Launched July 14, 2026
MDMA - GenUI for apps

MDMA - GenUI for apps

Turn LLM output into interactive UI, forms, tables, approval gates, straight from the chat.

FreeMCP961 impressions#20 of its week

MDMA (Markdown Document with Mounted Applications) is an open-source format that extends Markdown with interactive components, forms, tables, approval gates, charts, callouts, webhooks,declared in fenced ```mdma``` blocks. When an LLM knows the spec, it responds with validated, renderable UI instead of plain text. Ships 9 component types, a remark-based parser, React & React Native renderers, a validator, a prompt-pack (OpenAI/Anthropic/Google/xAI), a CLI, an MCP server, and its own fine-tuned authoring model on Hugging Face. Deterministic parsing, headless runtime with audit log + PII redaction. Open source, MIT. Docs & demo: https://mdma.software · Built by Mobile Reality (https://themobilereality.com).

MDMA extends Markdown with interactive UI components that LLMs can generate and render in React, Vue, and React Native.

for
Developers building LLM-driven interfaces and workflows.
pricing
open source
license
MIT
MobileReality/mdma 67 2TypeScriptupdated 20 days ago
works withReactVue 3React NativeAG-UIOpenAIAnthropicGooglexAI

Key features

8 features of MDMA - GenUI for apps
  • MDMA Parser — Converts fenced ```mdma``` blocks into a validated AST using a remark plugin.
  • Static Validator — Runs 17 lint rules with auto-fix, catching structural errors without token cost.
  • Stateful Runtime — Provides bindings, audit log, policies and PII redaction for interactive components.
  • Multi-Framework Renderers — React, Vue 3 and React Native renderers turn MDMA documents into live UI.
  • Component Library — Ten built-in interactive types (form, button, table, approval-gate, etc.) plus custom variants.
  • Prompt Pack & Fixer — Model-specific authoring prompts and an LLM fixer to repair invalid output.
  • Streaming & Agent Bridge — Supports streaming renders and AG-UI protocol for bidirectional agent interaction.
  • Token-Efficient Design — Forgiving Markdown format works reliably on small, cheap LLMs, reducing API costs.

Use cases

  • Render a contact form directly from LLM output
  • Build a budget-approval workflow with sign-off gates
  • Create an employee onboarding checklist with equipment selection
  • Generate a sortable sales dashboard table with charts
  • Implement an incident-triage flow with severity assessment and closure approval

MDMA - GenUI for apps vs alternatives

MDMA - GenUI for appsPipelexUnified MCP ServerScryerMetorial
Best forMarkdown-based generative UI for LLMsDeclarative AI workflow definitionsLarge tool integration marketplaceBlueprint-style agent guidanceAI agent integration infrastructure
PricingOpen sourceFreeSubscriptionFreeFree
DevHunt upvotes0402957
LaunchedJul 2026Oct 2025Jan 2026Mar 2026Oct 2025

MDMA - GenUI for apps FAQ

How does MDMA ensure the generated UI is valid?+

MDMA ships a deterministic validator with 17 lint rules and auto-fix strategies that run without an LLM, catching structural issues for free.

Can MDMA work with small, open-weight models?+

Yes, its forgiving Markdown format is designed to be reliably emitted by cheap models like Gemini Flash and Claude Haiku, keeping token costs low.

What rendering targets are supported?+

MDMA provides official renderers for React, Vue 3, and React Native, all sharing the same component spec and theming.

Is it possible to add custom UI components?+

Custom components can be emitted as type custom with a name, and the host registers the variant to render it.

How does MDMA handle approvals or gated actions?+

The approval-gate component enforces role-based sign-off before proceeding, with the runtime logging actions for audit.

Summarized by DevHunt from mdma.software · Sep 28, 2026. Details may change; check the official site.