CLI vs the web

The web app (nemar.org) and the CLI are the two surfaces a person uses to do dataset work. They share the same backend, the same accounts, and the same datasets, but each is optimized for a different workflow.

This page compares those two in depth. NEMAR has three more surfaces that are not aimed at a person clicking or typing: the backend API, the Zarr serving copy, and the Model Context Protocol (MCP) server. For the whole set, and which one a given task calls for, see Which surface should I use?.

What the web is best for

• Signing up. Sign in with ORCID (Open Researcher and Contributor ID) or a one-time email code at nemar.org/login; a first ORCID sign-in creates your account as part of that flow, no separate form and no admin review. ORCID is the one identity root for both surfaces: the CLI’s nemar auth login / nemar auth signup open a browser to the same ORCID sign-in, at app.nemar.org/cli/authorize, and land on the same account either way. See Getting started on the web.
• Browsing and exploring. Discover, dataset detail pages, READMEs, and the citation dashboard live there.
• One-off uploads. Drop a folder, walk through validation, request publication. No tooling to install.
• Collaborator management. Inviting and removing collaborators happens on the dataset’s collaborators page.
• Admin review. The publication-request queue is web-only.

What the CLI is best for

• Scripted publishes. CI pipelines that publish a dataset version when a tag is pushed.
• Large / parallel uploads. The CLI uses git-annex parallel transfers for very large datasets that would be tedious to drop in a browser.
• Server-side workflows. Running on a compute node or a shared lab server where the web UI isn’t convenient.
• Inspection and version checks. Comparing local vs remote, viewing manifests, and so on.
• Bulk operations. Managing many datasets where shell scripting wins.

You don’t have to pick one

Many researchers use the web for sign-in, discovery, and publication review, then reach for the CLI when they need to script something. Same account, same datasets, same permissions; switching back and forth is fine.

Accounts and permissions

The CLI and the web share state through the same backend. Signing in on the web issues a cookie-backed session; signing in on the CLI opens a browser to that same sign-in and, once you confirm, issues an API key instead, named for the machine it was minted on. A laptop and a compute cluster can each hold their own live key for the same account at once; manage the whole set with nemar auth keys (see Authentication). Both surfaces honor the same admin role, dataset ownership, and collaborator permissions.

# Quick start with the CLI (see the full guide under CLI > Getting Started)
nemar auth login          # sign in with your browser
nemar dataset list --mine # show your datasets
nemar dataset upload ./my-dataset

See Installation for the full CLI setup.