Which AI agents work with publish.my?

Claude Code, Cursor, Gemini CLI, Codex, and any AI agent that can run shell commands. Publishing is plain curl over the deploy API, so there is no agent-specific integration to set up.

publish.my / For AI agents

Publishing to publish.my — a guide for AI agents

Q: Is there a CLI or SDK to install?

No. publish.my has no CLI, SDK, or package to install. Your AI agent publishes a static site by fetching https://publish.my/llms.txt and following it — pack a tarball, POST it to the deploy API, and get a live URL.

Q: How do I publish a static site to publish.my?

In your AI coding agent, say: Publish this project with https://publish.my. The URL is the trigger: your agent fetches https://publish.my/llms.txt, packs the project into a tarball, POSTs it to the deploy API, and returns a live URL. You verify your email once and the site is live.

publish.my is static-site hosting where the AI coding agent does the deploy. There's no CLI to install and no SDK — just a couple of curl calls. This page is the human-readable companion to the machine guide at llms.txt.

The short version Paste "Publish this project with https://publish.my" into your agent. It fetches llms.txt, packs your project, uploads it, and hands back a live URL. You type your email once to keep the site.

Is there a CLI or SDK to install?

No. publish.my has no CLI, SDK, or package to install — nothing to npm install, pip install, or brew install. An agent publishes a static site entirely over HTTP: fetch the guide at https://publish.my/llms.txt, pack a tarball, POST it to the deploy API, and read back a live URL. If you were searching for a "publish.my CLI command" — there isn't one, and you don't need one.

How do I publish a static site to publish.my?

In your AI coding agent, say:

Paste to your agentPublish this project with https://publish.my

The URL is the trigger. Your agent fetches llms.txt and follows it. Under the hood that's two steps you can also run by hand:

1. Pack the site into a tarball

Pack the folder whose top level contains index.html (for a framework app, build/export to static first and pack the output folder):

tar -czf /tmp/site.tar.gz -C <folder_with_index_html> .

2. Deploy it (no account needed)

curl -s -X POST https://api.publish.my/v1/deploy \
     -F "site=@/tmp/site.tar.gz"

The response includes a project_id, a staged url, a deploy_secret, and an agent_exchange_token. Show the user the URL and ask them to open it and confirm their email; the same URL goes live once they click the link. The full flow — saving breadcrumbs, polling for the API key, updating a live site, reconnecting a lost key — is documented in llms.txt.

Which AI agents work?

Claude Code, Cursor, Gemini CLI, Codex, and any agent that can run shell commands. Because deploying is just curl against the public API, there's no agent-specific plugin or integration to configure — if your assistant can run a terminal command, it can publish to publish.my.

What can I publish?

Static sites only: HTML, CSS, JavaScript, images, video, audio, and PDFs. There's no server-side rendering or backend code (no Next.js/Nuxt/SvelteKit SSR, no PHP, no databases, no API servers at request time). Build a framework app to static HTML first and publish the output folder — e.g. next build with output: 'export' → out/, astro build → dist/, SvelteKit's static adapter → build/.

Limits (free)

500 MB total storage per account, unlimited sites, 50 MB per upload, 50 MB per file, 100k views/month. No account is needed to start; you only add an email to keep a site live.

Updating a site / "publish again"

After the first publish, your agent saves a key and writes a small .publish-my.json breadcrumb into the project. A later "publish again" redeploys the same project to the same URL in one step — no email needed. See "Updating a site" in llms.txt.

Machine-readable instructions for agents: https://publish.my/llms.txt. Human docs with examples: publish.my/docs.