Skip to content

Tutorial: your first jarvy.toml

Time: ~5 minutes · You'll need: a terminal, an internet connection, a directory you can write to.

By the end of this tutorial you'll have:

  • Jarvy installed on your laptop
  • A working jarvy.toml describing a small Node.js + Python project
  • A reproducible setup that any teammate can run with one command

We'll build a config for a hypothetical service that uses Node 20 for the API and Python 3.12 for ML scripts.

1. Install Jarvy

curl -fsSL https://raw.githubusercontent.com/Cliftonz/jarvy/main/dist/scripts/install.sh | bash

brew install jarvy

cargo install jarvy

irm https://raw.githubusercontent.com/Cliftonz/jarvy/main/dist/scripts/install.ps1 | iex

Verify:

jarvy --version

You should see something like jarvy 0.1.0.

Already have brew, apt, or winget?

Jarvy uses your system's native package manager — it doesn't replace it. If brew install node already works for you, Jarvy makes a thin call to the same machinery.

2. Make a project directory

mkdir hello-jarvy && cd hello-jarvy

This is the directory we'll provision. Pretend it's a real repo — every developer who clones it will run the same commands.

3. Write jarvy.toml

Open jarvy.toml in your editor and paste:

jarvy.toml
[provisioner]
git    = "latest"
node   = "20"
python = "3.12"

[npm]
typescript = "^5.0"
prettier   = "latest"

[hooks.node]
post_install = "node --version && npm --version"

[env.vars]
NODE_ENV = "development"

What each block means:

Block What it does
[provisioner] The CLI tools to install. Versions accept "latest", major ("20"), or semver ranges (">=3.10").
[npm] Global npm packages installed after Node is on the machine. There are matching [pip] and [cargo] blocks.
[hooks.node] A shell script that runs after Node is installed. Use it to print versions, run setup scripts, or seed databases.
[env.vars] Project-scoped environment variables. Jarvy writes them to .env and your shell rc.

4. Preview before provisioning

jarvy diff

diff compares the config to the current machine and prints what will change. No installs happen yet — this is your dry run. It should show three tools to install (or zero if you already have them) and one hook to run.

Tip

jarvy setup --dry-run shows the same plan with the actual command sequence Jarvy will execute.

5. Provision

jarvy setup

You'll see colored progress lines:

✓ git 2.45.0 already installed
✓ node 20.11.0 installed via brew
✓ python 3.12.1 installed via pyenv
✓ npm: [email protected] installed
✓ npm: [email protected] installed
✓ ran hook for node
✓ wrote .env
Setup complete in 14.3s

Tools that are already on the machine are detected and skipped — Jarvy is idempotent, so re-running jarvy setup is always safe.

6. Verify

jarvy doctor

doctor walks through every tool in your config and confirms it's installed at a satisfying version, and that the .env matches.

If anything is off, you'll see something like:

✗ node: expected ^20, found 18.17.0
  fix: jarvy setup

Run the suggested fix and you're back to baseline.

What just happened

You wrote a single file. Jarvy:

  1. Detected your OS and picked the right package manager (Homebrew on macOS, apt/dnf on Linux, winget on Windows).
  2. Resolved versions and skipped tools that already met the requirement.
  3. Installed missing tools, then global npm packages.
  4. Ran the post-install hook.
  5. Wrote a .env file matching [env.vars].
  6. Captured a snapshot of the result so drift detection can flag changes later.

Anyone else who clones the repo runs jarvy setup and gets the exact same machine state.

Next steps