Backend logic and APIs
Node.js, serverless functions, integrations, webhooks, scheduled work, imports, exports, and the glue between systems.
Technology
How we choose technology matters more than which technology we choose.
Tools change every year. The judgment behind choosing them does not. This page explains how the studio evaluates, adopts, and runs the platforms that power AI infrastructure and client applications.
How we decide
Five principles behind every technology decision.
01
Choose boring technology for the core
PostgreSQL, TypeScript, React. These tools have survived hype cycles, have deep ecosystems, and are understood by every engineer who might maintain the project after us. Novel tooling enters only when it solves a problem the proven stack cannot.
02
The client owns the infrastructure
Repositories, deployments, databases, API keys, billing, domains, and app distribution accounts belong to the client. The build runs on accounts they control, not ours. If the engagement ends, nothing is held hostage.
03
AI earns its place per feature
Every AI feature is evaluated on concrete business value: does it save time, reduce error, or enable a workflow that was previously manual? If the answer is unclear, the feature ships without AI and we revisit when the data supports it.
04
Ship the operational layer, not just the interface
Authentication, permissions, error tracking, backups, analytics, environment management, and handoff documentation ship as part of the build. These are not optional polish added if there is time left.
05
One codebase, multiple platforms
React Native and Expo let us write once for iOS, Android, and the web where it makes sense. When a native module is required, we write it. But we do not maintain three separate codebases when one will do.
How we build
Four standards on every project.
01
AI infrastructure
Frontier model APIs chosen per task. Vector databases and embeddings for semantic search and recommendation. Structured extraction pipelines for document intelligence. Evaluation and monitoring frameworks to measure accuracy, cost, and drift in production.
02
Application delivery
Expo, React Native, React, TypeScript, Next.js, Astro, Node.js, and Vercel workflows. Mobile apps are built primarily on Expo when the project calls for native app stores or device features.
03
Client-controlled infrastructure
Repositories, deployments, databases, API keys, billing, domains, and app distribution accounts are planned around client-controlled setup wherever the platform allows.
04
Production basics
Authentication, role permissions, observability, backups, analytics, error paths, and handoff documentation are part of the build, not optional polish.
Production app infrastructure
More than a frontend stack.
Useful applications need the operational layer behind the interface. Entoura.Studio plans the backend, access model, deployment path, integrations, monitoring, and handoff details early so the build can be maintained after launch.
Data and access control
PostgreSQL, Supabase, authentication, role permissions, audit trails, backups, and data portability planning.
Mobile distribution
Expo Application Services, native device features, app-store preparation, TestFlight-style review paths, and Google Play release planning.
Production operations
Sentry, analytics, deployment history, repository access, environment variables, documentation, and handoff routines.
Current stack
What we ship with today.
Tools are chosen per project. This is the current default stack, not a permanent commitment. GitLab is supported as a mirror when client infrastructure requires it.
Primary
Anthropic Claude AI models
OpenAI AI models
Google Gemini AI models
React Native Mobile
Expo Build & deploy
React Web
TypeScript Language
Next.js Fullstack web
PostgreSQL Database
Supabase Backend
Vercel Hosting
Supporting
Python ML pipelines
Stripe Payments
Sentry Error tracking
GitHub Source control
Tailwind CSS Styling
Astro Content sites
Node.js Runtime
Figma Design
Google Cloud Infrastructure where required
Responsible AI
AI features have to earn their place.
Claude is treated as an engineering accelerator, not an unreviewed code generator. AI features are added only when they create concrete business value: search, summarization, classification, internal assistants, drafting, routing, and workflow support.
For client applications, AI has constraints: clear purpose, scoped data access, human review where risk is meaningful, documented limitations, and no hidden automation in sensitive decisions.
Useful software should be fast to ship and boring to operate.
Tell us about the problem. The build gets scoped, and you learn what it costs.
Discover Your AI Opportunities →