Claude Inventory Automation POS Integration

I want to put Anthropic Claude at the heart of my retail stock workflow so I can finally stop living inside spreadsheets. My point-of-sale system already captures every sale, return and delivery; what’s missing is the intelligence layer that watches those numbers, predicts what comes next, and quietly places orders before shelves run dry.

Here is the scope I need you to cover:

• Connect Claude to the POS API, pull live SKU counts and sales velocity data in near-real time.
• Build prompt chains (or function calls if you prefer) that let Claude:
– flag low-stock items,
– forecast demand over selectable time windows, and
– draft or trigger automatic purchase orders with preferred suppliers.
• Configure error handling, rate-limit safeguards and logging so I can audit every recommendation Claude makes.
• Package the solution so it can run unattended on my existing cloud instance—Python or Node is fine as long as it’s well-documented.
• Deliver a short hand-over session and quick-start guide so my staff can tweak thresholds without calling a developer.

Acceptance will be based on:
1. Claude consistently reading live inventory from the POS sandbox and the production tenant.
2. Reorder proposals matching ±5 % of historical reorder quantities during a two-week parallel test.
3. Clear, step-by-step deployment instructions working on a fresh VM.

If this feels like a challenge you enjoy, tell me how you would approach the data mapping and which Claude features (e.g., function calling, tool use) you would lean on first.