Quick Start with Claude Code
The zango-app-developer plugin is the primary way developers enter Zango. Because the
platform is opinionated and consistent, the plugin knows exactly how to generate correct, idiomatic code
from a plain-English description. Install it once, then go from an empty directory to a running,
multi-tenant app in three steps.
What you're building
A Zango project is one place that can run many separate business applications, each with its own private data. You describe the Agentic AI app you want in plain English, and the plugin builds it for you: the screens, the user roles, the workflows, and the AI agents that read your data and act on it, production-ready from the start.
You make only two choices: how it runs on your machine, and what you want to build. Everything underneath, the data, the security, and the background work, is set up for you. All you need to begin is Claude Code open in an empty folder.
Prerequisites
Claude Code
The CLI that hosts the plugin. Install on macOS:
curl -fsSL https://claude.ai/install.sh | bashSetup docs →
Docker Desktop
Runs the Zango stack locally — Postgres, Redis, and the app server. Compose ships inside it, so there is nothing else to install. Install docs →
Install in Claude Code
The plugin is distributed through the zelthy marketplace by Healthlane Technologies. Install
it once and it activates automatically whenever you work on a Zango project.
claude plugin marketplace add Healthlane-Technologies/zelthy-claude-skills
claude plugin install zango-app-developer@zelthy
Open Claude Code in your project and the plugin loads on its own. You can also invoke it explicitly in any session:
/zango-app-developer
Use claude plugin list to see what is installed,
claude plugin update zango-app-developer@zelthy to upgrade, and
claude plugin uninstall zango-app-developer@zelthy to remove it. Not on Claude Code? The same
expertise ships as a skill file you can load into Cursor, GitHub Copilot, or Windsurf.
Want more starting points? Browse What You Can Build for ready-to-paste prompts that go further.
The developer loop
After the first build, every change is the same move: describe it in plain language, and the plugin maps it to the right primitive and implements it end-to-end.
| You say | The plugin knows |
|---|---|
| "Add a flag for urgent cases coordinators can set independently of status" | This is a workflow Tag, not a status. Adds tags + tag_transitions, runs zango update-apps. No migration needed. |
| "When a case moves to Resolved, send the patient an SMS" | This is a workflow done method + Communication package. Adds resolve_done(), calls the Communication API from inside it. |
| "Add an AI agent that reads case history and suggests next steps" | This is the AI module. Generates tools.py, sets up the Provider / Prompt / Agent, adds agent.run() in the view. |
| "The case list should look like a kanban board grouped by status" | A custom table body. Generates a customTableBody grouping records by status label, with showBody=false. |
What the plugin handles
The plugin does far more than write code. It also drives the platform's own APIs, so the configuration that used to be manual App Panel work happens automatically too. There is no separate "now go click through the App Panel" step.
Generates the code
Writes all Python · runs migrations · writes policies.json · wires views & URLs ·
creates workflows · generates React components · adds AI tool functions.
Configures the platform via API
Syncs policies · toggles the auth method · sets session timeout & password policy · creates the Provider, Prompt, and Agent · adds users & assigns roles · schedules async tasks.
Judgment, not plumbing: pasting real secret values (production API keys), reviewing what was generated, and approving anything that ships to production. The App Panel stays available whenever you want to inspect or adjust a setting by hand.