Ship full-stack apps.
Backend included.

One command, zero config. Creek provisions the database, cron, queues, and WebSockets — use any ORM, or bring your own DB. Free to start.

no signup · live in seconds · Claude + Codex + Cursor ready

or import from GitHub

01

Realtime in 6 lines

WebSocket sync, optimistic updates, multi-user rooms. Zero boilerplate.

server.ts
import { db } from "creek";
import { room } from "creek/hono";

app.use("/api/*", room());

app.get("/api/todos", async (c) => {
  return c.json(
    await db.query("SELECT * FROM todos WHERE room_id = ?", c.var.room)
  );
});

app.post("/api/todos", async (c) => {
  const { text } = await c.req.json();
  await db.mutate(
    "INSERT INTO todos (room_id, text) VALUES (?, ?)",
    c.var.room, text
  );
  return c.json({ ok: true });
});

db.mutate() auto-broadcasts to all connected clients. No WebSocket code.

02

Your repo is the config

creek.toml is optional. Creek reads your framework config, package.json, wrangler files — or just an index.html — and works it out. Framework, bindings, build command, output dir: inferred from what's already in your repo. Add a creek.toml only when you need triggers or explicit overrides.

React (Vite)

$ creek deploy

Detected: vite-react

Building...

Output: dist/

Live in 8.3s

Vue (Vite)

$ creek deploy

Detected: vite-vue

Building...

Output: dist/

Live in 7.9s

Svelte (Vite)

$ creek deploy

Detected: vite-svelte

Building...

Output: dist/

Live in 6.4s

Astro

$ creek deploy

Detected: astro

Building...

Output: dist/

Live in 6.7s

03

Coming from Vercel or Netlify?

We grew up on those platforms. They taught us what great deploy DX looks like — git push, PR previews, framework detection, env vars in the dashboard. Creek keeps all of it. The bill is different. The license is different. Everything else should feel familiar.

Same workflow

  • git push deploys.
  • PR previews on every pull request.
  • Framework auto-detection.
  • Env vars in the dashboard.

Nothing new to learn. Your muscle memory transfers.

Same frameworks

  • Vite, Astro, SvelteKit, Nuxt.
  • Every Vite meta-framework, auto-detected.
  • Next.js via adapter-creek (WIP).
  • Static sites with one index.html.

See the Frameworks section below for the full maturity tiers.

Different bill

  • $0 when idle. Pay per request.
  • No per-seat fees.
  • No surprise invoices when traffic spikes.
  • Apache 2.0: self-host on your own account anytime.

The point of Creek. The cost shape is structurally different, not marginally different.

04

Your framework is supported

Vite, React, Vue, Svelte, Astro, TanStack Start, React Router, Hono, static sites — zero config. Next.js via our adapter (WIP). Every framework written up honestly below.

Zero-config

  • Vite + React
  • Vite + Vue
  • Vite + Svelte
  • Vite + Solid
  • Astro
  • TanStack Start

Detected from package.json + vite.config. Deploy with no flags.

Supported

  • SvelteKit
  • Nuxt
  • React Router (v7 / Remix)
  • Hono
  • Static site

Works today. Server build for SvelteKit / Nuxt is experimental.

Work in progress

  • Next.js

Routed through @solcreek/adapter-creek, currently via an OpenNextJS workaround.

05

Built for AI agents

Remote MCP server, JSON output on every command, installable agent skills, and the Agent Challenge protocol so verified agents skip CAPTCHAs. All shipping today.

MCP Config
{
  "mcpServers": {
    "creek": {
      "url": "https://mcp.creek.dev/mcp"
    }
  }
}

Add one line. Any AI agent can deploy. No CAPTCHAs — verified via Agent Challenge.

creek deploy --json
{
  "ok": true,
  "url": "https://a1b2.creeksandbox.com",
  "deployDurationMs": 9234,
  "mode": "sandbox"
}

Structured output. No parsing needed.

Agent Skills
$ npx skills add solcreek/skills
# Installs the "creek" skill into
# Claude Code, Cursor, Copilot,
# Gemini CLI, OpenCode, …

Agents get the full CLI reference, deployment modes, trigger schema, and troubleshooting tree — loaded on demand via the open Agent Skills standard.

Apache 2.0 · maintained at solcreek/skills

06

Edge-native performance

Your app runs on 300+ edge locations. Millisecond cold starts, global TTFB.

RegionCreek (edge)Origin server
Tokyo
12ms
280ms
Frankfurt
18ms
45ms
Sao Paulo
25ms
350ms
Sydney
15ms
310ms
US East
8ms
22ms

07

Open source. Low lock-in, honestly.

Apache 2.0. Self-host on your own Cloudflare account. We're honest: Creek is built on Cloudflare Workers, so you're ultimately locked to Cloudflare. The point is you're not locked to us — your code keeps working if you eject to raw wrangler, because Creek deploys standard CF primitives, not proprietary ones.

Apache 2.0

Use it, modify it, self-host it. Full source on GitHub.

Self-hostable

Run Creek on your own Cloudflare account with a single command.

Eject anytime

Creek deploys standard Cloudflare primitives. Your code runs on raw wrangler too.

BYO stack

Three modes: use Creek's built-in db, layer Drizzle or Prisma on Creek's auto-provisioned D1, or connect Neon / Supabase / any external database. Your data, your ORM, your choice.

01

10-second deploys

From CLI to live URL on the global edge.

02

GitHub auto-deploy

Push to main. Preview URLs on every pull request. Commit status in the diff.

03

Realtime sync

db.mutate() auto-broadcasts. useLiveQuery() auto-refetches.

04

Cron triggers

Schedule background jobs in creek.toml. No extra services.

05

Queues

Per-project queue, auto-provisioned. Producer + consumer wired up.

06

Per-tenant analytics

Requests, errors, p50/p99 latency. Cron execution log.

07

Custom domains

Automatic SSL. One CLI command to set up.

08

Environment variables

Encrypted at rest, injected at runtime.

09

Framework detection

React, Vue, Svelte, Astro, Solid — auto-detected.

Deploy in 10 seconds

No account needed. Try it right now.