If you sell on Amazon, Walmart, or eBay, price is the single biggest lever on your buy-box share — and it moves by the minute. Managing that from spreadsheets or a disconnected third-party repricer means your ERPNext Item costs, landed freight, and margin targets live in one system while the actual selling price is decided in another. ERPNext core has Item Price, Price List, and Pricing Rules, but they are static: nothing watches your competitors, nothing understands the buy-box, and nothing safely pushes a new price back to the channel every few minutes without a human in the loop. That gap is where sellers either leave the buy box on the table or race to the bottom and erode margin.

The ERPNext AI Repricing Engine is a proper Frappe app (installable on your bench, v15 and v16 supported) that closes that gap inside ERPNext instead of alongside it. We model the domain as first-class DocTypes — a Marketplace Listing linking your Item to its per-channel SKU/ASIN, a Repricing Strategy holding your rule set (beat-lowest-by, match-buy-box, hold-if-you-own-it), a Competitor Price log capturing observed offers per channel, and a Repricing Run audit record for every price change with old price, new price, the winning rule, and the reason. Margin-floor and ceiling guardrails are enforced at the DocType validation layer, so no rule — and no AI suggestion — can ever push a price below the floor you set against the item's landed cost.

Technically, the engine runs on Frappe's own primitives. scheduler_events in hooks.py drive the polling cadence (every N minutes, hourly, daily) to pull competitor offers and recompute target prices; long pulls and bulk pushes run as background jobs on the Frappe worker queue so the UI never blocks. doc_events hooks fire on cost or strategy changes to recheck floors immediately. The AI layer takes demand signals — velocity, buy-box win rate, price elasticity across your own Repricing Run history — and produces a suggested price that the rule engine then clamps to your guardrails; suggestions can run in advisory mode (queued for approval) or fully autonomous per strategy. Every channel push goes out through the marketplace API and is written back as a Repricing Run, and whitelisted methods exposed over the Frappe REST API let your other systems trigger a reprice, read the current target, or pull the audit trail. Client Scripts add one-click "reprice now" and live competitor context on the Listing form; Role Profiles and DocType permissions keep the margin-floor and autonomous-mode controls to authorized users only.

Because this is build-to-order, we start from your real catalog and channels, not a generic template. After a short scoping call we confirm which marketplaces, which repricing rules, your margin-floor logic, and your approval workflow, then build the app against Frappe/ERPNext v15 or v16 to match your bench. You get the installable source, a UAT pass on a staging site, and a rollback plan before anything touches production. Typical delivery is one working week from confirmed scope, and you receive the full git repository so the app is yours to run, extend, or hand to another team.