One-prompt root-Linux hosting for Claude Code / Cursor / Codex. Free forever. Tell your agent: “Fetch thevibehosting.com/SKILL.md and deploy it there” — it gets a Linux box and your project goes live. No MCP needed.
# no signup, no key, no MCP — your agent just runs this$ curl -sX POST https://thevibehosting.com/create↳ returns an SSH command (root), a live https://abc123.thevibehosting.com, and your api_key (vibe_…).
That's the whole instruction. Your coding agent — Claude Code, Codex, anything — takes it from there. No dashboard, no signup, no setup for you.
$ curl -sX POST https://thevibehosting.com/create↳ root SSH · api_key. Then it SSHes in and ships.
One POST /create — no signup, no key. An api_key is issued on the first call.
The response hands back a root SSH command and a live HTTPS subdomain (valid TLS, auto-issued).
ssh root@… · https://abc123.thevibehosting.com
It's a real Ubuntu box — apt install, run a DB, a worker, a bot. Serve on :80 from /root/www or /root/run.sh.
apt install … → run your app on :80
By default a server sleeps when idle (RAM freed, wakes in 1–3 s) — still reachable 24/7, and it barely touches your free credit. Need zero cold starts? create it with {"never_sleep":true} — billed by real RAM 24/7.
No dashboard to babysit. List, rename, expose a port, sleep/wake, check live usage & billing — all over plain REST. GET /account/billing shows your balance, burn rate and runway.
A genuinely free tier, then pay only for what runs. RAM is billed only while the server is awake — CPU is always free.
Anonymous & instant — great for a quick deploy. A small app that sleeps when idle stays free.
One SMS unlocks the real free tier — 2.5× the free credit, bigger servers, as many as you want. A 512 MB server runs 24/7 all month free.
Only past your free credit, and you pay for the RAM you actually use: a 1 GB box 24/7 ≈ $1.35/mo after credit, 2 GB ≈ $5, 4 GB ≈ $12 — anything that sleeps when idle is ~free. From prepaid credit.
no plans · set {"never_sleep":true} · first $2.25/mo free, then billed by real RAM · + disk $0.25 / GB-mo
Sleeping or within your free credit → free; you only pay for RAM actually used while awake. Top up any time via POST /account/recharge. Verifying a phone (free) unlocks the bigger free credit + never-sleep.
A real session, replayed: one call for root + TLS, a file into /root/www, a 200 from your subdomain. Then it's your call — let it sleep, or keep it always-on.
Missing an endpoint, a base image, a higher limit? Send it — we read every one. Add an email if you'd like a reply.
# what would make this better for you?$ curl -sX POST https://thevibehosting.com/feedback -d '{"kind":"feature","message":"…","email":"you@x.com"}'A real root Linux server (Ubuntu) on its own HTTPS subdomain, created with one curl — no dashboard to babysit. The twist: you choose whether it stays always-on or sleeps when idle to save credit.
No signup, no API key to configure, no MCP to install. You get root in seconds and an api_key (vibe_…) is issued automatically on the first POST /create call. Save it — it's your account credential.
Your choice, per server. By default it scales to zero when idle (RAM freed, wakes in seconds) so you pay only for awake time. Or keep it always-on. Because RAM is billed only while awake at $0.005/GB-hour, your monthly free credit ($2.25 verified) runs a 512 MB server 24/7 for a whole month free — bigger servers can sleep to stay free, or stay pinned and pay for the RAM they use (1 GB 24/7 ≈ $1.35/mo after the credit).
Create it with {"never_sleep":true} (verified phone). It never scales to zero and has no cold starts — billed by the RAM it actually uses, 24/7, at $0.005/GB-hour (e.g. 1 GB ≈ $3.60/mo gross, ≈ $1.35 after the free credit) from prepaid credit you top up via POST /account/recharge. No plans.
Each project runs in a gVisor-sandboxed container with hard CPU/RAM limits, fully isolated from the host and every other tenant. Outbound SMTP is blocked and abusive CPU/egress is auto-suspended.
That's the whole idea. Point it at curl -s https://thevibehosting.com/SKILL.md for the full end-to-end guide, or read /llms.txt. This page is just the friendly mirror of it.