EveryBite is a nutrition intelligence platform that sits between restaurants and the applications guests use. We ingest nutrition data from multiple sources, process it with AI/ML, and serve it through APIs that power personalized dining experiences.This page walks through the architecture from top to bottom, starting with a high-level view and then drilling into each layer.
About these diagrams: The integrations shown (e.g., Olo, Thanx, PAR) are examples of supported platforms. Not all integrations are live for every deployment. Contact your EveryBite representative for current availability.
Large language models and conversational AI that help guests find food through natural language:
“Find me something vegan under 500 calories”
“What can I eat here if I’m allergic to peanuts?”
“Show me high-protein options”
AI agents connect through AI Tools, built on the Model Context Protocol (MCP). This provides tool definitions, context management, and session handling, allowing AI to call our APIs conversationally without building custom integrations.
Traditional mobile and web applications built by our partners (ordering platforms, loyalty providers, restaurant chains). These connect directly to the SmartMenu API (GraphQL) with full control over queries and data fetching.Key difference: AI agents use MCP for conversational access. Partner apps use GraphQL for programmatic access. Both hit the same underlying SmartMenu API.
Upstream: Data Sources (external systems), EveryBite Platform (read requests)Downstream: Database (writes), Platform Services (reads)Data In: Raw nutrition files, PDF documents, API feeds from ordering/loyalty systems, manual correctionsData Out: Standardized dish records, validated nutrition panels, allergen classifications, diet tags
The engine behind accurate, real-time nutrition data.The Data Platform ingests data from ordering systems, nutrition providers, and loyalty platforms—then uses AI/ML to match, classify, and validate everything into a unified database. Each data source has its own dedicated pipeline:Each data source flows through its own pipeline into the unified database.
Potential Future Additions: Health Apps (Apple Health, Fitbit) for personalized nutrition goals, and Inventory for real-time availability.
The ordering system is the source of truth for what guests can actually order. We sync menu items, categories, modifiers, and real-time availability so nutrition data only appears for items that are actually on the menu.
Source
Integration
Data
Olo
Real-time API
Menu sync, item availability, pricing
PAR
API
Menu structure, POS integration
Toast(coming soon)
Webhook
Menu updates, inventory status
Square(coming soon)
API
Menu catalog, location data
How it works:
LLM matching links ordering items to nutrition data (95%+ accuracy)
Unmatched items are flagged as exceptions and published to the Developer Portal
Partners review exceptions and update item names in their ordering system and/or nutrition solution
This is where the magic happens. We ingest nutrition data from multiple sources, use AI/ML to classify allergens and diet tags, and link everything to the menu items guests are ordering. The result: accurate, actionable nutrition information for every dish.All data flows through our seven-layer ingredient hierarchy: Menu → Dish → Recipe → Prep Recipe → Ingredient → Ingredient Data → Ingredient Specification. This deep structure is how we trace allergens through house-made components and deliver precise nutrition calculations.
Source
Integration
Data
MenuCalc
Direct (sister company)
Full nutrition panels, recipe analysis
Trustwell
API
Genesis/Food Processor exports
Restaurant PDFs
OCR processing
Nutrition documents, allergen sheets
AI/ML Processing:
NLP extraction to parse ingredients (92%+ accuracy)
Allergen classifiers to detect FDA Big 9 (98%+ accuracy)
Diet tagging for Vegan/Vegetarian/Pescatarian (97%+ accuracy)
Loyalty data helps us understand guests before they even set preferences. With consent, we use purchase history and stated preferences to personalize recommendations from the first interaction—and for new guests, behavioral modeling provides intelligent defaults.
Source
Integration
Data
Spendgo
API
Customer preferences, loyalty program data
Thanx(in progress)
API
Customer segments, stated preferences
Punchh(coming soon)
API
Purchase history, reward status
AI/ML Processing:
Audience segmentation and cohort analysis
Behavioral modeling for new guest personalization
Industry trends across platforms (app vs kiosk vs web)
Every pipeline follows the same journey from raw data to API-ready intelligence. This consistency ensures data quality and makes it easy to add new data sources as we grow.
Stage
What Happens
Ingest
Connect to source systems and pull raw data on a scheduled or real-time basis
AI/ML
Match items across systems, classify allergens and diets, extract insights
Pipelines
Transform, standardize, validate, and run quality assurance checks
Database
Store in unified schema, ready to serve through the SmartMenu API
Partners must not cache API responses. Allergen data can change at any time, and stale data poses a health risk. We handle caching internally so you don’t have to.
Source of Truth
The ordering system (e.g., Olo) is the source of truth for availability. SmartMenu never returns dishes that aren’t in your system.
Exact Matching
Dishes are matched by name between ordering systems and our nutrition database. Mismatches appear in exception reports.
Session-Based
Preferences are applied per-session, not stored permanently (unless using Passport). Every request requires a session ID.
AI-powered assistants that can interact with restaurant menu data through natural language. These include:
Agent Type
Description
Use Case
Claude / GPT-4
Large language models
Conversational menu browsing, dietary recommendations
Custom Agents
Partner-built AI assistants
Branded experiences, specialized workflows
Chatbots
Text-based interfaces
Customer support, order assistance
Voice Assistants
Speech interfaces
Drive-thru, accessibility, hands-free ordering
AI agents connect via AI Tools, built on the Model Context Protocol (MCP), which provides tool definitions and context management for seamless integration.
