withvibe
Getting started

Quickstart

From zero to your first shared AI env in under five minutes. We'll run withvibe locally, create a workspace, and spin up an env seeded with a sample repo.

Prerequisites

  • Node.js 20+ (ships with npm)
  • Docker Desktop (or Docker Engine on Linux), with the daemon running
  • An Anthropic API key with access to Claude — you'll add it from the UI

1. Install & start

npm install -g withvibe && withvibe init --yes

This installs the CLI, writes ~/.withvibewith generated secrets and free ports, pulls the prebuilt images, and brings up three containers — the web app, the API, and Postgres for the API's metadata. Per-env containers are spawned later, on demand.

First-run secrets
withvibe init generates the Postgres password and session secret automatically. The Anthropic key is not prompted at install — you add it from the workspace UI in the next steps (or any time, under Settings).

2. Open the web app

Visit http://localhost:3000, register the first user, and you'll land in your new workspace as its owner. Invite teammates by email; they register from the link.

3. Add your Anthropic key

Go to Settings and paste your ANTHROPIC_API_KEY. Agent runs stay disabled until this is set.

4. Create your first env

Click New environment. Pick a Git repo (or use the bundled sample), give it a name, and confirm. Within ~30 seconds the API will:

  1. Clone the repo into a fresh Docker container
  2. Open a Claude Agent SDK session inside it
  3. Drop you into a chat + terminal view

5. Vibecode together

Type a feature request into the chat: “Add a /health endpoint that returns the commit SHA.” The agent will plan, edit files, and run tests. Click Share link to bring teammates in — they see the same chat, the same terminal, and can steer the agent themselves.

6. Ship

When everyone is satisfied, hit Submit for review. The agent gate kicks off: security, code review, tests, and policy agents run in parallel against the env's branch. If they all pass, the env opens a pull request to main and your existing CI/CD takes over.

What's next?
Read Workspaces & envs to understand the data model, The agent gate to configure the pre-prod checks, or Install for a production deployment.