Table of Contents

  1. What Amazon Announced
  2. Who Can Use It Right Now
  3. What the Plugin Reads
  4. It Can Act, With Your Approval
  5. Seller Assistant Is the Engine
  6. The Free Amazon Quick Plus Year
  7. What It Reportedly Does Not Cover
  8. The Gap for Arbitrage Sellers
  9. Running a Sourcing MCP Next to It
  10. A Morning With Both Connected
  11. Guardrails Before You Approve Actions
  12. Next Steps

1. What Amazon Announced

At Accelerate in September 2026, Amazon announced the Amazon Selling Partner plugin. It connects your Seller Central data to an AI assistant, launching first in Amazon Quick and in beta with Anthropic's Claude (About Amazon announcement).

Amazon says more integrations are coming. Today, the two places it runs are Quick and Claude.

This sits on top of MCP, the standard that lets an AI app call outside tools. If that term is new, start with what an MCP server is and come back.

2. Who Can Use It Right Now

The plugin is in beta for sellers in Amazon's US stores. Amazon says international expansion follows later.

Amazon also says connecting Amazon to Claude takes about 60 seconds and needs no code. You sign in, approve the connection and your Seller Central data shows up in the chat.

If you sell in the US, you can try it now. If you sell only in the UK, EU or elsewhere, wait for Amazon to announce your marketplace.

3. What the Plugin Reads

Amazon lists four kinds of data the plugin connects:

  • Listings you contribute to
  • Real-time performance metrics
  • Inventory levels
  • Sales analytics

That's your account, seen from the inside. You can ask "which of my SKUs sold fastest this month?" or "what's sitting with no sales in 60 days?" and get answers from your own numbers instead of an export you cleaned up by hand.

Pair the slow-mover question with your sell-through rate targets and you get a weekly inventory review without the spreadsheet.

4. It Can Act, With Your Approval

The plugin goes past reading. Amazon says the AI can take actions on your Amazon business, inside defined limits on what it can access.

Two protections come with that: a human approval step before an action runs, and a full audit trail. Seller Assistant workflows follow the same pattern. For each workflow you choose whether it only recommends or also acts, and actions wait for your review before they run.

Keep approvals on. A pricing change you didn't read is still your pricing change, and the audit trail tells you what happened after the fact. It doesn't undo it.

5. Seller Assistant Is the Engine

Amazon built Seller Assistant on Amazon Bedrock with Anthropic's Claude models, per the same announcement. The new version keeps a memory of your business, such as pricing patterns, inventory cycles and growth goals, and Amazon says that memory carries across Seller Central, Amazon Quick and Claude.

Workflows run on their own schedule, even when you're not logged in. Amazon's prebuilt templates cover account health, order summaries, pricing trends and product expansion, and you can describe custom ones in plain language with your own guardrails.

Every recommendation includes its reasoning. Read it. If the reasoning doesn't match how you run your account, reject the recommendation and tighten the guardrail.

6. The Free Amazon Quick Plus Year

Amazon is giving every primary account holder a free 12-month Amazon Quick Plus subscription. You have to sign up by December 31, 2026.

Primary account holders can also name two co-workers, and each gets a free year too. If you have a VA or a partner on the account, that covers them.

Quick is Amazon's general AI work assistant. It writes documents and spreadsheets, does research and runs on phone, web and desktop. With the plugin, Amazon says you can mix your Amazon data with financial or marketing data and build agents like a pricing, listing or restock agent. Quick asks for approval before it acts.

7. What It Reportedly Does Not Cover

Amazon's announcement doesn't spell out every gap, so treat this section as reported, not official. Third-party analyses say the plugin doesn't cover Amazon Ads data or margin data.

The Ads gap has a separate fix: the Amazon Ads MCP Server, in open beta since February 2, 2026, which needs Amazon Ads API access.

The margin gap is harder. Amazon knows your sale price and fees. It doesn't know what you paid at the retailer, your cashback or your prep cost. Without buy cost, nothing inside Seller Central can tell you real profit. Get that math right with the ROI calculation formula.

8. The Gap for Arbitrage Sellers

The plugin looks inward at products you already sell. An online arbitrage seller spends most of the day looking outward: what competitors added, what's in stock at which retailer, and whether your account can even list a product before you buy it.

"And if you don't know what Arbitrage Stalker is, it's a software that I created to help you track sellers, right? So you don't have to do the storefront stalking yourself. So whenever they add a new product, you get a notification."

I made $347 in 5 minutes with Online Arbitrage | Live Amazon FBA Product Research

That's the sourcing side, and Amazon's plugin isn't built for it. Competitor storefronts, source retailers and pre-buy eligibility live in a different set of tools. The manual version of this is reverse sourcing.

9. Running a Sourcing MCP Next to It

Claude can hold more than one connector at a time. Run Amazon's plugin for your account and a sourcing server for buying decisions in the same chat.

I built Arbitrage Stalker for the sourcing half. The Arbitrage Stalker MCP reads your tracked sellers' new products, runs profit at your buy cost with your saved prep and inbound costs, checks eligibility on up to 100 ASINs against your account and tells you where a product has been seen for sale. It's on every paid plan, starting with Scout at $29 a month.

Setup takes a few clicks in claude.ai, covered step by step in connecting Claude to your Amazon seller data. If you'd rather watch the whole sourcing loop first, join the free Thursday training.

Free Live Training

Watch me run this system live every Thursday

Every Thursday at 8 PM EST I run a free 60-minute training where I source, analyze, and ship a real product. Reserve a seat and watch the whole thing.

Reserve My Free Seat →

10. A Morning With Both Connected

A sample routine with both connectors on:

  1. Ask Amazon's plugin which SKUs are running low and which haven't sold in 60 days.
  2. Ask the sourcing connector what your tracked sellers added since yesterday, filtered to the categories you're restocking.
  3. Run "is this worth buying at $14?" on the two or three that look good.
  4. Check eligibility on the shortlist before you place an order.

Step 4 matters most. If a brand needs approval, the sourcing side hands you Amazon's approval link instead of submitting a request for you, since failed attempts count against you. The ungating guide covers how to apply.

11. Guardrails Before You Approve Actions

The approval step only protects you if you read what you approve. Before you let any AI touch prices or listings:

  • Set a floor price for every SKU it can reprice, so a race to the bottom stops at your minimum profit
  • Start with recommend-only workflows for two weeks and compare its calls to yours
  • Check the audit trail every week until you trust the pattern

"Starting today, I will not be using the AI repricing rules on BQool anymore in my online arbitrage business. In today's video, I'm going to tell you exactly what I'm going to do instead of doing this and why I'm stopping."

Why I Quit Using Bqool's AI Repricing Rules - My New Strategy Revealed

I stopped using BQool's AI repricing rules in my own business and switched to a different approach. Hold Amazon's agent to the same standard. The OA repricer guide covers floor and ceiling settings.

12. Next Steps

The plugin covers your account. These five posts cover the rest of the system around it: