decantr-ai/decantr
Native LLM-focused webapp building for the UI
Platform-specific configuration:
{
"mcpServers": {
"decantr": {
"command": "npx",
"args": [
"-y",
"decantr"
]
}
}
}Add the config above to .claude/settings.json under the mcpServers key.
[](https://github.com/decantr-ai/decantr/actions/workflows/ci.yml) [](https://www.npmjs.com/package/decantr) [](LICENSE)
AI-first web framework. Zero dependencies. Native JS/CSS/HTML. v0.9.6
Decantr is designed for LLMs to generate, read, and maintain — not for human readability. Every API is optimized for token efficiency: terse atomic CSS atoms, proxy-based tag functions, and a machine-readable registry so agents can look up props and exports without parsing source files.
Every Decantr app begins as a conversation. The Decantation Process is a structured pipeline that turns a plain-English description into a production app — one layer at a time.
POUR → SETTLE → CLARIFY → DECANT → SERVE → AGE
you AI AI+you AI AI ongoing
describe decomposes crystallizes resolves generates guards
intent layers the Essence layout code against drift---
> *You describe what you want in plain English — the AI handles the rest.*
No config files, no boilerplate, no decisions about folder structure. Just say what you need.
"I need a SaaS analytics dashboard with authentication,
real-time data feeds, and a dark command-center aesthetic."---
> *Your intent is decomposed into five independent layers.*
The AI breaks your description into five orthogonal layers, each independent and composable:
Terroir → saas-dashboard (what kind of app — sets default patterns and pages)
Vintage → command-center / dark (how it looks — visual style, color mode, shape)
Character → tactical, data-dense (personality — controls spacing, density, and tone)
Structure → overview, settings (pages — the views your app needs)
TLoading reviews...