Skip to content

solcreek/creek

BranchesTags

Repository files navigation

Creek

Open-source deployment platform built on Cloudflare Workers.

npm SDK License

One command deploys your full-stack app — frontend, SSR, database, cron, queue — to 300+ Cloudflare locations worldwide. Self-host on your own account, or use Creek Cloud.

npx creek deploy

What you get

  • One-command deploys — auto-detects your framework, builds, and deploys
  • Built-in resources — D1 database, R2 storage, KV cache, AI, Queues — provisioned per project
  • Cron + Queue triggers — declare in creek.toml, Creek wires everything up
  • Realtime sync — D1 writes auto-broadcast to connected clients via WebSocket
  • Per-tenant analytics — requests, errors, latency, cron execution log
  • Agent-first — JSON output, breadcrumb hints, MCP server, no CAPTCHAs
  • Open source — Apache 2.0, self-host on your own Cloudflare account

Zero config, one command

cd my-vite-app
npx creek deploy
#  Detected: creek.toml (Vite + React + D1 + KV)
#  Building...
#  Deploying to edge...
#  ⬡ Deployed! https://my-vite-app-myteam.bycreek.com

Creek detects your framework, provisions resources, builds, and deploys.

Supported: React · Vue · Svelte · Solid · Astro · VitePress · Hono · TanStack Start · React Router · Next.js · static HTML WIP: Nuxt · SvelteKit

Try it — deploy these in ~15 seconds

No signup, no config. Click a badge, watch it go live.

Repo Stars What you'll see
Deploy to Creek satnaing/astro-paper ⭐ 4.4k Minimal, SEO-friendly Astro blog — the most popular Astro theme
Deploy to Creek withastro/docs ⭐ 1.6k Astro's official documentation site
Deploy to Creek vuejs/docs ⭐ 3.2k Vue 3's official documentation — VitePress
Deploy to Creek vitejs/vite /docs ⭐ 58k Vite's official documentation — VitePress in a monorepo subpath
Deploy to Creek sanity-io/sanity-template-astro-clean ⭐ 173 Astro + Sanity starter — headless CMS pipeline
Deploy to Creek danielcgilibert/blog-template ⭐ 995 Openblog — elegant, SEO-optimised Astro blog

First click of any repo runs a full build (~60–90s). Every click after the first hits the build cache and deploys in ⚡ ~7s — the repo only rebuilds when someone pushes a new commit.

Add it to your own repo's README:

[![Deploy to Creek](https://creek.dev/deploy.svg)](https://creek.dev/deploy/gh/YOUR-ORG/YOUR-REPO)

Supports GitHub tree/{branch}/{path} URLs for monorepos — point at a subfolder the same way you'd link to it on GitHub.

Features

Realtime in 6 lines

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

// Server
import { db } from "creek";
import { room } from "creek/hono";

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

app.post("/api/todos", async (c) => {
  // db.mutate() auto-broadcasts to all connected clients
  await db.mutate("INSERT INTO todos (room_id, text) VALUES (?, ?)",
    c.var.room, await c.req.json().then(b => b.text));
  return c.json({ ok: true });
});
// Client
import { LiveRoom, useLiveQuery } from "creek/react";

function App() {
  const { data: todos, mutate } = useLiveQuery("/api/todos");
  // Auto-refetches when data changes. Optimistic updates with auto-rollback.
}

Cron + Queue triggers

# creek.toml
[project]
name = "my-app"

[resources]
database = true

[triggers]
cron = ["0 */6 * * *"]   # Every 6 hours
queue = true              # Auto-provisions a CF Queue
// worker/index.ts
import { db, queue } from "creek";

export default {
  async fetch(request, env) {
    await queue.send({ type: "process", id: "123" });
    return new Response("queued");
  },
  async scheduled(event, env, ctx) {
    // Runs every 6 hours
    await db.prepare("DELETE FROM stale WHERE created < ?").bind(Date.now() - 86400000).run();
  },
  async queue(batch, env, ctx) {
    for (const message of batch.messages) {
      // Process message
    }
  },
};

Per-tenant analytics

Built-in dashboard shows requests, errors, CPU latency (p50/p99) for the last 24h / 7d / 30d. Cron execution log included.

Built-in resources

[resources]
database = true   # D1 (SQLite at the edge)
storage = true    # R2 (object storage, zero egress)
cache = true      # KV (key-value)
ai = true         # Workers AI

Creek provisions per-tenant resources via the Cloudflare API. Each project gets its own isolated D1 database, R2 bucket, KV namespace.

Built for AI agents

# JSON output auto-enabled in non-TTY / CI
npx creek deploy --json
{
  "ok": true,
  "url": "https://my-app-myteam.bycreek.com",
  "deploymentId": "a1b2c3d4-...",
  "cron": ["0 */6 * * *"],
  "breadcrumbs": [
    { "command": "creek status", "description": "Check deployment status" }
  ]
}
  • No CAPTCHAs — Agent Challenge protocol issues verified agent tokens (60/hr vs 10/hr)
  • Breadcrumb hints in error responses guide agents to next steps
  • MCP server at mcp.creek.dev (in development) — any AI agent can deploy with a single tool call

Open source

Apache 2.0 licensed. The entire platform is open source — CLI, control plane, dashboard, runtime, sandbox, dispatch worker.

Self-host on your own Cloudflare account, or use Creek Cloud (managed).

git clone https://github.com/solcreek/creek.git
cd creek && pnpm install
# Copy wrangler.*.example → wrangler.*, fill in your CF account
pnpm --filter @solcreek/control-plane deploy

Enterprise governance features (SSO, approval workflows, policy engine) will be available under /ee in a separate license.

CLI Reference

creek deploy [dir]                  Deploy project or directory
creek deploy --template landing     Start from a ready-made Vite + React template
creek deploy --dry-run              Show the deploy plan without executing
creek deploy --json                 Output structured JSON (auto in CI)
creek deploy --skip-build           Deploy without running build step
creek dev                   Local dev server with D1/KV/R2 simulation
creek status                Project status, triggers, deployment info
creek deployments           List deployment history
creek logs                  Read recent log entries (R2 archive)
creek logs --follow         Live tail via WebSocket until Ctrl+C
creek logs --deployment <id> --outcome exception   Filter by deploy / outcome / branch / level / search
creek rollback              Rollback to a previous deployment
creek env set <key> <val>   Set an environment variable
creek env ls                List environment variables
creek domains add <host>    Add a custom domain
creek login                 Authenticate with Creek
creek init                  Create creek.toml configuration
creek claim <sandboxId>     Convert sandbox preview to permanent project

How it works

                 CLI / API / MCP
                       |
          +------------+------------+
          |                         |
    Sandbox API               Control Plane
   (no auth, 60min)         (auth, permanent)
          |                         |
          +------------+------------+
                       |
            Workers for Platforms
           (dispatch namespace)
                       |
              +--------+--------+
              |        |        |
           Worker   Static    D1/R2/KV/Queue
           Script   Assets    (per-tenant)
                       |
                Cloudflare Edge
               (300+ locations)

Sandbox path: creek deploy / creek deploy --template <name> → Sandbox API → live URL (60 min, no auth) Production path: creek deploy → Control Plane → permanent URL + custom domain

Project structure

creek/
  apps/
    dashboard/              Vite + React + TanStack Router (app.creek.dev)
    www/                    Next.js marketing site via adapter-creek (creek.dev)
  packages/
    cli/                    CLI — npm: creek
    sdk/                    TypeScript SDK — npm: @solcreek/sdk
    runtime/                creek runtime for deployed apps — npm: @solcreek/runtime
    deploy-core/            Shared WfP deployment logic
    control-plane/          Hono API + Better Auth + D1 (api.creek.dev)
    sandbox-api/            Public sandbox deploy API
    sandbox-dispatch/       Sandbox routing + banner injection
    dispatch-worker/        Production tenant routing
    realtime-worker/        WebSocket via Durable Objects
    remote-builder/         CF Containers remote build orchestrator
    build-container/        Build environment Docker image
    ui/                     Shared UI (shadcn v4 + Base UI)
    create-creek-app/       Template scaffolder — npm: create-creek-app
    mcp-server/             MCP server (mcp.creek.dev)

Tech stack

  • Runtime: Cloudflare Workers
  • Framework: Hono (API), Vite + React (dashboard), Next.js (www)
  • Database: Cloudflare D1 (SQLite) + Drizzle ORM
  • Storage: Cloudflare R2
  • Queue: Cloudflare Queues
  • Auth: Better Auth (GitHub, Google, email/password, API keys)
  • Multi-tenancy: Workers for Platforms (dispatch namespaces)
  • Monorepo: pnpm workspaces + Turborepo
  • Testing: Vitest (790+ tests)

Self-hosting

Creek is designed to run entirely on a single Cloudflare account.

Prerequisites

  • Cloudflare account with Workers for Platforms enabled
  • Node.js >= 18, pnpm >= 9
git clone https://github.com/solcreek/creek.git
cd creek
pnpm install

# Configure: copy each wrangler.*.example to wrangler.* and fill in
# your Cloudflare account ID, dispatch namespace, and resource bindings

# Deploy workers
pnpm --filter @solcreek/control-plane deploy
pnpm --filter @solcreek/sandbox-api deploy
pnpm --filter @solcreek/sandbox-dispatch deploy
pnpm --filter @solcreek/dispatch-worker deploy

See Self-Hosting Guide for detailed instructions.

Development

pnpm install
pnpm test            # Run all tests (790+)
pnpm typecheck       # TypeScript checks across all packages
pnpm dev             # Start all dev servers

Next.js Adapter

Creek ships its own Next.js adapter — @solcreek/adapter-creek — purpose-built for Cloudflare Workers for Platforms.

npm install @solcreek/adapter-creek
creek deploy

The CLI auto-detects Next.js >= 16.2.3 and invokes the adapter. No config needed.

What it does differently from generic adapters:

  • Streaming SSR via TransformStream — compatible with Workers' request lifecycle (no node:http server dependency)
  • Embedded manifests — all Next.js route/build/prerender manifests are inlined into the worker bundle via a custom fs shim, so no filesystem reads at runtime
  • Node.js API shims — http, https, net, inspector, fs, vm, process EventEmitter stubs — purpose-built for the APIs Next.js actually touches, not generic polyfills
  • Turbopack runtime patching — replaces dynamic R.c() chunk loading with a static requireChunk() switch for Workers' no-filesystem constraint
  • Zero config — the adapter hooks into Next.js's onBuildComplete lifecycle and produces a .creek/adapter-output/ bundle the CLI uploads directly

Dogfooding: creek.dev is a Next.js app — its source lives in apps/www in this repo — and it's deployed on Creek's own infrastructure via creek deploy --yes. Same platform, same adapter, same pipeline that every other Creek user gets. If the adapter breaks, our own marketing site breaks first.

Ecosystem

  • Awesome Creek — Curated list of templates, integrations, guides, and community projects. Submissions welcome.
  • Templates Gallery — Live previews of official starter templates.

Contributing

We welcome contributions. Please see CONTRIBUTING.md for guidelines.

License

Apache 2.0 — see LICENSE.

creek.dev · Docs · Templates · Discord

Built by SolCreek.

About

Open-source edge deployment platform. Deploy full-stack apps to Cloudflare Workers in seconds.

creek.dev

Topics

react open-source typescript serverless mcp nextjs realtime paas cloudflare ai-agents edge-computing deployment-platform cloudflare-workers durable-objects workers-for-platforms

Resources

Readme

License

Apache-2.0 license
Activity
Custom properties

Stars

2 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

50 tags

Packages

 
 
 

Contributors

Languages