opencosmo-portal 0.1.1

CLI and local MCP server for the OpenCosmo platform

OpenCosmo Portal CLI (ocp)

Command-line interface and local MCP server for the OpenCosmo platform. Query cosmological simulation datasets at DOE Leadership Computing Facilities from your terminal.

Installation

pip install opencosmo-portal

Or for development:

cd cli
uv sync

Quick Start

The default profile points to http://localhost:8000 (local development). For development against a local backend:

# Authenticate via browser (Globus OAuth)
ocp auth login

# Browse available tasks
ocp task list
ocp task info <slug>

# Submit a task and monitor the run
ocp task run <slug>
ocp run status <run-id> --watch
ocp run results <run-id>

Commands

ocp auth — Authentication

Command	Description
ocp auth login	Authenticate via browser (Globus OAuth)
ocp auth logout	Clear stored tokens
ocp auth status	Show token status and expiry

ocp config — Profile Management

Command	Description
ocp config list	List all profiles
ocp config add-profile <name> <url>	Add or update a profile
ocp config set-profile <name>	Set default profile
ocp config remove-profile <name>	Delete a profile
ocp config show [name]	Show profile details

ocp task — Tasks

Command	Description
ocp task list	List available tasks
ocp task info <slug>	Show task details and input parameters
ocp task run <slug>	Submit a task (interactive or --input)

ocp run — Runs

Command	Description
ocp run list	List your runs
ocp run status <id> [--watch]	Check run status
ocp run logs <id>	View run logs
ocp run results <id>	Download results
ocp run cancel <id>	Cancel a run
ocp run archive <id>	Archive a run

ocp whoami — User

Show current user info.

ocp mcp — MCP Server

Start a local stdio MCP server for AI assistants like Claude Desktop.

ocp mcp start

Configure in Claude Desktop (claude_desktop_config.json):

{
  "mcpServers": {
    "opencosmo": {
      "command": "ocp",
      "args": ["mcp", "start"]
    }
  }
}

Global Options

Option	Description
--profile, -p	Use a specific profile
--format, -f	Output format: table (default) or json
--version	Show version
--help	Show help for any command

Releasing

The CLI is published to PyPI as opencosmo-portal via GitHub Actions (trusted publisher / OIDC).

Automated (recommended)

Use the Release CLI workflow in GitHub Actions:

1. Go to Actions → Release CLI → Run workflow
2. Select the bump type (patch, minor, or major)
3. The workflow bumps cli/pyproject.toml, commits, tags, and pushes — which triggers the publish workflow automatically

Manual

1. Bump version in cli/pyproject.toml (or run cd cli && uv version --bump patch)
2. Commit: git commit -am "release: CLI v0.2.0"
3. Tag: git tag cli-v0.2.0
4. Push both: git push origin main cli-v0.2.0

The publish-cli.yml workflow will:

• Run the full test suite
• Verify the tag version matches pyproject.toml
• Build and publish to PyPI

First-time setup

Register a trusted publisher on pypi.org:

Field	Value
Package name	opencosmo-portal
Owner	ArgonneCPAC
Repository	OpenCosmoPortal
Workflow	publish-cli.yml
Environment	pypi

Then create a pypi environment in GitHub repo settings → Environments.

Configuration

Config and tokens are stored in ~/.opencosmo/:

~/.opencosmo/
├── config.json          # Profiles (name → API URL)
└── tokens/
    └── <profile>.json   # OAuth tokens per profile