Deploy anything AI builds to a live URL. Works with Claude Code, Cursor, ChatGPT, Gemini, and 9 more coding agents. Free to start.

VibeHost is a hosting platform built for the way people actually code in 2026: half by hand, half by AI agent. The same CLI works whether a human is typing vibehost deploy, any of the nine first-class coding agents (Claude Code, Codex, Cursor, Antigravity, Antigravity CLI, Windsurf, OpenCode, Copilot CLI, Hermes Agent) is shelling out to it, or a chat client like ChatGPT is calling our MCP server.

What VibeHost is

A managed hosting platform that runs your built artifacts behind an HTTPS URL.

Static sites — any directory with index.html. Served directly from the edge.
Channels, not branches — preview deploys are named channels (production, dark-mode, pr-42). The server has zero git knowledge.

The 60-second mental model

flowchart LR
  subgraph You["You / your agent"]
      CLI[CLI]
      Agent[Claude / Cursor]
      SDK[REST / SDK]
  end
  subgraph CP["Control plane"]
      API[api.vibehost.com]
  end
  subgraph Edge["Edge runtime"]
      Router[Tenant router]
      Static[Static assets]
  end
  CLI -- Bearer --> API
  Agent -- OAuth 2.1 / PKCE --> API
  SDK -- PAT --> API
  API -.deploy.-> Router
  Router --> Static
  Static --> User((Visitor))

Three things to internalize from this picture:

Your machine builds. Our server just receives. No build steps run in the API hot path — those have killed hosting platforms before.
Every interface goes through the same REST API. CLI, agent (MCP), and dashboard all hit api.vibehost.com/api/v1 with the same auth checks. The same per-app permissions apply.
Tenant code never shares a process with us. Apps run in isolated edge sandboxes. Static is served straight from our asset storage — no app-specific code path at all.

What's different

hammer

Client builds, not server builds

npm run build runs on your machine. The server only accepts the artifact. No build steps in our API hot path.

box

Tarball deploys

CLI packs your dir into .tar.gz and POSTs it. Content-addressed blob dedup makes repeat uploads cheap.

terminal

Machine-readable everywhere

Every CLI command takes --json and emits {ok, data | error} with stable error codes. LLMs branch on code, not message.

key

Agent-first auth

OAuth 2.1 + PKCE for MCP clients, PATs for CI, sessions for browsers. Same per-app grants across all three.

How to drive VibeHost

VibeHost is built around coding agents — IDE / terminal agents that can both call our MCP tools AND shell out to our CLI for scripted control. Most users land through one of these:

Nine first-class integrations — Claude Code, Codex CLI, Cursor, Antigravity, Antigravity CLI, Windsurf, OpenCode, Copilot CLI, Hermes Agent. Each picks up the full VibeHost tool surface via our MCP server and can also run vibehost directly for scripts. See MCP server for the underlying details.

curl -fsSL https://vibehost.com/install.sh | sh, then vibehost login and vibehost deploy. The headless path — works from any shell, scriptable, what your coding agent shells out to under the hood. See the CLI guide.

ChatGPT and Claude Desktop don't have a CLI channel — they reach VibeHost only through MCP. Setup is one URL.

VibeHost Chrome extension adds a Deploy button inside Gemini Canvas, AI Studio, ChatGPT Canvas, Claude Artifacts, and v0.dev — one click from "I just generated this" to "here's a live URL". Popup can also snapshot any webpage.

https://api.vibehost.com/api/v1 with a PAT. Same backend everything else uses.

When to use VibeHost (and when not)

You want a URL fast and you don't want to think about build infra.
You're shipping AI-generated code and need agents to be able to deploy on your behalf.
You want to share previews per branch/PR/experiment without git hooks.
You need per-client / per-customer / per-feature isolated URLs.
You're hosting more than a handful of small apps (Vercel et al. price per project; we price per workspace).

You need a long-running stateful backend (databases, queues, persistent WebSockets). Pair VibeHost with a managed database / queue, or use a different platform.
You need server-side cron or arbitrary worker processes — we don't host those.
You need an egress-heavy media platform (video streaming). Our deploy model is artifact-based, not file-server.

Pricing in one paragraph

Free tier: unlimited static apps, up to 100 apps per workspace, *.vibehost.space URLs, OG card watermark on, no custom domains. Business ($20/month per workspace, not per app): custom domains, watermark off, password gates, share links, higher rate limits. Enterprise: anything above plus SLA, dedicated support, raised tarball caps. See pricing.

Next step

If you've never deployed to VibeHost: jump to the quickstart. It gets you from zero to a live URL in under two minutes.

If you're integrating an agent: jump to MCP or personal access tokens depending on whether you want OAuth or static bearer tokens.

If you came here from another platform: check recipes for migration walkthroughs and common end-to-end scenarios.

VibeHost — Hosting for Vibe-Coded Apps