Every app says "AI-powered." Few actually are.
Every food app claims to be AI-powered now. Usually that means a chatbot wrapper around a recipe database, or a recommendation engine that noticed you searched for "chicken" twice.
That's not AI planning. That's autocomplete with better marketing.
When we built Eatgent, we made a deliberate architectural choice: instead of bolting AI onto an existing recipe app, we built an agentic system. Multiple specialized AI agents collaborate to produce your meal plan. Each agent has a specific job, its own tools, and the ability to make decisions on its own.
So what actually happens when you tap "Plan My Week"?
What "agentic" actually means
A chatbot waits for you to ask a question, then answers it. An agent takes a goal and figures out how to accomplish it. It decides what information to gather, what tools to use, and how to handle tradeoffs along the way.
Eatgent doesn't wait for you to pick recipes. You give it a goal ("plan my week") and it orchestrates a multi-step process across four specialized AI agents, each handling a different part of the problem.
The full architecture:
Agent 1: Get to Know You
The first agent you'll meet is the onboarding agent. It's a conversational AI that builds your food profile through a natural dialogue, not a form.
It asks about your dietary restrictions, cuisine preferences, health goals, household size, cooking skill level, and how much effort you're willing to put in on a Tuesday versus a Saturday. The conversation adapts based on your answers. If you mention you're vegetarian, it doesn't ask about your favorite cuts of meat. If you say you have kids, it asks about picky eaters.
The output is a structured food profile that every other agent in the system uses as context. Not a static questionnaire result. It's a living representation of your household's food preferences that gets refined every time you use Eatgent.
Agent 2: Plan Your Week
The meal planning agent is the core of the system. It takes your food profile, loads your meal memory (what you've eaten over the past four weeks), and orchestrates a day-by-day search for recipes that fit your life.
What makes this different from a recipe recommendation engine:
It searches the real web. The agent doesn't pull from a fixed database. It searches across real recipe sources on the internet, evaluating thousands of options against your profile. Every recipe in your plan is a real, tested recipe from a real source, not something the AI generated or hallucinated.
It plans day by day. The agent doesn't just find seven good recipes and call it a week. It builds the plan sequentially, Monday through Sunday, balancing nutrition, variety, cuisine diversity, cooking complexity, and ingredient reuse as it goes. Wednesday's meal is chosen with Monday and Tuesday already in context.