A CLI that plans your McDonald's party group order, with real prices

What it does

mcd-party-planner is a Python CLI and WorkBuddy Skill that answers the surprisingly fiddly question of feeding a group at McDonald’s: which nearby store is open, what to order for N people, and what it actually costs within budget. It pulls live menus and pricing from McDonald’s official MCP server (mcd-mcp), assembles economy/balanced/lavish combos, and returns several comparable plans with totals and per-person costs. It also chains a five-step query flow for booking themed birthday parties, and cross-references the campaign calendar with nearby stores to suggest when to organize.

The interesting bit

The discipline around data honesty is the real craft: every price comes from the official calculate-price tool, and when a combo hits a business rule like “limit 1 per order,” it skips and flags it rather than fabricating a number. Write operations (ordering, address creation, cancellation) default to dry_run and never auto-retry on 5xx — only 429s get exponential backoff. The repo contains zero static menu or price snapshots, so nothing can silently go stale.

Key highlights

• Zero dependencies — pure Python 3.9+ standard library, no pip install needed
• Built-in sliding-window rate limiter (300/min default vs. the official 600 cap) with 429 backoff
• 99 offline unit tests run in CI across Python 3.9–3.13; real-MCP smoke tests run locally with a token
• A probe command pulls the server’s inputSchema into docs/mcp-schema.json as a single calibration point for tool parameters
• User input stays in memory; only city and keyword are sent to the MCP endpoint, nothing else leaves the machine

Caveats

• It’s a contest entry (McDonald’s Programmer’s Day 2026), so expect it to be a showcase rather than a long-term maintained product
• CI green means offline unit tests only — real-endpoint verification requires a manually configured token and the local smoke test
• Server-side menus and prices rotate daily, so outputs differ between runs; the README is upfront that this is normal, not simulation

Verdict

Worth a look if you organize team meals or want a clean reference for orchestrating many MCP tools responsibly (throttling, dry-run writes, honest pricing). If you were hoping for a general-purpose restaurant planner, this is McDonald’s-only glue — well-made glue, but glue.