Plan your meals with OpenClaw

The hard part of cooking at home is the deciding, done five nights a week at the worst possible time, when you are already hungry and out of ideas, and then turning whatever you decided into a shopping list that does not forget the garlic.

Meal preparation, planning and often prepping food ahead of time, fixes both problems in theory. In practice the planning itself is a chore, and the reward is a second chore, writing out everything you need to buy. People who batch cook on Sundays or prep lunches for the work week report less decision fatigue during the week, but someone still has to pick the meals and build the list first.

Your Operator.io agent can take both of those off your plate. You send it one prompt and once a week it proposes a set of dinners built around what you like, then turns those meals into a single clean grocery list.

How it works

You send the meal planner prompt, and the agent sets up the plan and the list together:

• It keeps the week's plan in a CSV in its workspace, with the day, the meal, and the recipe or main ingredients.
• Once a week it fills that in with dinners chosen around your tastes, your dietary needs, and how much effort you want to spend.
• It reads back across those recipes to build the grocery list, folding a repeated ingredient into one line with the total amount instead of scattering it.
• When you tell it what you already have or what you ended up cooking, it adjusts both the plan and the list rather than starting over.

The deduped list is what makes it usable in the store: three meals that each call for onions become one entry that tells you how many to buy instead of three lines you have to add up at the shelf. Mealime markets its Pro tier partly on reducing food waste by planning around overlapping ingredients, and the same logic applies here, except the agent works from recipes you name rather than a fixed catalog.

The prompt

This is the instruction the agent acts on:

Be my meal planner. Create a spreadsheet in your workspace, a CSV with columns for
the day, the meal, and the recipe or main ingredients. Once a week, plan my
dinners for the week based on what I like, then build a grocery list from those
meals and send it to me. When the same ingredient shows up in more than one
recipe, combine it into a single line with the total amount rather than listing it
again. When I tell you what I already have or what I actually cooked, adjust the
plan and the list. Before you start, ask me how many people I am cooking for, any
dietary preferences or foods to avoid, how many dinners to plan, and which day to
send the plan and list.

Instead of retyping it, you can pull this prompt from the prompts library and send it to your agent.

Using it day to day

Once a week the plan and the list arrive together in your channel. You scan the dinners, swap the one you are not feeling for something else in a sentence, and take the list to the store. During the week it bends to real life: tell it you already have rice and eggs and it drops them from the list, tell it Wednesday fell through and you want to stretch Tuesday's leftovers, and it reshuffles. When you find a recipe you love, you hand it over and it goes into the rotation for next time.

Because the plan is a file the agent keeps, you can lean on its memory instead of yours. Ask it what you have not made in a while, to keep next week under a certain budget, to plan around the chicken you need to use before it turns, or to lean vegetarian for a stretch.

Your preferences and the number of dinners are set at the start, so you change them by asking. If you follow a specific diet, keto, Mediterranean, dairy free, tell the agent once and it filters suggestions accordingly, the same way Mealime offers diet filters in its Pro tier but without locking you to Mealime's recipe library.

It is worth separating a food someone would rather not eat from one they cannot. If anyone at your table has a real allergy, tell the agent it is an allergy when it asks what to avoid, not just a dislike, so it keeps that ingredient out of the dinners it picks and out of any swap it suggests later when an ingredient runs short.

The agent plans from recipes and what it knows about common ingredients, so it cannot read the label on the jar in your cupboard or tell whether a packaged sauce was made on equipment that also handles nuts. For a meal that involves an allergen for someone eating it, read the recipe and the packaging yourself before you cook, since the agent is planning from recipe text and cannot see the jars on your shelf.

Why this beats another meal app

The meal apps each handle one piece of this.

The Paprika and AnyList prices come from Eat This Much's 2026 comparison. Each one asks you to live inside its library or do the planning by hand, which is the step that stops happening.

The agent covers all three jobs without making you live in one app's library. The plan is built around the food you like rather than a fixed recipe database, the grocery list comes out deduped and ready, and the whole thing arrives in the channel you already use, so a change is a reply rather than a trip through menus.

The plan and the list are plain files of your own, with nothing to subscribe to beyond the OpenClaw you already run. You still do the cooking, but what it takes off you is the deciding and the list, the two chores that end most attempts at eating in.

Open the prompts library and send the meal planner to your agent to begin. It asks how many people you cook for and any foods to avoid, then the weekly plan and grocery list arrive together in your channel.