Getting started

Shmastra is a Mastra template. You bootstrap a fresh Mastra project from it, then start the dev server and open Mastra Studio.

Requirements

• Node.js ≥ 22.13.0 (required by both Mastra and Shmastra).
• npm or pnpm — Shmastra auto-detects pnpm and prefers it when available.
• At least one LLM provider account — OpenAI, Anthropic, or Google AI. You can either log in via OAuth (OpenAI / Anthropic) during the setup wizard, or paste an API key. Google AI needs an API key.

Bootstrap

npx create-mastra@latest --template https://github.com/just-ai/shmastra

Follow the on-screen prompts (project name, package manager). The template is copied into the folder you chose and dependencies install automatically.

First run

From the project directory:

npm run dev

On the first start Shmastra runs its setup wizard in the terminal. You'll be asked to:

1. Optionally log in to OpenAI or Anthropic with OAuth (so the coding agent uses your existing Pro/Max subscription instead of burning API credits).
2. Add at least one provider API key. The key is written to .env.
3. Optionally add a Composio API key to unlock 200+ third-party toolkits (Gmail, Drive, Slack, Notion, …).

Once the wizard finishes, the Mastra dev server starts on http://localhost:4111 and Mastra Studio opens with the Shmastra chat widget already attached.

Say hello

In the widget, try:

Create an agent called "scribe" that summarises any text I send it in three bullet points.

Shmastra will:

1. Write src/mastra/agents/scribeAgent.ts.
2. Register it and run a dry-run build.
3. Hot-reload the dev server — the agent shows up in Mastra Studio immediately.

Click the new agent in Studio to chat with it directly.

Other commands

npm run build   # production build (mastra build)
npm run start   # run the production server
npm test        # unit tests (Vitest, pure helpers)

Next

• Setup wizard — what runs on first start.
• The chat widget — how vibe-coding works.
• Features — full menu of what you can ask for.