Claude Code on RootCX
The fastest way to ship internal software with Claude Code. You describe, it builds, and RootCX runs it in production with a managed database, enterprise auth, permissions, and a full audit trail out of the box.
Build an inventory management app. Products with SKU, name, price, stock level, and a supplier link. A dashboard with low-stock alerts, a product list with search and filters, and a form to receive new stock.
● Loading rootcx-manifest — entities: products, suppliers, stock_movements
● Loading rootcx-sdk-hooks — useAppCollection with where filters, sorting
● Loading rootcx-ui — AppShell, DataTable, Forms, Badge
● Loading rootcx-backend-worker — RPC handler for stock receive
Wrote manifest.json — 3 entities, 6 fields, 5 permissions
Wrote src/App.tsx — AppShell with sidebar navigation
Wrote src/pages/Dashboard.tsx — low-stock alerts, stats cards
Wrote src/pages/Products.tsx — DataTable with search, filters, pagination
Wrote src/pages/ReceiveStock.tsx — form with supplier select, quantity input
Wrote backend/index.ts — receive_stock RPC, stock level update
/rootcx-deploy
Manifest installed, schema synced (3 tables created)
Backend uploaded, worker started
✓ inventory is live — SSO, RBAC, audit log included
Without RootCX, that code still needs a database, auth, permissions, hosting, and monitoring
With RootCX, it deploys and runs in production immediately. Every app gets all of this from the first deploy. You write zero infrastructure code.
Managed PostgreSQL
Auto schema sync, no migration files
CRUD APIs
Generated on every entity, query DSL included
SSO (OIDC)
Okta, Entra ID, Google, Auth0
Role-based access
Namespaced permissions, wildcards, inheritance
Immutable audit log
Database trigger level, every change captured
Encrypted vault
AES-256 for API keys and secrets
Background jobs
Durable queue with automatic retry
Crash recovery
Worker supervisor with exponential backoff
Live log streaming
Real-time stdout/stderr over SSE
Integrations
Apollo, GitHub, Stripe, Slack, MCP
Self-hostable
Docker + any PostgreSQL, no feature gap
Source-available
FSL-1.1-ALv2, converts to Apache 2.0
How does Claude Code know your stack?
Six official skills teach it the RootCX SDK, components, manifest format, backend protocol, API, and agent framework. It loads the right one automatically.
rootcx-manifestKnows how to define your data model, entity relationships, and permissions. Writes manifest.json so the database creates itself on deploy.
rootcx-sdk-hooksKnows every React hook for reading and writing data. Filters, sorting, pagination, cross-app queries. No REST boilerplate.
rootcx-uiKnows 30+ components: layouts, tables, forms, dialogs. Builds production interfaces with dark mode out of the box.
rootcx-backend-workerKnows how to write server-side logic: RPC handlers, background jobs, database access, authenticated API calls.
rootcx-rest-apiKnows every Core API endpoint: CRUD, bulk operations, integrations, job queue. For when the SDK hooks are not enough.
rootcx-agentKnows how to build AI agents: LangGraph config, system prompts, tools, supervision modes. From scaffold to production agent.
The only three commands you need
/rootcx-connect <url>Connect to a Core
Point at your local Docker instance, a cloud project, or a remote server. Handles auth automatically (password, OIDC, or bearer token for CI).
/rootcx-new app | agent <name>Scaffold a project
Creates the manifest, frontend (Vite + React + Tailwind), and backend skeleton. For agents, adds agent.json and a system prompt template.
/rootcx-deployShip it
Reads the manifest, syncs the database schema, uploads backend and frontend archives, starts the worker process. One command, full deploy.