Deploy with zero friction
Made for agents to assist humans.
Drop one prompt into Claude Code, Codex, Cursor, or any AI coding agent. It reads the project, picks up the specs, and ships the deploy.
$ claude # or codex · cursor — your AI assistant › install github.com/ImpossibleFinance/impossible-hosting-skill and deploy this project to ifhost agent reads your project, installs what it needs, picks the right size, and ships it live — nothing for you to type. → live at my-api.host.impossi.build
That's it. The skill picks RAM, region, and scale; writes the config; runs the deploy.
A new UX for code
For decades, deploys meant humans. ifhost flips it. No terminal. No dashboard. No CLI to learn. You say "ship it" — it ships. This is what coding should look like.
Things you can build
Three little apps, all live right now. Click one to try it, then grab the code.
All three are open source in one repo → impossible-hosting-demo
…or host an AI agent that never sleeps
Telegram & Discord bots, autonomous agents, schedulers — anything that needs to stay awake. Bring your API key or bot token; ifhost keeps it running.
$ claude # or codex · cursor — your AI assistant › install github.com/ImpossibleFinance/impossible-hosting-skill and host hermes-agent on ifhost agent sets it all up, then asks you for the bits only you have — your API key, Telegram bot token, chat ID — and wires them in. ✓ running 24/7 at hermes-agent.host.impossi.build
Get started
3 steps, 2 minutes.
-
Install the CLI$ curl -fsSL https://host.impossi.build/install | sh
-
Sign in with Google$ ifhost loginOpens your browser to pick your Google account. Or grab a token from the dashboard.
-
Tell your agent to deployinstall github.com/ImpossibleFinance/impossible-hosting-skill
and deploy this projectOpen your project's folder in Claude Code, Codex, Cursor, or any AI coding agent first (no project yet? ask it to build one) — then paste that prompt. "This project" is whatever folder you have open. The skill teaches it the full workflow — it picks specs, writes config, and deploys in 2-4 min. What the agent learns →
Works for
Whatever your agent can install — these are the common shapes.
Static sites
Portfolios, docs, landing pages, SPAs. Served from a tiny nginx container, with auto-scaling and a free *.host.impossi.build URL.
SSR web apps
Next.js, Remix, SvelteKit, Astro, Rails, Django, anything with a start script, or let your agent write one.
APIs & backends
REST, GraphQL, webhooks, workers, long-running jobs. Any language, any runtime. ifhost deploys it.
AI agents & bots
Telegram, Discord, Slack, scrapers, LLM gateways. ifhost deploy --runner boots a shell; your agent installs through the console.
Stateful services
SQLite-backed tools, embedded caches, single-node databases, session stores. storage = "local" gives a persistent /data that survives redeploys.
Cron & always-on workers
Nightly reports, polling loops, schedulers, queue consumers. Pin min-machines = 1 + autostop = false and the worker never sleeps.
Built for agents to assist humans
Every surface is shaped for the coding agent.
Agent-first CLI
Every flag is declarative. Every command returns structured JSON with --json. No prompt gets in the way when args are fully specified. Your agent never has to guess a default or block on stdin.
Interactive setup via console
Run install wizards, config prompts, and long-lived daemons inside a tmux-backed session your agent drives through console input/output.
Persistent /data
storage = "local" auto-creates a volume that survives redeploys. Perfect for SQLite, session history, agent memory.
Auto-scaling
Scale to zero when idle, scale up under load, or pin machines for zero-cold-start daemons.
Secrets, not env vars
Pass sensitive values with --secret KEY=VALUE or machines secrets set. Injected at runtime, never stored in our DB.
Persistent disk, on demand
Need to keep data between deploys? Set storage = "local" and you get a 3 GB /data volume. Perfect for SQLite, file caches, or anything single-machine.
Simple pricing
Start free. Scale when you need to.
Free
- 1 app
- 256 MB RAM pool
- 1 GB volume pool
- 1 shared CPU
- 100 GB bandwidth
- Scale to zero
Hobby
- Unlimited projects
- 2 GB RAM pool
- 5 GB volume pool
- 2 shared CPUs/app
- 100 GB bandwidth
- 3 custom domains
Pro
- Unlimited projects
- 8 GB RAM pool
- 20 GB volume pool
- 4 shared CPUs/app
- 500 GB bandwidth
- 10 custom domains
Team
- Unlimited projects
- 24 GB RAM pool
- 50 GB volume pool
- 8 shared CPUs/app
- 1 TB bandwidth
- Unlimited domains
Pool = total across all your machines. Spend it on one fat machine or many small ones. Volume above pool: $0.30 / GB·month. Egress above pool: $0.05 / GB (NA/EU), $0.08 (APAC), $0.18 (Africa/India). No surprise shutoffs — set a spend cap with ifhost billing alert.
Questions people ask
If you don't see yours here, the docs go deeper.
What is ifhost?
It's a place to put your code online without learning a hosting tool. You tell an AI coding assistant — Claude Code, Codex, Cursor, or another one — to deploy your project, and it handles the setup. What you get back is a URL anyone can visit.
Do I need to be a developer to use it?
No. If you can describe what you want and let an AI assistant build it, you can ship it on ifhost. The assistant handles the parts that usually need a terminal — picking specs, writing config, running the deploy.
Which AI assistants does it work with?
Claude Code, Codex, and Cursor are the ones we've tested most. The deploy instructions are two short lines of plain English, so any assistant that can run a shell command should be fine.
What kinds of projects can I deploy?
Static sites, full web apps, APIs, AI agents, Telegram bots, cron jobs, and small databases. If it runs on a Linux machine, it'll run here.
How is ifhost different from Vercel, Netlify, or Heroku?
Those are tuned for developers using a dashboard or git push. ifhost is tuned for one workflow: an AI assistant doing the work for you. There's no dashboard to navigate, no config to memorize — the assistant picks specs, writes the config, and ships. If you already know your way around a hosting dashboard, you may prefer one of those. If you'd rather just ask an assistant, this is for you.
Is there a free plan?
Yes. One app, 256 MB of RAM, 1 GB of storage, 100 GB of bandwidth a month. It scales to zero between requests.
Can I cancel anytime, and what happens if I do?
Yes. Cancel from the dashboard and you stay on your paid plan until the end of the current billing period, then drop back to free. Your apps and data aren't deleted — they just run within the free limits from there on.
Do you accept crypto payments?
Yes — USDC on Base. The plans are the same either way. Pick crypto at checkout, and your assistant can also handle the upgrade for you when you ask it to.