The optimal default stack for AI-native products.
Vibe Stack is an opinionated gentleman's kit: a curated set of self-hostable, production-ready technologies that covers the common 95% of product use cases without forcing teams or agents to compare equivalent alternatives.
It is designed for both a smooth fast start and long-running autonomous development and maintenance through AI agents.
| Need | Default |
|---|---|
| Language | TypeScript |
| Package manager | pnpm |
| UI | React + Next.js |
| Agents | Mastra |
| AI UI layer | assistant-ui |
| Auth | Better Auth |
| Data access | Drizzle ORM + PostgreSQL |
| Validation | Zod |
| Logging | Pino JSON logs to stdout |
| Components | Mantine |
| Styling | Tailwind CSS v4 |
| Quality | TypeScript strict mode + Ultracite + Oxlint |
| Animation | Motion |
| Need | Default | Replaces |
|---|---|---|
| Deployment | Dokploy | Vercel-style hosted deployment |
| Database | PostgreSQL | Supabase-style managed Postgres platform |
| Error tracking | Bugsink | Sentry; compatible with Sentry SDKs |
| MVP observability | Structured stdout/stderr logs via Docker / Dokploy | Heavy log platforms too early |
| MVP+ observability | OpenObserve + OpenTelemetry | Grafana-style observability stack |
| Agent observability | OpenObserve AI/LLM observability | A separate agent-observability backend |
Start with structured logs and Bugsink. Add OpenTelemetry + OpenObserve when traces, metrics, retention, and correlation become worth the extra resources. For agent products, send model, tool, retrieval, workflow, and agent-turn spans to the same OpenObserve backend; its built-in Online Evaluations are an optional Enterprise feature.
Full guides Detailed docs: Web application, TypeScript, Vibe Infra, Security Practices, and Common guide.
| Area | Status |
|---|---|
| TypeScript web apps | Covered |
| Vibe Infra | Covered |
| Common agent-ready practices | Covered |
| Security practices | Covered |
| Existing web apps to iOS/Android | Capacitor 8.5 agent skill |
| Apple App Store submission review | App Store reviewer skill |
| Python | Planned, not selected yet |
| .NET | Planned, not selected yet |
| Backend-only stacks | Planned, not selected yet |
agentic-capacitor helps agents add Capacitor 8.5 to an existing web app, especially React 19.2, maintain iOS/Android integrations, and prepare for a later Capacitor 9 upgrade. It preserves the existing web architecture and includes offline validation scripts.
See the skill README for installation, example requests, validation commands, and coverage limits.
apple-app-store-reviewer audits Apple-platform release candidates before submission or resubmission, including Capacitor apps. It combines deterministic checks with evidence-based manual review; it does not guarantee App Store approval.
Use this repo as a decision shortcut, not as a comparison catalog.
- Read this README first.
- Open the relevant guide for the project type:
Rules:
- Use the named default unless the project has a concrete constraint that makes it wrong.
- Do not ask users to choose between equivalent alternatives when this repo already names a default.
- Do not add a second equivalent tool just because it is popular.
- Prefer self-hostable, production-ready tools with a low operational burden.
- Keep MVP choices simple: choose the smallest stack that supports fast feedback, clear errors, and autonomous maintenance.
- When deviating from a default, write down the reason in the project docs or implementation notes.
- If a category is not covered yet, make the smallest reversible choice and mark it as project-specific.
Pull requests are welcome for better recommended defaults across languages, app types, infrastructure, observability, auth, deployment, and templates.
The bar is intentionally high: recommendations should stay small, useful, proven, production-ready, and self-hosted first whenever possible. See CONTRIBUTING.md.