Domino CLI and R package

Applies to Domino 6.3 and Domino Cloud.

Product pages: https://docs.domino.ai/cloud/reference/cli (index, install, reference, download files, force-restore, move a project, proxy) and https://docs.domino.ai/cloud/reference/r-package/domino-r-package. The 6.3 pages are the same content under /6.3/.

What it is, and what it is not

The CLI is a Java application (the headless install needs a JRE); it keeps its project link in .domino/config.json and its proxy settings in domino.vmoptions.

Install and sign in

1. Account > Download CLI in the Domino UI, pick the OS, run the installer (macOS prompts for admin credentials; cancel and add /Applications/domino to PATH yourself if you lack them). Headless Linux: apt install openjdk-8-jre-headless, wget https://<domino-url>/download/client/unix -O /tmp/domino-install-unix.sh, sh /tmp/domino-install-unix.sh.
2. Close and reopen the terminal, then domino login https://<your-domino-url>. The CLI opens a browser to the Domino sign-in page and receives the authorization itself.
3. No browser on this machine (SSH, a Workspace)? domino login --device https://<your-domino-url> prints a verification URL and a short code; open it on any device, sign in, Grant Access. The code expires quickly; rerun the command for a new one.

A CLI older than 6.2.3 uses a retired authentication client and cannot log in; reinstall from Download CLI. After domino login the CLI keeps its own session; there is no token or API-key login for the CLI. Full install notes, OS support and proxies: CLI-INSTALL.md.

Everyday workflow (DFS project)

domino get owner/project          # first copy of an existing project; creates ./project
cd project
domino run train.py 20            # upload changed files, start a Job with arguments, return at once
domino run --wait train.py        # block until the Job ends
domino run --direct "pip freeze | grep pandas"   # shell command instead of a file
domino run --tier "Large" train.py               # hardware tier by name
domino download                   # pull results and collaborators' changes
domino sync                       # download then upload
domino upload -m "add data"       # push without starting a Job
domino download-results --filter 'results/*' 42  # files from run number or id 42 into results/run-42

File and run commands act on the project folder you run them in (the CLI keeps the link in .domino/config.json); domino get and domino create make that folder. domino init [name] turns the current folder into a new project without uploading; domino create [name] makes a new folder and project; domino restore reconnects a folder (for example a git clone) to its project. --owner <org> on create and init assigns the project to an organisation. --no-sync runs server-side code without uploading local changes; --local records a run that executes on your machine. Every command, flag and example: CLI-REFERENCE.md.

Datasets from the CLI

domino upload-dataset --fileUploadSetting Overwrite --targetRelativePath raw/2026-10 owner/project/dataset-name /local/folder
domino create-snapshot owner/project/dataset-name
domino create-dataset-from-snapshot owner/project/dataset-name 3 new-dataset-name

upload-dataset overwrites existing files unless --fileUploadSetting Ignore|Rename (values are case-sensitive). Tune big uploads with DOMINO_UPLOAD_CHUNK_BYTES (default 3 MB) and DOMINO_UPLOAD_THREADS (default 8). Quotas and limits: domino-datasets.

R package in one screen

install.packages("domino", repos = "http://cran.r-project.org", type = "source")
library(domino)
domino.get("owner/project")             # or setwd() into an existing project folder
domino.run("myScript.R", 1, "someString")   # arguments become the script's command-line args
domino.download()                       # bring results back, then load() them
domino.sync(); domino.init("name"); domino.upload("message")

CRAN domino 0.3.1 (2017) runs the installed CLI with system(), so install and log in with the CLI first (the browser or device flow above). Its domino.login(username, password, host) feeds a username and password to the CLI's old interactive prompt and may fail against the current CLI; prefer a terminal domino login. Exported functions, signatures and limits: R-PACKAGE.md.

Pitfalls

• The CLI syncs the Domino File System. On a Git-based project that is the Artifacts section; code goes through git and Jobs through the UI or API (domino-jobs).
• "blob file not found" on domino run: run domino sync first.
• More than about 1,000 files in one upload can time out; upload in batches.
• "too many open files" during sync: add an empty .noLock file at the project root.
• Keep files out of sync with .dominoignore at the project root; .git/ is always ignored.
• Broken local metadata: rm -rf .domino then domino restore (force-restore); files on the server that are missing locally come down, local changes go up on the next sync.
• Moving a project to another deployment copies only the latest files, never history or collaborator settings.

Related skills

• domino-projects: DFS versus Git-based projects
• domino-datasets: Datasets, snapshots, quotas
• domino-jobs: what a run becomes, logs, scheduling
• domino-workspaces: the dom SSH tool kit for Workspaces
• domino-python-sdk, domino-api-intro: API automation instead of the CLI