Creek — Deploy to the edge

Global Flags

Every command supports these flags:

Flag	Description
`--json`	Output structured JSON (auto-enabled in non-TTY)
`--yes` / `-y`	Skip confirmation prompts (auto-enabled in non-TTY)
`--help`	Show help
`--version`	Show version

Commands

`creek dev`

Start a local development server with full Creek runtime (D1, KV, R2, realtime sync).

creek dev                       # Start on port 3000
creek dev --port 8080           # Custom port
creek dev --reset               # Clear local data before starting

What it runs:

Worker — Your Hono/Worker code via Miniflare (Cloudflare's local runtime)
Database — Local D1 (SQLite) with data persisted in .creek/dev/
Realtime — Local WebSocket server for useLiveQuery and usePresence
Vite HMR — Client-side hot reload (auto-detected)

All behind a single port. Matches production behavior.

creek dev

App:       http://localhost:3000
Worker:    worker/index.ts (D1)
Realtime:  ws://localhost:3000
Data:      .creek/dev/

Ready in 340ms

Supported project types:

Type	Worker	Vite	Realtime
Worker + SPA	Yes	Yes	Yes
Worker only	Yes	No	Yes
SPA only	No	Yes	No

SSR frameworks (Next.js, TanStack Start) use their own dev servers. Coming soon: creek dev proxy mode to unify SSR dev with Creek bindings.

`creek deploy [dir]`

Deploy a project or directory.

creek deploy                           # Auto-detect framework, build, deploy
creek deploy ./dist                    # Deploy a pre-built directory
creek deploy --template landing        # Deploy a ready-made starter
creek deploy --dry-run                 # Show the plan without executing
creek deploy --skip-build              # Skip the build step
creek deploy --json                    # Structured JSON output (agent-friendly)

Deploy from GitHub URL:

creek deploy https://github.com/user/repo          # Clone + detect + deploy
creek deploy https://github.com/user/repo#branch   # Specific branch
creek deploy github.com/user/repo --path apps/web   # Monorepo subdirectory

Behavior:

Has creek.toml + token → authenticated production deploy
Has package.json → auto-detect framework, build, sandbox deploy
Has dist/, build/, or out/ → deploy directory as sandbox
Nothing found → show guidance

Deploy to Creek Button

Add a one-click deploy button to any GitHub repo README:

[![Deploy to Creek](https://creek.dev/button.svg)](https://creek.dev/new?repo=https://github.com/YOUR/REPO)

Clicking the button takes visitors to a landing page with a ready-to-run deploy command. No account needed — deploys to a 60-minute sandbox preview.

`creek status [id]`

Check project or sandbox status.

creek status                    # Current project (from creek.toml)
creek status a1b2c3d4           # Sandbox status (no auth needed)
creek status --json

`creek projects`

List all projects.

creek projects
creek projects --json

`creek deployments`

List recent deployments for the current project.

creek deployments
creek deployments --project my-app
creek deployments --json

`creek logs`

Read recent log entries for a project. Console output (console.log/warn/error) and uncaught exceptions from every request your Worker handled are captured per-tenant and archived for 7 days.

creek logs                           # Last 1 hour (default)
creek logs --since 30m               # Last 30 min (relative: s/m/h/d)
creek logs --since 2026-04-13T18:00Z # ISO timestamp also accepted
creek logs --limit 500               # Up to 1000 per call
creek logs --json                    # ndjson, pipe to jq
creek logs --project other-app       # Not in cwd

Filter by what went wrong:

creek logs --outcome exception              # Only failed invocations
creek logs --outcome exceededCpu,exception  # Multiple outcomes (comma-separated)
creek logs --level error                    # Only entries with error-level console output
creek logs --search "TypeError"             # Substring match on messages, exceptions, URLs

Filter by deploy variant:

creek logs --deployment a1b2c3d4            # Specific deployment preview (8-hex id)
creek logs --branch feat-checkout           # Specific branch preview
creek logs --script-type production         # Exclude preview deploys

Live tail via WebSocket, until Ctrl+C:

creek logs --follow                         # Historical context + stream new entries
creek logs --follow --outcome exception     # Alert on errors only, live
creek logs --follow --json | jq '.outcome'  # Pipe to jq or any shell filter

Live tail prints the most recent historical entries first (so you don't stare at an empty terminal waiting for traffic), then streams new entries as they arrive. Dedup is automatic — you won't see the same entry twice across the historical → live boundary.

Output shape:

Each log line is a JSON object with the producer Worker's outcome, request metadata, and captured console/exception output:

{
  "v": 1,
  "timestamp": 1776108325010,
  "team": "acme",
  "project": "my-app",
  "scriptType": "production",
  "outcome": "exception",
  "request": { "url": "https://my-app-acme.bycreek.com/api/x", "method": "GET", "status": 500 },
  "logs": [
    { "level": "error", "message": ["TypeError: x is undefined"], "timestamp": 1776108325009 }
  ],
  "exceptions": []
}

Sensitive request headers (Authorization, Cookie, etc.) are redacted at ingest — Creek never stores them.

Authenticate with Creek.

creek login                     # Browser-based (default)
creek login --headless          # Paste API key manually
creek login --token <key>       # Non-interactive (CI/CD)
creek login --json

`creek init`

Create a creek.toml configuration file.

creek init
creek init --name my-app
creek init --json --yes

`creek claim <sandboxId>`

Convert a sandbox preview into a permanent project.

creek claim a1b2c3d4
creek claim a1b2c3d4 --json

`creek env`

Manage environment variables.

creek env set DATABASE_URL "postgres://..."
creek env ls                    # Values redacted by default
creek env ls --show             # Show plaintext values
creek env rm DATABASE_URL
creek env ls --json

`creek domains`

Manage custom domains. See Custom Domains for full guide.

creek domains add app.example.com   # Add a custom domain
creek domains ls                    # List all domains
creek domains activate app.example.com  # Manually activate
creek domains rm app.example.com    # Remove a domain
creek domains ls --json             # JSON output
creek domains add app.example.com --project my-app  # Specify project

`creek whoami`

Show the currently authenticated user.

creek whoami
creek whoami --json

Configuration

`creek.toml`

[project]
name = "my-app"
framework = "vite-react"     # Optional — auto-detected

[build]
command = "npm run build"
output = "dist"
worker = "worker/index.ts"   # Optional — enables creek dev + Worker deploy

[resources]
database = true     # D1
cache = false       # KV
storage = false     # R2

For AI Agents

Creek is designed for programmatic use:

Non-TTY auto-detection — pipes, CI, and agents automatically get JSON output
--json --yes — explicit structured output, skip all prompts
Exit codes — 0 success, 1 failure
Error format — {"ok": false, "error": "...", "message": "..."}

# These are equivalent in CI/agent:
creek deploy ./dist --json --yes
creek deploy ./dist              # auto-detects non-TTY

CLI Reference