Getting started with Mstro

1. Install the CLI

Run this in any project directory. That project becomes browser-accessible as an orchestra. No Docker, no VM, no config file.

npx mstro-app

Want it globally:

npm install -g mstro-app
mstro

2. Sign in

Open https://app.mstro.app and sign up or sign in with Clerk. You can use email, GitHub, or Google. The browser app pairs with the CLI via a one-time device code.

3. Start your first session

In the browser, pick your orchestra and open a new chat tab. Type a prompt like “add a skip-to-content link to the landing page”. Claude Code runs on your machine, edits files, runs tests, and reports back.

The Security Bouncer auto-approves safe tool calls (reading files, running lint, git status). Risky calls (rm, push --force) still require explicit approval.

4. Go further

• Turn one prompt into a PM Board of parallel agents.
• Understand how Mstro uses git worktrees for safe parallelism.
• Tune the Security Bouncer for your workflow.

Requirements

• Node.js 20+ (for the CLI)
• An Anthropic API key (Claude Code is BYOK)
• macOS, Linux, or Windows (via WSL)