" height="24"> NSF (National Science Foundation) - Supporting scientific computing research and AI integration initiatives

we welcome more sponsorships. please contact the Principal Investigator

Ways to Contribute

• Submit Issues: Report bugs or request features via GitHub Issues
• Build MCP Servers: Contribute Python, Node.js or Go servers (guide)
• Write Skills: Share reusable scientific workflows (guide)
• Share Plugins or Marketplaces: List tools maintained in your own repository (guide)
• Improve Documentation: Help make guides clearer
• Share Use Cases: Tell us how you're using CLIO Kit in your research

Full Guide: CONTRIBUTING.md

Community & Support

• Chat: Zulip Community
• Join: Invitation Link
• Issues: GitHub Issues
• Website: https://toolkit.iowarp.ai/
• Project: IOWarp Project