Track your calories with OpenClaw

Most calorie tracking dies the same way. You download an app, log everything for four days, then forget on day five and never open it again. The app was never really the problem. Keeping a food diary is the part that produces results, and it is also the part people quit first, because nothing reminds you and the logging feels like a chore.

The evidence on that point is old but still cited everywhere for a reason. A 2008 Kaiser Permanente study of 1,685 adults in the Weight Loss Maintenance trial found that participants who kept daily food records lost roughly twice as much weight as those who kept none over six months, with an average loss of about 13 pounds across the group. The trial also had people follow the DASH diet, attend weekly group sessions, and exercise at moderate intensity for at least 30 minutes a day, so the diary was one piece of a larger program.

Lead author Jack Hollis told ScienceDaily that the more food records people kept, the more weight they lost. The published trial data backs that up: frequency of diet records was one of the strongest behavioral predictors of weight change.

Later reviews paint the same picture at scale. A 2022 systematic review in Public Health Nutrition looked at 59 behavioral weight loss studies and found that most interventions using dietary self monitoring, whether on paper, a website, or a mobile app, produced statistically significant weight loss compared with controls. An earlier 2011 review in the Journal of the American Dietetic Association called self monitoring the centerpiece of behavioral weight loss programs and noted a consistent association between logging and outcomes, even if the evidence quality was limited by self reported adherence.

The logging itself is what works. The hard part is keeping it going.

Your Operator.io agent can take both jobs: the spreadsheet and the nudge. You hand it one prompt, and it sets up a running log of what you eat plus an automation that checks in with you a few times a day so the log stays current.

How it works

You send the calorie tracker prompt, and the agent sets up the log and the reminder that keeps it current:

• It creates a CSV in its workspace with the date, time, item, calories, protein, carbs, and fat.
• When you tell it what you ate, it works out the calories and macros, adds a row, and replies with your running total for the day against your goal.
• A few times a day it messages you to ask what you have eaten, a scheduled job that runs whether or not you remember to think about it.

The file is yours: it lives in OpenClaw's files, you can open it any time, and you can hand it to a coach without prying it out of someone else's app. The reminder is what keeps you logging, and you can pause or change it on the Automations page whenever you want.

The prompt

This is the instruction the agent acts on:

Be my calorie and macro tracker. Create a spreadsheet in your workspace, a CSV
with columns for date, time, item, calories, protein, carbs, and fat. When I
tell you what I ate, add a row and reply with my running total for the day
against my goal. Then set up an automation that messages me here a few times a
day to ask what I have eaten, so the log stays current. Before you start, ask me
my daily calorie goal, my protein target, and what times you want to check in.

You can hand this prompt to your agent from the prompts library instead of retyping it.

Picking a calorie and protein target

The prompt asks for your daily calorie goal and protein target before it writes the first row. If you already have numbers from a dietitian or a coach, give those.

If you are starting from scratch, the NIH Body Weight Planner is a free tool that models how calorie intake and activity level interact with weight change over time, accounting for metabolic adaptation in a way the old "3,500 calories equals one pound" shortcut does not. You enter your current weight, height, age, sex, activity level, and a goal weight with a timeline, and it returns a daily calorie target for losing weight and a separate maintenance number for after you get there.

Protein targets are usually set as grams per day or as a share of total calories. The DASH eating plan, which the Kaiser trial used alongside food diaries, emphasizes lean protein alongside fruits, vegetables, and low fat dairy, but the exact gram count depends on your body weight and goals. Many people land somewhere between 0.7 and 1.0 grams of protein per pound of body weight when they are trying to preserve muscle during a cut, though a registered dietitian can give you a number tied to your labs and medical history. Tell the agent your target once and it tracks against it on every reply.

Using it day to day

After the first setup, you just talk to it. Tell it "two eggs, toast, and a coffee" in your channel and it works out the calories and macros from what it knows about those foods, asking you only when it cannot make a reasonable estimate, like the size of the portion or how the eggs were cooked. The figures come from standard food databases like the USDA's free FoodData Central, which covers foundation foods, branded products, and restaurant style entries, with a public API if you ever want to cross check a number yourself. Anything that looks off is easy to spot check against the label or the database entry.

Portion ambiguity is where most estimates drift. "A bowl of oatmeal" could mean half a cup dry or two cups with brown sugar and peanut butter mixed in. The agent will ask when the difference matters, and you can answer in the same message thread without opening a separate form. Cooked versus raw weight matters too: 100 grams of raw chicken breast and 100 grams of cooked chicken breast are not the same calorie count because cooking drives off water. If you weigh food on a kitchen scale, say the weight and whether it is raw or cooked, and the log stays tighter.

Then it appends the row and tells you where you stand against your goal for the day. At the times you picked, it comes to you with a short message asking what you have had so far, so a skipped meal does not turn into a skipped week.

Because the goal and the reminder times are things you set at the start, you change them the same way you set them. Tell it to bump your protein target, add a fourth check in, or go quiet on weekends, and it adjusts the spreadsheet and the automation for you.

Asking it questions

The part that a notes app cannot match is what happens once a few weeks of meals pile up. Since the data is a real spreadsheet, you can put questions to it:

• your average protein this week
• which day you went furthest over
• how your calories trended across the month
• whether you hit 150 grams of protein on any day last week

It reads its own file and answers, and because it can send email, you can have it mail you or your trainer a weekly summary without you exporting anything by hand. A summary of totals and averages is usually what a coach actually needs, so you can have it send that rather than the full row by row log of everything you ate, which keeps the granular record in your hands. The log itself is a record of your eating, sitting in the workspace CSV and in the chat thread where you typed each meal, so a private conversation with the agent is the comfortable place to keep it.

The CSV format also means you can pull the file into Excel, Google Sheets, or Python if you want charts the agent has not built yet. The agent keeps writing to the same file, so any external analysis you run is working from the same source of truth.

Why a spreadsheet beats another app

The reason this sticks where apps do not is ownership and friction. The log is a file you control, so a year of meals is yours to keep even if you cancel a subscription or switch phones. Logging drops to sending a text in the channel you already use for everything else, and the scheduled reminder handles the part most people skip, which is remembering to log at all.

The popular apps have steadily moved the basics behind a paywall. MyFitnessPal moved its barcode scanner to Premium in October 2022, and the free tier today is ad supported manual search. Premium runs $19.99 per month or $79.99 per year for barcode scanning, meal photo logging, voice logging, and custom macro goals. In March 2026, MyFitnessPal acquired Cal AI, the photo first calorie app built by teenagers Zach Yadegari and Henry Langmack, in a deal that closed in December 2025. Cal AI continues as a standalone product with its own subscription, and its AI photo estimates now draw on MyFitnessPal's nutrition database. Both apps solve the speed problem well. Neither gives you a plain file you can query in natural language or hand to a coach without an export step.

Open the prompts library and send the calorie tracker to your agent to get started. It asks for your daily goal and when to check in, and from there you log meals by texting the channel you already use.