Overview
UIX Protocol is designed to be AI-native. Every part of the schema is machine-readable, and AI agents interact with UIX through the Model Context Protocol (MCP).
MCP tools
UIX exposes 58 MCP tools across two servers:
@uix/mcp (34 tools — document operations)
| Category | Tools |
|---|---|
| Document | list, get, create, update, delete, history |
| Query | query, node get, nodes by type, token resolve |
| Mutation | node set, node insert, node remove, apply diff |
| Analysis | mind build, soul annotate, render |
| Validation | validate |
| Version Control | init, commit, branch create, log, checkout, diff, merge |
| Collaboration | crdt ops get |
| Image-to-App | image analyze, image to generation context |
@uix/kernel (24 tools — workflow operations)
| Category | Tools |
|---|---|
| Agent | register, list |
| Workflow | create, execute, status |
| Pipeline | trigger, status |
| DAG | build, validate, generate documents |
| Integration | configure, status, test, webhook replay |
| Backend | plan, generate, security audit |
| QA | functional tests, full report |
Soul layer
Annotate documents with semantic meaning:
typescript
const withSoul = annotateSoul(doc, {
purpose: 'Authenticates existing users',
userGoal: 'Get back to work quickly',
journey: 'onboarding.step-1',
emotion: 'trust',
});Mind layer
Build a pre-computed AI reasoning surface:
typescript
const mind = buildMind(doc);
mind.screenType; // 'authentication'
mind.summary; // 'Authenticates existing users'
mind.dataRequirements; // ['auth.email', 'auth.password']
mind.intents; // ['auth.login', 'auth.setEmail']
// Focused context for one node
const ctx = getMindContext(mind, 'submit-btn');Image-to-App pipeline
UIX can generate a complete multi-screen app from a design image:
- Image analysis — semantic detection of layout, palette, typography, navigation
- Design system extraction — tokens derived from the image
- Screen planning — 7 domain templates (social, ecommerce, productivity, etc.)
- App Flow DAG — complete app topology: screens, states, popups, transitions
- Document generation — UIXDocuments for every screen and state
- Code generation — multi-platform output from the generated documents
- QA validation — automated test generation + coverage analysis
App Flow DAG
The central artifact for autonomous app building — a directed acyclic graph of:
- Screen nodes — main app screens
- State nodes — loading, error, empty, success states for each screen
- Popup nodes — modals, alerts, bottom sheets
- Transitions — navigation with triggers and animations