psr-cloud-mcp 0.5.2

MCP server for PSR Cloud (local stdio)

PSR Cloud MCP Server

MCP server that exposes PSR Cloud HPC operations as AI tools, built on top of pycloud (psr-cloud on PyPI).

Prerequisites

1. Python 3.10+
2. Credentials. The server picks one of two paths automatically — you only need to set up the one that fits you:
   • Personal Access Token (recommended). Go to sso.psr-inc.com/profile, log in with your corporate PSR Google account, generate a token under PSR Cloud — Personal Access Tokens, and save it (shown only once). Then export PSR_CLOUD_EMAIL and PSR_CLOUD_ACCESS_TOKEN (see below).
   • PSR Cloud Desktop fallback. If you already have PSR Cloud Desktop installed and signed in, leave both env vars unset — the MCP server will reuse the credentials cached at %APPDATA%\PSR\PSRCloud\EPSRConfig.xml. This path only works on Windows.

Installation

From PyPI

pip install psr-cloud-mcp

From the Git repository

pip install git+https://github.com/your-org/psr-cloud-mcp.git

Local development clone

git clone https://github.com/your-org/psr-cloud-mcp.git
cd psr-cloud-mcp/local
pip install -e .

Configuring Claude Code

Add to ~/.claude/settings.json (or settings.local.json for local-only):

{
  "mcpServers": {
    "psr-cloud": {
      "command": "psr-cloud-mcp",
      "env": {
        "PSR_CLOUD_EMAIL": "[e-mail]",
        "PSR_CLOUD_ACCESS_TOKEN": "<your-token>"
      }
    }
  }
}

If you prefer not to store the token in the config file, omit the env block and set the variables in your shell profile instead. Or, if you already have PSR Cloud Desktop signed in, omit the env block entirely — the server will fall back to the credentials saved by Desktop.

Configuring Claude Desktop

Open the Claude Desktop config file:

• Windows: %APPDATA%\Claude\claude_desktop_config.json
• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

Add the psr-cloud entry under mcpServers:

{
  "mcpServers": {
    "psr-cloud": {
      "command": "psr-cloud-mcp",
      "env": {
        "PSR_CLOUD_EMAIL": "[e-mail]",
        "PSR_CLOUD_ACCESS_TOKEN": "<your-token>"
      }
    }
  }
}

If psr-cloud-mcp is not on the system PATH (common on Windows), use the full path to the executable:

{
  "mcpServers": {
    "psr-cloud": {
      "command": "C:\\Users\\<user>\\AppData\\Local\\Programs\\Python\\Python3xx\\Scripts\\psr-cloud-mcp.exe",
      "env": {
        "PSR_CLOUD_EMAIL": "[e-mail]",
        "PSR_CLOUD_ACCESS_TOKEN": "<your-token>"
      }
    }
  }
}

To find the exact executable path on Windows, run:

where.exe psr-cloud-mcp

After saving the file, restart Claude Desktop. The PSR Cloud tools will appear in the tools panel (hammer icon).

Available tools

Tool	Description
list_cases	List cases from the last N days
get_cases	Get details for specific case IDs
get_case_status	Poll execution status of a case
get_case_log	Retrieve execution log text
list_download_files	List result files available for a case
run_case	Submit a new model run
cancel_case	Cancel a running or queued case
download_case_files	Schedule a background download to a local path; returns a job_id
get_download_status	Poll a background download by job_id
list_downloads	List all background downloads tracked by this server
list_clusters	List the PSR Cloud clusters available to your account
get_programs	List available programs (SDDP, OPTGEN, …)
get_program_versions	List versions for a program
get_execution_types	List execution types for a program + version
get_memory_per_process_ratios	List valid memory ratio strings

Choosing a cluster

Every cluster-scoped tool above accepts an optional cluster argument. Omit it to use your account's default cluster — the right choice for most users. Pass a name (run list_clusters to see them) to operate against a specific cluster; unknown names are rejected with the list of valid options.

Typical AI workflow

get_programs()
  → get_program_versions("SDDP")
  → get_execution_types("SDDP", "18.0")
  → run_case(name="my-run", data_path="/path/to/data", program="SDDP", ...)
  → get_case_status(case_id)         # poll until SUCCESS
  → list_case_files(case_id)
  → download_case_files(case_id, output_path="/path/to/results")  # returns job_id
  → get_download_status(job_id)      # poll until status is "done"