Upload a folder of static files — your build output or a downloaded workflow artifact — to harvis.dev from GitHub Actions, using the harvis npm package.

Deploy the build output straight from the job that built it:

name: Deploy

on:
  push:
    branches: [main]

jobs:
  deploy:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-node@v4
        with:
          node-version: 20
      - run: npm ci && npm run build
      - name: Deploy to harvis.dev
        uses: harvis-io/static-deploy-action@v1
        with:
          directory: dist
          token: ${{ secrets.HARVIS_DEPLOY_TOKEN }}

The deploy token is all you need: it identifies the site by itself, so every run updates the same one. Get it from the site's page in the harvis.dev dashboard and store it as a repository secret.

Omit token to get a fresh, anonymous site on every run — handy for pull request previews. Unclaimed sites expire 24 hours after the last deploy; the single-use claim link is exposed as the claim-url output (and kept out of the log, since whoever opens it owns the site).

When build and deploy are separate jobs, pass the output between them with actions/upload-artifact / actions/download-artifact, then point this action at the downloaded folder:

jobs:
  build:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-node@v4
        with:
          node-version: 20
      - run: npm ci && npm run build
      - uses: actions/upload-artifact@v4
        with:
          name: site
          path: dist

  deploy:
    needs: build
    runs-on: ubuntu-latest
    steps:
      - uses: actions/download-artifact@v4
        with:
          name: site
          path: site
      - name: Deploy to harvis.dev
        id: deploy
        uses: harvis-io/static-deploy-action@v1
        with:
          directory: site
          token: ${{ secrets.HARVIS_DEPLOY_TOKEN }}
      - run: echo "Deployed to ${{ steps.deploy.outputs.url }}"

A full copy-paste version lives in examples/build-and-deploy.yml.

• Limits: 500 files, 50 MB per site. Hidden files (anything starting with .), node_modules, and OS junk files are skipped automatically.
• The runner needs Node.js 18+ on PATH (all GitHub-hosted runners have it; on self-hosted runners add actions/setup-node first).
• Pinning cli-version below 0.0.6 breaks updates: those versions cannot resolve a site from the token alone and will create a new site on every run instead.
• The CLI records the site it deployed to in a harvis.json file at the project root (the site name only, no secrets). On a checked-out repository that leaves an untracked — or, if you commit harvis.json and the token names a different site, a modified — file in the workspace. Worth knowing if a later step is sensitive to a dirty working tree.
• token wins over a committed harvis.json: the action deploys to the site the token names, whichever site the file points at.
• On public repositories, workflow logs and PR comments are public — don't print claim-url or the deploy token anywhere.
• If a linked unclaimed site has expired, the CLI says so and creates a new site; check the url output rather than assuming the subdomain stayed the same.