On this page

Cookbook

Practical, end-to-end walkthroughs. Each recipe starts from a blank terminal and ends with something running. Pick the one closest to your use case and adapt from there.

Agent builds an app

Your AI agent receives a prompt, writes code into a sandbox, runs the build, exposes a preview URL, then publishes to production. End deliverable: a deployed Next.js app from a single prompt.
Streaming code execution

Run a long Python script in a sandbox and stream stdout to your UI in real time via SSE. Includes backpressure handling. End deliverable: a code-execution console-style UI snippet.
Multi-tenant SaaS

Embed MIOSA into your SaaS. Tag every sandbox with your own workspace and user IDs, query usage per tenant, and issue browser tokens so your customers interact directly with their sandbox without seeing your API key.
Snapshot & fork

Checkpoint a long-running task, fork into multiple branches, compare outcomes, and restore to a known-good state on failure.
Browser agent

Boot a Computer (full Linux desktop), screenshot a website, click a button, fill a form, take another screenshot. Implements the screenshot → action → screenshot loop with your own LLM.

Before you start

All recipes share the same prerequisites:

  • A MIOSA workspace API key (msk_live_*) — see API Keys
  • Node 22+ or Python 3.11+ installed locally
  • The MIOSA SDK installed for your language:
# Python
pip install miosa

# TypeScript / Node
npm install @miosa/sdk

Set the key in your environment once — every recipe reads it from there:

export MIOSA_API_KEY="msk_live_..."

Was this helpful?