Skip to content

Getting Started

Get a research workstation running in under 5 minutes.

1. Install Prism

brew install scttfrdmn/tap/prism

scoop bucket add scttfrdmn https://github.com/scttfrdmn/scoop-bucket
scoop install prism

Download the latest release from the releases page, extract, and add to your PATH.

Verify: 

prism version

2. Authenticate with AWS

Prism uses your AWS credentials to launch instances in your account.

Step 1 — Log in (requires AWS CLI v2.32+):

aws login

This opens a browser window. Sign in with your IAM user or federated identity. Credentials are cached for up to 12 hours and refresh automatically.

No browser? Use aws login --remote for cross-device authentication, or aws configure to set up long-term access keys (see AWS Setup Guide).

Verify it works: 

aws sts get-caller-identity

Step 2 — Add a Prism profile:

prism profile add

This interactive wizard links a Prism profile to your AWS credentials and default region. You only need to do this once.

3. First-time wizard (optional)

If this is your first time running Prism, prism init covers credential validation, profile creation, and a test launch in one flow:

prism init

4. Launch a workspace

# See what's available
prism templates

# Launch a Python + Jupyter environment (~2 minutes)
prism workspace launch python-ml my-project

# Check it's ready
prism workspace list

5. Connect

prism workspace connect my-project

Prints the SSH command and any web service URLs (Jupyter, RStudio, etc.).

6. Stop when done

prism workspace stop my-project

Stopped workspaces have no compute cost. Resume later with prism workspace start my-project, or delete permanently with prism workspace delete my-project.

Hibernation preserves RAM state and reduces cost further: 

prism workspace hibernate my-project
prism workspace resume my-project

Common templates

Template What's included
python-ml Python, PyTorch, TensorFlow, Jupyter
r-research R, RStudio Server, Bioconductor
genomics GATK, BWA, samtools, STAR
bioinformatics Conda, Snakemake, BioPython
deep-learning CUDA, cuDNN, GPU-ready PyTorch
data-science Pandas, scikit-learn, DuckDB
hpc-base MPI, OpenMP, GCC, CMake

Run prism templates info <name> for details on any template.

Two interfaces

CLI

The prism command is the primary interface — scriptable and pipeline-friendly: 

prism workspace launch python-ml my-project --size L --spot --ttl 8h

GUI (desktop app)

A visual desktop app for managing workspaces, storage, and settings: 

prism gui
Or launch from your Applications folder (macOS) / Start menu (Windows).

Common workflows

Data science

prism workspace launch python-ml data-analysis --size L
prism workspace connect data-analysis
# Jupyter available at http://localhost:8888 via SSH tunnel

R / statistics

prism workspace launch r-research stats-project
prism workspace connect stats-project
# RStudio Server at http://localhost:8787 via SSH tunnel

Shared storage (EFS)

prism volume create shared-datasets
prism volume attach shared-datasets my-project

Time-limited workspace

prism workspace launch python-ml my-project --ttl 8h --dns my-ws
# Auto-stops after 8 hours; accessible at my-ws.abc123.prismcloud.host

Troubleshooting

"Daemon not running" 

prism admin daemon status
prism admin daemon stop     # Then run any prism command to auto-restart

"AWS credentials not found" 

aws sts get-caller-identity   # Verify credentials are valid
aws login                     # Re-authenticate if expired

Instance launch fails 

prism workspace launch python-ml my-project --region us-east-1  # Try different region

For more, see the Troubleshooting Guide.

Next steps