Amazon MCF explained: how multi-channel fulfillment orders move, what they cost, where they show up in your data, and how to run them from Claude.
Amazon Multi-Channel Fulfillment (MCF) lets you ship orders from your own website, another marketplace or a social shop out of the same FBA inventory that serves your Amazon orders. You send Amazon the order, Amazon picks, packs and ships it, and the customer never has to know where the stock sat.
This guide covers how an MCF order moves, what it costs, the packaging and channel rules worth checking first, where MCF orders show up in your data (and where they deliberately don't), and how to preview, create and track MCF orders from Claude with DataDoe.
With FBA, Amazon fulfills orders placed on Amazon. With MCF, Amazon fulfills orders placed anywhere else, from the same pool of units in its fulfillment centers. There is no separate inventory to send in and no second warehouse to manage. The trade-off is cost and control: you pay a per-unit fulfillment fee on each MCF order, and Amazon decides how the parcel travels within the speed you choose.
Two details catch people out. The order ID is yours, not an Amazon order ID, so keep it unique and traceable to the order on your store. And an MCF order is cancellable only until Amazon starts working on it.
The MCF fee is charged per unit and depends on the item's size and weight, the number of units in the order, and the shipping speed. Faster speeds cost more, and multi-unit orders cost less per unit than single-unit ones. Storage is not separate: MCF units sit in the same FBA inventory and pay the same storage fees.
Rather than reading a rate card, ask Amazon for the exact figure on a real basket before you commit. An order preview returns Amazon's estimated fees for those SKUs, that address and that speed. A delivery-offer lookup shows which speeds and delivery windows are available for a SKU. A dedicated post on MCF fees, with worked examples, follows later in this series.
Some marketplaces do not accept Amazon-branded boxes or Amazon's own delivery network on their orders. Amazon offers unbranded packaging for MCF, and you can steer delivery when a channel requires it. Read each channel's fulfillment policy before you route its orders through MCF, because the rule belongs to the channel, not to Amazon.
It also pays to check sellability per SKU. Not every FBA unit can ship through MCF: hazmat items, some categories and some regions have limits. A sellability check returns, for each SKU, whether Amazon will fulfill it and in what quantity.
This is the part most tools gloss over, and it decides whether your numbers mean anything. In DataDoe, MCF orders appear in three places, each with its own rule:
amazon_order_items_with_cogs). Off-Amazon MCF lines belong to the seller account, not to a marketplace, so their marketplace columns are empty. sales_channel keeps Amazon's value, such as Non-Amazon. The price on those lines is a declared or fulfillment amount, not what your customer paid on your store. If you export several marketplaces of one account and stack the files, the same MCF line can appear more than once.amazon_notifications_fulfillment_order_status). Amazon's order, shipment and package status changes for MCF, available within about two hours of Amazon publishing them. Tracking numbers and estimated delivery times are there only when Amazon sends them. Returns are not part of this feed.amazon_mcf_fulfillment_coverage). One row per shipment item Amazon reported as shipped: SKU, quantity, facility, event time. It counts confirmed MCF units and nothing else.And one place where they are left out on purpose: the profit tables exclude MCF lines from sales, units, orders, tax and product cost. Amazon never sees the revenue from your store, so any "MCF profit" built only from Amazon data would mix a declared amount with real fees. The coverage table says so in its own columns: external revenue, product cost and profit are marked unavailable, and unavailable is not zero.
Two practical consequences:
With DataDoe connected to Claude, the whole MCF lifecycle is available as Actions, for Seller Central accounts:
A typical conversation:
DataDoe does not connect to your store. The order details come from you: pasted into the chat, or passed along by your AI client from wherever it already reads your orders. Everything after that runs on your Amazon account.
Creating, releasing and cancelling orders are write Actions. Actions stay off until you enable them, most AI clients ask before a change, and every change is logged. For an MCF order to go out, an organization owner turns on its Action type, and the Seller account is set to "Read and write" in DataDoe. Previews, delivery offers and sellability checks are read Actions and work with "Read only" access.
Once the notifications table is on, the questions you used to answer by clicking through orders one by one become a single prompt:
For a connected India Seller account shipping to an address in India, MCF orders can be paid on delivery. Prepaid stays the default. Preview the option first: Amazon marks each offer as enabled or disabled for payment on delivery. To create such an order, every line needs its own price in Indian rupees. DataDoe rejects the option for any other marketplace or destination.
The full list of write operations lives on DataDoe Actions. For the fees that MCF units share with FBA, see the FBA fees breakdown and Amazon storage fees explained. To keep MCF volume in your stock plan, see FBA restock planning.
Multi-Channel Fulfillment is Amazon's service for shipping orders from other sales channels out of your FBA inventory. You create the order, and Amazon picks, packs and ships it.
Yes, once the MCF status notifications table is enabled. Tracking numbers appear when Amazon includes them in its shipment and package updates.
No. Profit tables leave MCF lines out of sales, units and product cost, because the revenue from your other channels never reaches Amazon. MCF units and statuses are in their own tables.
Yes, through DataDoe Actions on a Seller Central account, after the Action type is enabled and the account has "Read and write" access. Start with a preview and a dry run.
No. DataDoe works on the Amazon side. The order details come from you or from your AI client.