The heximon CLI

Building and running a Heximon app in development is pnpm dev — no CLI needed. The heximon package is a separate, unscoped install for the parts that bracket everyday development: starting a new project, shipping a build to a real platform, provisioning the resources that build needs an id for, and driving migrations without a hand-written drizzle.config.ts. It ships exactly four commands.

pnpm create heximon my-app     # day zero — nothing installed yet
pnpm add -D heximon            # in an existing project
pnpx heximon deploy

npm create heximon@latest my-app     # day zero — nothing installed yet
npm install -D heximon               # in an existing project
npx heximon deploy

yarn create heximon my-app     # day zero — nothing installed yet
yarn add -D heximon            # in an existing project
yarn dlx heximon deploy

bun create heximon my-app     # day zero — nothing installed yet
bun add -d heximon            # in an existing project
bunx heximon deploy

The command surface

That's the whole surface. There is no heximon dev, heximon build, or heximon serve — the dev server and the build step are the vp tasks (vp dev / vp build) your vite.config.ts already runs via the heximon() Vite plugin. The heximon CLI brackets those: create before your first pnpm dev, and deploy at "ship this to a platform."

Every command shares a few flags

--cwd (project root, default .) and --env (deploy environment — production / staging / preview:<id>, deploy/provision only) work the same way everywhere.

Beyond that, deploy and provision read your heximon.config.ts config file to resolve which platform you're targeting — a --platform cloudflare flag overrides it when your config doesn't declare one — and migrations reads your drizzle config(s) directly, no platform involved.

Interactive or scripted, either way

Every command renders through clack prompts when your terminal is a TTY — a missing credential is an interactive question, a multi-database migrations run is a select menu.

In CI (no TTY), the same commands run non-interactively: a missing required credential halts with the env var name to set instead of prompting, and a multi-database migrations run defaults to acting on all of them. Nothing about the underlying steps changes between the two modes — only how a decision gets made.

No embedded infrastructure engine

heximon deploy and heximon provision don't reimplement Terraform. They resolve your credentials, run vp build, then hand off to the platform's own CLI — wrangler deploy, vercel deploy --prebuilt, netlify deploy, deployctl deploy — capturing its output and reporting it back through clack. That's the whole ship step: native-CLI handoff, not a second deploy tool to learn.

See also

• Deploy a Heximon app — heximon deploy's full flag reference: --provision, --create, --migrate, --profile, --verify, --destroy.
• Migrations from the CLI — generate / migrate / status / push without a hand-written drizzle config.
• Database migrations — the MigrationRunner / MigrationConfig mechanism the CLI drives at the drizzle-kit layer, and the injectable runner that applies migrations at boot.