CLIO Kit MCP Servers Reference

15+ MCP Servers with 150+ Tools for Scientific Computing

What is MCP?

The Model Context Protocol (MCP) enables AI agents to interact with external tools through standardized interfaces. IOWarp provides 15+ specialized MCP servers that expose scientific computing capabilities to AI agents, enabling them to read HDF5 files, submit Slurm jobs, analyze I/O patterns, and much more.

Key Benefits

  • • Standardized Interface: All servers follow the MCP protocol
  • • Tool Discovery: Agents automatically discover available capabilities
  • • Composable: Mix and match servers for different workflows
  • • Extensible: Easy to add custom servers

Installation

Install MCP servers in your preferred IDE. Each IDE has a slightly different configuration format.

Claude Code

claude mcp add clio-hdf5 -- uvx clio-kit mcp-server hdf5
claude mcp add clio-slurm -- uvx clio-kit mcp-server slurm

Cursor

Add to .cursor/mcp.json:

{
  "mcpServers": {
    "clio-hdf5": {
      "command": "uvx",
      "args": ["clio-kit", "mcp-server", "hdf5"]
    },
    "clio-slurm": {
      "command": "uvx",
      "args": ["clio-kit", "mcp-server", "slurm"]
    }
  }
}

VS Code

Add to settings.json:

{
  "mcp.servers": {
    "clio-hdf5": {
      "command": "uvx",
      "args": ["clio-kit", "mcp-server", "hdf5"]
    }
  }
}

Detailed Server Reference

1. mcp-hdf5

Read/write HDF5 files with full dataset and attribute support

GitHub

Tools

  • • read_dataset — Read HDF5 dataset
  • • write_dataset — Write HDF5 dataset
  • • list_groups — List HDF5 groups
  • • get_attributes — Read attributes
  • • create_group — Create new group

Use Cases

  • • Climate simulation data analysis
  • • Scientific dataset exploration
  • • HDF5 file manipulation

2. mcp-slurm

HPC job management with Slurm integration

GitHub

Tools

  • • submit_job — Submit Slurm job
  • • cancel_job — Cancel running job
  • • queue_status — Check queue status
  • • node_info — Get node information
  • • job_history — Query job history

Use Cases

  • • Automated job submission
  • • Resource monitoring
  • • Workflow orchestration

3. mcp-darshan

I/O profiling analysis with Darshan logs

GitHub

Tools

  • • parse_log — Parse Darshan log
  • • get_io_summary — I/O statistics
  • • compare_runs — Compare multiple runs
  • • identify_bottlenecks — Find I/O issues

Use Cases

  • • Performance optimization
  • • I/O pattern analysis
  • • Bottleneck identification

4. mcp-lmod

Environment modules management

GitHub

Tools

  • • load_module — Load module
  • • list_modules — List available modules
  • • module_info — Get module details
  • • unload_module — Unload module

Use Cases

  • • Environment setup automation
  • • Dependency management
  • • Software version control

5. mcp-jarvis

Software deployment and service management

GitHub

Tools

  • • deploy_pipeline — Deploy software pipeline
  • • start_service — Start service
  • • check_status — Check service status
  • • rollback — Rollback deployment

Use Cases

  • • Automated deployments
  • • Service orchestration
  • • Infrastructure management

6. mcp-node-hardware

Hardware query and system information

GitHub

Tools

  • • get_cpus — CPU information
  • • get_gpus — GPU information
  • • get_memory — Memory details
  • • get_storage — Storage devices

Use Cases

  • • Resource discovery
  • • Capacity planning
  • • Hardware monitoring

7. mcp-pandas

DataFrame operations and analysis

GitHub

Tools

  • • read_csv — Read CSV file
  • • describe — DataFrame statistics
  • • filter — Filter rows
  • • aggregate — Aggregation operations

Use Cases

  • • Data analysis
  • • Statistical operations
  • • Data transformation

8. mcp-parquet

Columnar data format operations

GitHub

Tools

  • • read_parquet — Read Parquet file
  • • write_parquet — Write Parquet file
  • • get_schema — Get schema information
  • • query_columns — Columnar queries

Use Cases

  • • Big data analytics
  • • Columnar storage operations
  • • Efficient data storage

9. mcp-arxiv

Paper search and retrieval

GitHub

Tools

  • • search_papers — Search arXiv
  • • get_abstract — Get paper abstract
  • • download_pdf — Download PDF
  • • get_citations — Citation information

Use Cases

  • • Literature review
  • • Research discovery
  • • Citation analysis

10. mcp-compression

Data compression and decompression

GitHub

Tools

  • • compress — Compress data
  • • decompress — Decompress data
  • • estimate_ratio — Compression ratio
  • • list_algorithms — Available algorithms

Use Cases

  • • Storage optimization
  • • Data transfer efficiency
  • • Archive management

11. mcp-parallel

Distributed compute operations

GitHub

Tools

  • • scatter — Scatter data to nodes
  • • gather — Gather data from nodes
  • • reduce — Reduce operation
  • • broadcast — Broadcast to all nodes

Use Cases

  • • Parallel processing
  • • Distributed computing
  • • Map-reduce operations

12. mcp-plot

Data visualization and plotting

GitHub

Tools

  • • line_plot — Create line plot
  • • scatter_plot — Create scatter plot
  • • heatmap — Create heatmap
  • • histogram — Create histogram

Use Cases

  • • Data visualization
  • • Result presentation
  • • Exploratory analysis

13. mcp-adios

Streaming I/O operations

GitHub

Tools

  • • open_stream — Open ADIOS stream
  • • read_step — Read time step
  • • write_step — Write time step
  • • list_variables — List stream variables

Use Cases

  • • Real-time data streaming
  • • Time-series analysis
  • • Simulation I/O

14. mcp-chronolog

Distributed shared log for multi-agent coordination and event ordering

GitHub

Tools

  • • append_log — Append to distributed log
  • • read_log — Read log entries
  • • query_events — Query events by time range
  • • get_ordering — Get event ordering

Use Cases

  • • Multi-agent coordination
  • • Event ordering
  • • Distributed consensus

15. mcp-ndp

Near-Data Processing for in-situ computation at storage layer

GitHub

Tools

  • • execute_compute — Execute computation at storage
  • • filter_data — Filter data in-place
  • • transform_data — Transform data during I/O
  • • get_capabilities — Get NDP capabilities

Use Cases

  • • In-situ data processing
  • • Storage-level computation
  • • Reduced data movement

16. mcp-paraview

ParaView scientific visualization and rendering

GitHub

Tools

  • • load_dataset — Load dataset for visualization
  • • create_visualization — Create visualization
  • • render_image — Render visualization to image
  • • export_animation — Export animation sequence

Use Cases

  • • Scientific visualization
  • • Volume rendering
  • • Animation generation

Meta-MCP Protocol

Dynamic routing to appropriate servers: The Meta-MCP protocol enables intelligent routing of requests to the most appropriate MCP server based on context, capabilities, and availability.

How It Works

  • • Agent requests a capability (e.g., "read HDF5 file")
  • • Meta-MCP router identifies available servers
  • • Selects optimal server based on load, latency, capabilities
  • • Routes request and returns result

Benefits

  • • Load Balancing: Distribute requests across servers
  • • Failover: Automatic failover to backup servers
  • • Transparency: Agents don't need to know which server handles requests
  • • Extensibility: Add new servers without changing agent code

Additional Resources

CLIO Kit

Overview of the MCP server ecosystem

GitHub Repository

Source code and detailed documentation