For agents

Static site hosting for AI agents

Last updated: 2026-11-02

A coding agent runs npx flypod ./dist --json and gets back one line of JSON with a live URL in it. There is no account to create first, no API key to paste and no project to configure, so the agent can finish the job without stopping to ask you.

npx flypod ./dist --json

An answer an agent can parse

• Success is one line on stdout with ok: true, the url, the site_id, expires_at and any warnings, each with the fix to make.
• Failure is one line with ok: false, a stable code and a fix the agent can follow. Nothing is printed to stderr.
• The exit code follows the code: 0 ok, 1 runtime, 2 usage, 3 auth, 4 retry later. An agent can branch on it without reading prose.

One folder, one address

The first flypod in a folder creates the site. Every later run in that folder ships a new version to the same URL, and the receipt says "action": "updated". flypod rollback brings the previous version back, and flypod --new makes a second site on purpose.

Keeping what the agent deployed

A deploy made without an account lasts 14 days. To keep it, the agent runs flypod login --json and shows you the link it returns. Once you confirm in the browser, the agent's next flypod command finishes signing in and keeps every site that machine deployed. The agent does not wait or poll, and it never handles a password.

Teach your agent once

Run npx flypod install skill in the repo. It writes an Agent Skill to .claude/skills/ and .agents/skills/, which Claude Code, Codex, Cursor, Gemini CLI and Copilot read, so asking for "a link to this" leads to flypod. In a chat app with no terminal, add https://flypod.dev/mcp as a remote MCP server instead.

The full guide is Deploy from an AI agent, and the raw HTTP API is in the API reference.

What it does not do

flypod serves static files only: HTML, CSS, browser JavaScript, images and markdown. There is no server code, no database, no build step on the server and no custom domain. A deploy can be up to 10 MB and 200 files. The full list is on the limits page.

Questions

Can an AI agent deploy a website on its own?

Yes. An agent runs npx flypod ./dist --json and gets one line of JSON with the live URL. No account, API key or config file is needed first.

What does the agent get back?

One line of JSON on stdout: ok, url, site_id, expires_at, warnings and next steps. A failure is ok:false with a code and a fix the agent can run, and the exit code says whether to retry.

How does the person keep a site their agent deployed?

The agent runs flypod login --json and shows the person the link it returns. Once they confirm in the browser, the agent's next flypod command finishes signing in and keeps every site that machine deployed.

How long does an agent's deploy last?

A deploy made without an account lasts 14 days. Signed in, sites stay up until you delete them.

Can the agent update the same URL?

Yes. Running flypod again in the same folder ships a new version to the same address, and flypod rollback brings the previous one back.

Does it work from a chat app with no terminal?

Yes. Add https://flypod.dev/mcp as a remote MCP server and the chat app can deploy with the same rules as the CLI.

Related: publish from Cursor, deploy without an account, share an HTML file as a link, and what flypod can't do.