Introduction

What is OvenLovin?

The family recipe box your AI can actually use, and it knows who can't eat what.

Open-source recipe box and meal planner for families and their agents. Per-person allergen checks (US Big 9 + EU 14). Not medical advice - always read ingredient labels before serving.

Why OvenLovin?

Traditional meal planning tools aren't built for AI agents. OvenLovin is designed from the ground up to work seamlessly with:

• Grok Bot - X's AI assistant
• ChatGPT - OpenAI's conversational AI
• Claude Desktop - Anthropic's desktop AI
• Generic MCP clients - Any Model Context Protocol compatible agent

Key Features

• Agent-first APIs - REST /api/v1 with OpenAPI, MCP Streamable HTTP at /mcp, /llms.txt
• Safety-critical rules - Per-person allergy and dietary restrictions with deterministic checking
• Import - URL, paste, and photo. Default model deepseek/deepseek-v4.1-flash (RK_MODEL_DEFAULT / RK_MODEL_VISION / RK_MODEL_EXTRACT). DeepSeek failures retry openai/gpt-6-luna. Voice and third-party app export are on the roadmap
• Multi-household - Clerk organizations as households

How It Works

1. Configure dietary rules - Set up per-person allergies and dietary restrictions with severity levels (block/warn/prefer_avoid)
2. Import recipes - Add recipes from a URL, photo, or paste
3. Plan meals - Let your AI agent propose weekly meal plans with automatic rule checking
4. Generate shopping lists - Get aggregated ingredient lists with allergen warnings

Architecture

OvenLovin is built as a TypeScript monorepo with clear separation of concerns:

• @ovenlovin/core - Pure TypeScript business logic, zod schemas, and services
• @ovenlovin/web - Next.js 15 App Router UI with REST API routes
• @ovenlovin/mcp - Model Context Protocol server for AI agents
• @ovenlovin/importers - Recipe importers (URL, photo, paste)
• @ovenlovin/allergens - Allergen taxonomy and rule checker
• @ovenlovin/db - Drizzle ORM schema and migrations for Neon Postgres
• @ovenlovin/brand - Wordmark, deploy URL, tokens

CLI and a published SDK are on the roadmap.

All business logic lives in @ovenlovin/core as pure TypeScript, ensuring that external bots get identical behavior through any interface.

Next Steps

• Quickstart - Deploy to Vercel or run locally in minutes
• Self-hosting - Env vars and Vercel, Neon, and Clerk setup
• Public demo - No-signup sandbox and how to run it
• Connect Your Agent - Set up Grok Bot, ChatGPT, or Claude Desktop
• Concepts - Learn about households, rules, and meal planning