Can McDonald's MCP Really Save Money? This Skill Even Tallies the Cash You Are Quietly Wasting

What it does

The "McDonald's Money-Saving Assistant" is an ordering-cost-optimization skill built on McDonald's China MCP (mcp.mcd.cn). It does two things. In the short term, it pulls the real-time menu and prices of your current store, pulls the coupons actually available for your account at that store, enumerates every candidate — single items, combos, coupon combinations, and add-on bundles — then uses the MCP calculate-price call to price each one for real (not a guess), and finally outputs the cheapest way to order, explaining why it does not recommend the cheaper-but-not-filling option. In the long term, it runs the numbers on you: which of your coupons are about to expire (that is money leaking away), whether your recent spending habits push your average ticket too high, and which items should really be swapped for combos. In one line, it answers a question others skip: "how much money are you quietly wasting?"

Why it blew up

Money-saving tools are a dime a dozen, but almost all only answer "how do I order cheapest this time." This project is smart because it takes one more step: it uses MCP to fetch the real store prices and your real coupons, instead of approximating with a hardcoded menu. McDonald's MCP itself is brand new, and tying it to "saving decisions" rode that wave while hitting a genuine pain point — workers who want to eat cheap but hate studying coupon rules. The clever bit is that it has zero third-party dependencies and needs no token to run tests — demo.py reads recorded real MCP responses offline, so anyone can try it instantly. That low barrier drove fast sharing among developers.

Technical highlights

It turns "saving money" into verifiable engineering rather than a slogan. The price source is the MCP calculate-price call, meaning every cent is computed by the system, avoiding the misleading "I guess about 30 yuan" of other tools. It handles four combination dimensions — single items, combos, coupon combos, add-on bundles — and explicitly states why it skips the cheaper-but-not-filling plan. That matters, because pure cheapest-first could hand you a combo you would never actually eat; it treats "being full" as a constraint too.

The long-term ability is more interesting: through order-list history and query-my-coupons balances, it draws you a "silent cost" map — how many coupons are quietly expiring, how much you spend monthly, which habits stealthily raise your average ticket. It upgrades "saving assistant" into a "spending check-up." The whole project uses only the standard library, runs on Python 3.8+, needs no network, and the demo replays real responses from fixtures — clean and self-contained.

Who it is for

If you order McDonald's often but cannot be bothered to study which coupons your store accepts today and how to bundle best, this skill saves you 3–13 yuan per order and flags the tens of yuan in coupons you are letting expire. For engineers building Agent or MCP applications, it is a great "real MCP deployment case" — it shows how to turn an official MCP service into a decision-making assistant rather than a dumb data fetcher. It also works as a template for anyone wanting a personal spending audit. Note it depends on McDonald's China MCP being available; fully offline you only get the demo replay.

Quick start

Zero dependencies, no token needed — clone and run the demo:

git clone https://github.com/Ayzzz123/mcd-coupon-strategist.git
cd mcd-coupon-strategist
python demo.py

demo.py reads the recorded real MCP responses under fixtures/ and runs both the "single-order saving" and "silent cost" chains.

Compared with alternatives

Similar tools in the wild are mostly "coupon aggregators" or "hardcoded menu calculators" — either they just list coupons without deciding for you, or they guess with stale prices. The Money-Saving Assistant differs in three ways: (1) it connects to the live McDonald's MCP, so prices and coupons are real-time real data; (2) it does combination enumeration plus real pricing, not single-point comparison; (3) it adds the long-term "silent cost / spending profile" view. If you only want to see what coupons exist, the official app is enough; if you want "how to order cheapest this time plus where exactly I am leaking money," this open-source skill is currently the most practical pick.

Where it fits and what to watch

A couple of practical realities: the assistant is only as good as McDonald's China MCP availability, so if that service is down or changes its schema, the live path stops working while the offline demo keeps running. The "silent cost" view is genuinely the part I find most useful — most people have no idea how many coupons expire untouched, and seeing it quantified is a small shock. Treat the output as advisory, not gospel: menu prices and coupon rules shift, and the skill's job is to surface the cheapest compliant option, not to guarantee it matches a promotion you heard about elsewhere. For developers, the architecture is the real lesson — it shows how to wrap an official MCP service with a decision layer instead of just forwarding raw calls. That pattern transfers to any domain with a pricing or inventory API.