The vibiz CLI cheat sheet: every command you will use in week one
A complete cheat sheet of every vibiz CLI command you need in week one to authenticate, configure environments, manage databases, and deploy.
On this page 8 sections
The vibiz CLI gives you a single command-line tool to authenticate your machine, provision projects, manage isolated development databases, deploy sites, and configure AI agents. You run it directly with npx @vibiz/cli on Node 20+ or install it globally with npm install -g @vibiz/cli. Every command maps straight to the underlying HTTP API using the key saved during login, bridging your local shell and AI coding agents.
This guide collects the core vibiz commands you will use in your first week, organized by task.
Installing and authenticating the CLI
You do not need a separate account or login per service. Authentication creates an API key tied to your computer.
# Log in through your browser; opens an authorization window
vibiz login
# Log in without opening a browser (prints the URL and code)
vibiz login --no-browser
# Log in using an existing secret key directly
vibiz login --key vz_sk_123
# Check which account and key the current machine is using
vibiz whoami
# Forget the local key and revoke it if generated by login
vibiz logoutWhen you run vibiz login, the CLI prints an authorization link and waits up to 10 minutes. Opening the link lets you approve the CLI with either Full access (pre-selected) or Limited access. A Limited key can create projects, run deploys, send emails, and use AI, but it is blocked from production secrets and live money movements. If you have no account yet, create one at thelastapi.dev/start first.
Initializing and configuring projects
Initializing a project writes a ./vibiz.json file containing the project id. Subsequent commands in that folder read that id automatically.
# Create a standard project with hosting, database, auth, and analytics
vibiz init "Acme Support"
# Create a project using the Next.js starter template
vibiz init "Acme Support" --template nextjs-saas
# Provision only specific resources for an app hosted elsewhere
vibiz init "Acme Backend" --resources database,ai
# Add a domain during initialization
vibiz init "Acme Shop" --domain acme.example.com
# Check project status and provisioning progress
vibiz statusBy default, vibiz init provisions hosting, serverless Postgres, sign-in, and analytics. If you need other capabilities later, add them to your project as needed:
# Add opt-in capabilities to an existing project
vibiz add email
vibiz add replay
vibiz add aiFor a step-by-step walkthrough of starting your first project, see the Quickstart guide.
Working locally with isolated environments
Every project includes separate production and development environments. Full-access keys point to production by default, while Limited keys point to development. You can pass the --dev flag to target development explicitly from any key.
# Create a development copy and write variables to .env.development.local
vibiz dev
# Run a local process injected with development environment variables
vibiz dev -- npm run dev
# Reset the development copy and branch fresh from production tables
vibiz dev --reset
# List variable names configured on the project
vibiz env vars
# Set an app variable from your shell or a hidden prompt (write-only)
vibiz env set THIRD_PARTY_KEY --dev
# Generate a console link for a teammate to paste a secret directly
vibiz env link THIRD_PARTY_KEY --devThe development copy provisions an isolated database branch matching your production schema with zero rows, an isolated sign-in instance that accepts http://localhost:3000, and a dedicated AI key. App secrets are write-only: their values are never printed back to the shell or terminal. To understand how permissions boundary checks work across environments, review Keys and environments.
Deploying, checking diffs, and rolling back
Publishing a static or full-stack site takes one command from your project directory.
# Upload and deploy the current folder live
vibiz deploy -m "Ship auth flow"
# Deploy to an isolated preview URL
vibiz deploy --preview -m "Draft redesign"
# Promote a deployment to the live site
vibiz promote dep_123
# View deployment history, newest first
vibiz versions --limit 10
# Compare differences between the live version and the previous one
vibiz diff
# Roll back immediately to the previous production deployment
vibiz rollbackRunning vibiz rollback instantly shifts live incoming traffic back to the deployment before the live one. Application code reverts immediately, but the database schema and rows are not rolled back.
Calling AI models and generating media
thelastapi routes OpenAI-compatible calls across multiple frontier labs with one key, but you can also run completions, speech, and image commands directly from the shell.
# Stream an answer from a default model
vibiz ai "Draft a release note for version 1.2" --model vibiz/smart
# List available models and inspect their per-token prices
vibiz ai models
# Search for specific model families
vibiz ai models claude
# Check or set the daily AI dollar budget for the project
vibiz ai budget 5 --project prj_123
# Generate an image and save it to disk
vibiz image "A clean minimal logo for a coffee brand" -o logo.png
# Generate speech audio from text
vibiz say "Welcome to Acme" -o welcome.mp3
# Transcribe an audio file to text
vibiz transcribe call.mp3 --language enEvery generated media asset is saved automatically in your project's Files storage with a permanent link. You can inspect recent files with vibiz files latest or search previous runs using vibiz files search "logo". To learn how this setup simplifies backend code, read our post on one API key for an AI agent: hosting, database, email, payments and models.
Connecting AI coding agents and skills
If you develop using AI coding agents such as Claude Code, Codex, or Cursor, you can wire tools and workflows using the CLI instead of editing configuration files by hand.
# Connect coding agents to the remote MCP server automatically
vibiz mcp add --clients "Claude Code,Cursor,Codex"
# Run the local stdio MCP server directly
vibiz mcp
# Install all standard agent skills globally across supported agents
vibiz skills add --global
# List currently installed agent skills
vibiz skills listThe command vibiz mcp add detects your client configuration and registers the remote MCP endpoint at https://thelastapi.dev/mcp without overwriting existing entries. To see how these protocols connect to editor workflows, read what is MCP, and how do you give Claude Code real tools? or review how to connect a remote MCP server to Claude, ChatGPT, Cursor and Codex.
Quick reference table
Here is a summary of the most common commands used during your first week:
| Command | Action | Key requirement |
|---|---|---|
vibiz login | Sign in through browser and save machine key | Any account |
vibiz init <name> | Create project and generate ./vibiz.json | Limited or Full |
vibiz dev | Populate .env.development.local with isolated dev variables | Limited or Full |
vibiz dev -- <cmd> | Execute a command loaded with dev environment variables | Limited or Full |
vibiz deploy | Build, package, and deploy site to live URL | Limited or Full |
vibiz rollback | Revert live site to prior deployment version | Limited or Full |
vibiz env link <NAME> | Generate browser link for entering secrets securely | Limited or Full |
vibiz keys create <name> | Provision a new API key (--limited or --read-all) | Full access |
vibiz usage | Inspect spend by tool, model, or project | Limited or Full |
vibiz mcp add | Register remote MCP server inside agent config files | Any |
vibiz skills add --global | Install workflow instructions for Claude, Codex, Cursor | Any |
For complete command-line parameter breakdowns and advanced switches, reference the official CLI documentation.
You can begin testing these workflows by running npx @vibiz/cli login in your terminal or pasting the setup instructions into your agent.
Frequently asked questions
What is the difference between a Full-access key and a Limited key in the CLI?
A Full-access key can read production secrets, withdraw balances, manage custom domains, and revoke credentials. A Limited key can build, deploy, manage development environments, use AI models, and run actions, but it receives a `403 owner_only` error when attempting to alter live financial settings or view raw production secrets.
How do I pass environment variables to my local application without saving them to a file?
Run `vibiz dev -- <command>`, such as `vibiz dev -- npm run dev`. This passes all development environment variables directly into the child process memory without writing a `.env.development.local` file to disk.
Can an AI agent run the vibiz CLI directly in its shell?
Yes. When an agent runs `vibiz login`, the CLI outputs an authorization link and waits for the human developer to approve it in a browser, keeping full credentials out of the agent's context window.
How do I undo a bad production deployment with the CLI?
Run `vibiz rollback` to point production traffic immediately back to the deployment version that preceded the current live one. Application code reverts at once, though database tables and records are left untouched.