What is this?
mcd-saver-strategist (aka 'Mai Men Sheng Qian Jing Suan Shi') is an AI Agent Skill built on the McDonald's China MCP server that pulls a store's full menu and every available coupon, then runs a deterministic algorithm to find the cheapest valid order combo — for example turning a 47 yuan Big Mac meal into 21.9 yuan (53% off).
mcd-saver-strategist (aka 'Mai Men Sheng Qian Jing Suan Shi') is an AI Agent Skill built on the McDonald's China MCP server that pulls a store's full menu and every available coupon, then runs a deterministic algorithm to find the cheapest valid order combo — for example turning a 47 yuan Big Mac meal into 21.9 yuan (53% off).Is this a standalone app, or does it plug into an AI agent?It's an Agent Skill (built for WorkBuddy) paired with the McDonald's China MCP server, not a standalone chatbot — the skill is pure Python code with a deterministic optimizer, and an AI agent calls the MCP tools while this logic decides the actual order.
It's an Agent Skill (built for WorkBuddy) paired with the McDonald's China MCP server, not a standalone chatbot — the skill is pure Python code with a deterministic optimizer, and an AI agent calls the MCP tools while this logic decides the actual order.How can I try the optimizer without setting anything up?There's a built-in CLI demo with bundled sample data that runs the entire optimization flow offline, no MCP token required. To use it for real orders you plug the skill into WorkBuddy (or another agent) connected to the real McDonald's China MCP server.
There's a built-in CLI demo with bundled sample data that runs the entire optimization flow offline, no MCP token required. To use it for real orders you plug the skill into WorkBuddy (or another agent) connected to the real McDonald's China MCP server.Do I need a McDonald's MCP token to use it?Only for live menus, coupons, and real orders — that requires a McDonald's China MCP token and account access. The offline CLI demo needs nothing else: it's pure Python standard library with zero third-party dependencies.
Only for live menus, coupons, and real orders — that requires a McDonald's China MCP token and account access. The offline CLI demo needs nothing else: it's pure Python standard library with zero third-party dependencies.Is the pricing data in this repo real-time?No — the menu prices and calorie figures shipped in the repo are example snapshots matching the MCP response structure for demo purposes; actual coupon availability and prices follow the live data returned by the real MCP server for your store and account.
No — the menu prices and calorie figures shipped in the repo are example snapshots matching the MCP response structure for demo purposes; actual coupon availability and prices follow the live data returned by the real MCP server for your store and account.Who is this especially useful for?Price-sensitive students and office workers, people watching calories while dieting, families or groups splitting a multi-person order, and frequent coupon users who want to know which coupons actually save money versus which ones are traps.
Price-sensitive students and office workers, people watching calories while dieting, families or groups splitting a multi-person order, and frequent coupon users who want to know which coupons actually save money versus which ones are traps.What optimization goals besides 'cheapest' does it support?It supports four order-planning goals — lowest price, maximum food quantity, lowest calories, and fewest store trips — always anchoring on the lowest achievable price and only trading up within a tolerance band, so it never suggests a wildly pricier 'more balanced' option.
It supports four order-planning goals — lowest price, maximum food quantity, lowest calories, and fewest store trips — always anchoring on the lowest achievable price and only trading up within a tolerance band, so it never suggests a wildly pricier 'more balanced' option.What does the 'coupon audit' feature do?It checks each coupon against three figures — nominal face value, actual savings, and opportunity cost — to flag cases where a small coupon looks good alone but actually blocks you from a bigger combo deal like the '1+1' meal offer.
It checks each coupon against three figures — nominal face value, actual savings, and opportunity cost — to flag cases where a small coupon looks good alone but actually blocks you from a bigger combo deal like the '1+1' meal offer.How does it decide what to add when I'm close to a discount threshold?It runs a subset-sum dynamic-programming search to find the minimum-cost item that closes the gap to the next discount tier, then reports the net change — which can even be negative, meaning adding an item actually saves you more money.
It runs a subset-sum dynamic-programming search to find the minimum-cost item that closes the gap to the next discount tier, then reports the net change — which can even be negative, meaning adding an item actually saves you more money.Why use a deterministic algorithm instead of letting the AI model estimate the price?All totals are computed by the optimizer or returned directly by the MCP server rather than guessed by the language model, avoiding hallucinated prices. The project backs this with 46 self-tests, 7 CI gates, and 17 checks keeping the web demo and skill logic in sync.
All totals are computed by the optimizer or returned directly by the MCP server rather than guessed by the language model, avoiding hallucinated prices. The project backs this with 46 self-tests, 7 CI gates, and 17 checks keeping the web demo and skill logic in sync.What are similar projects?zhijian-ai-bluebook-workbuddy-harness, nuwa-skill, Kami, awesome-jev-live, china-travel-kit, Agent-Reach, Easel, narrator-ai-cli-skill
zhijian-ai-bluebook-workbuddy-harness, nuwa-skill, Kami, awesome-jev-live, china-travel-kit, Agent-Reach, Easel, narrator-ai-cli-skillWhich topics is it in?workbuddy, agent-skill, mcp-server, ai-agent, mcp, python
workbuddy, agent-skill, mcp-server, ai-agent, mcp, python