harvis

Deploy the current folder to harvis.dev with one command. No account, no config.

pip install harvis
harvis

or without installing (with uv):

uvx harvis

Output:

Uploading 3 files (12.4 KB)...

  Live site:   https://happy-panda-482.harvis.dev/
  Claim link:  https://harvis.dev/claim/xxxxxxxx-...

The claim link is private and single-use: open it and sign in to
manage the site. Run `harvis claim` to open it in your browser.

Commands

Options

• --name <name> — set the site name (defaults to the page <title>)
• --new — create a fresh site even if this project is already linked to one
• --token <token> — deploy token (or set HARVIS_DEPLOY_TOKEN)
• --subdomain <sub> — deploy to this site (or set HARVIS_SUBDOMAIN)
• --claim — open the claim link in your browser right after deploying

Updating a site

The first deploy creates a site and records it in two places:

• harvis.json at your project root — the site name only, no secrets. Commit it, and every clone of the project deploys to the same site.
• ~/.config/harvis/credentials.json — the deploy token, stored per user with 0600 permissions. Because it lives outside the project, rebuilding or deleting your output folder never loses it. (%APPDATA%\harvis\credentials.json on Windows; honours XDG_CONFIG_HOME, or HARVIS_CONFIG_DIR to override.)

Every later deploy from the same project sends that token and updates the same site — before and after you claim it. Claiming attaches the site to your account; the token keeps working, and you can view or regenerate it in the dashboard.

The project root is found by walking up from the folder being deployed, looking for harvis.json, .git, package.json, pyproject.toml and friends — so harvis deploy dist links the project, not the build output.

To deploy an existing site from another machine, link the project once:

harvis link happy-panda-482 --token <token-from-dashboard>
harvis

In CI, set the environment instead of running link:

export HARVIS_DEPLOY_TOKEN=<token-from-dashboard>
export HARVIS_SUBDOMAIN=happy-panda-482   # optional if harvis.json is committed
harvis deploy dist

To start over with a fresh URL, run harvis --new.

Notes

• Uploads are zipped when the folder compresses well (HTML/CSS/JS shrink ~60-70%); folders that are mostly already-compressed media (images, video, fonts) upload as raw multipart instead. A progress bar is shown on interactive terminals.
• Limits: 500 files, 50 MB per site. Hidden files and folders (anything starting with ., like .env or .git), node_modules, and OS junk files are skipped automatically.
• Unclaimed sites expire 24 hours after the last deploy. Claim a site to keep it online permanently. If a linked site has expired, the next deploy tells you and creates a new one.
• The deploy token is a secret and stays out of your project: it is written only to ~/.config/harvis/credentials.json, mode 0600. The harvis.json file in your project holds nothing but the site name and is safe to commit.
• The claim link is single-use and can't be recovered. If you want to keep the site, claim it before it expires.

Development

This package is managed with uv:

cd python
uv sync
uv run harvis --help
uv build

About

This CLI is the open-source companion to harvis.dev. Deploys are anonymous by default; opening the claim link and signing in attaches the site to your account so you can manage it from the dashboard.

License

MIT