Amazon's Seller Assistant Now Runs Workflows While You're Logged Out, and Plugs Into Claude. Here's What to Set Up and What to Lock Down First

On September 23, 2026, Amazon announced the next evolution of Seller Assistant: workflows that run around the clock, persistent memory of your business, and a new selling partner plugin that brings all of it into Amazon Quick and, in beta, Claude. One deadline attached to it matters right now, and a few settings matter more than the deadline.

The Problem: Routine Seller Work Never Stops, and You Are the Only Trigger

Amazon frames the issue in its official announcement. Sellers make dozens of connected decisions every day across inventory, advertising, listings, and compliance. The data exists, but stitching it together has been manual work.

Until now, Seller Assistant also had a limit: it acted when you asked. You could say "adjust price for this ASIN" or "draft a restock plan" and it would reason and execute with your permission. But checking inventory thresholds, watching competitor pricing, and reviewing account health alerts follow the same patterns every week. Those tasks are important, repetitive, and easy to forget on a busy day. A missed stockout alert or a slow reaction to a competitor's price move costs real money, and the only thing standing between you and that outcome is remembering to ask.

The second problem is tool sprawl. Amazon says 90% of its selling partners already use third-party AI tools to run their business. Those tools have historically not had clean access to your live Amazon data, which is why sellers end up pasting exports into chat windows.

What Amazon Is Proposing

Amazon announced three connected pieces. Here is what each one does, using the specifics from the announcement.

1. Persistent memory

Seller Assistant now carries memory of your pricing patterns, inventory cycles, and growth goals. Amazon says that memory follows you across Seller Central, Amazon Quick, and Claude, so context is not lost when you switch tools. Every recommendation comes with its reasoning, so you can evaluate or adjust it. Sellers can also open a canvas, a visual workspace for modeling scenarios, comparing pricing strategies, and diagnosing campaign performance with live data.

2. Seller Assistant workflows

These are custom automations that run continuously, monitor conditions, interpret changes, and respond, even when you are not logged in to Seller Central. You describe what you need in plain language and set guardrails. Amazon's examples:

• "Alert me if any of my top 10 products drop below a 4-star rating and draft a response plan."
• "Monitor my top category for competitive openings. If you see one, adjust my pricing and refresh the listings."

Prebuilt templates cover account health monitoring, order summaries, pricing trend analysis, and product expansion opportunities. You can also build from scratch. For any workflow, you choose whether it only surfaces recommendations or takes actions. When it takes actions, Amazon says you review and approve them before they are carried out, and every action is logged with a complete audit trail.

3. The Amazon Selling Partner plugin

The plugin connects your Amazon data, including listing contributions, real-time performance metrics, inventory levels, and sales analytics, to your preferred AI agent, and that agent can act on it the way Seller Assistant does inside Seller Central. It launches in Amazon Quick and in beta with Claude, with more integrations to follow. Amazon says Claude users get Seller Assistant connected in roughly 60 seconds with no coding required. The beta is available to sellers in Amazon's U.S. stores, with international expansion to follow.

Amazon also says plugin interactions are protected by clear boundaries on what can be accessed, human approval on actions, and complete audit trails.

Why Amazon Is Doing This

Amazon's own numbers show the logic. It says Seller Assistant has rolled out to over 90% of selling partners worldwide in their native language, has hundreds of thousands of active users, and sees sellers accept its recommendations over 90% of the time. Its conclusion is that sellers want AI that works for them, not only with them.

The plugin follows from a plain observation: if 90% of sellers already use outside AI tools, Amazon would rather be the trusted data source inside those tools than watch sellers export data to them. Amazon built Seller Assistant on Amazon Bedrock with Anthropic's Claude models, having worked with AWS and Anthropic on it, so putting it in Claude is a natural extension. Amazon says Seller Assistant keeps seller data within Amazon's infrastructure and never shares it outside Amazon.

There is also a business incentive worth naming. The more your daily operating decisions run through Amazon's assistant and Amazon's data, the more your workflow depends on Amazon, whichever front end you use. That is not a criticism, it is the reason to configure this deliberately. It fits the pattern we covered in our piece on why AI shopping is moving from free to paid: Amazon is investing heavily in agents on both the shopper side and the seller side.

How Sellers Will Be Impacted

Small teams get real leverage. A lean operation can now monitor ratings, inventory, and pricing continuously without adding headcount. Amazon says beta testers report meaningful time savings. James Oliver, founder of Atlas Bar, said connecting Amazon data to AI tools used to take an afternoon and now happens instantly. That is one testimonial supplied by Amazon, so treat it as a signal, not a benchmark.

Automation shifts your risk, it does not remove it. A workflow that "adjusts my pricing and refreshes the listings" can touch margin, Buy Box position, and listing content while you sleep. Amazon says you can require approval before actions execute, and that is the setting to think hardest about. A pricing rule with loose guardrails can chase a competitor's mistake down to a loss.

Your account access widens. Connecting a plugin to an outside AI agent is a new access path into your seller data. Amazon describes boundaries, approvals, and audit trails, but who inside your company can connect the plugin, and to which AI account, becomes a real internal policy question.

Memory helps, and you should check what it learned. Persistent memory of your pricing patterns and growth goals will make recommendations sharper over time. It will also carry any bad habit or outdated assumption forward. Review recommendations with the reasoning Amazon says accompanies each one.

There is a free offer with a deadline. Every primary account holder globally can sign up for a free 12-month Amazon Quick Plus subscription through December 31, 2026, and can designate two additional co-workers to receive a free year at no extra cost.

What Sellers Should Do This Week

1. Claim Quick Plus if you plan to use it. The offer runs through December 31, 2026. As a primary account holder, decide which two co-workers get the extra seats.
2. Start with recommend-only workflows. For every workflow, prebuilt or custom, begin with the setting that surfaces recommendations instead of acting. Move to action mode only after you have watched it for a few weeks.
3. Use the templates first. Turn on the account health monitoring and order summary templates before writing custom automations. They are lower risk and teach you how the assistant reasons.
4. Write pricing guardrails as numbers. If you ever let a workflow adjust price, set a hard floor tied to your real landed cost and margin, a maximum daily change, and a list of ASINs that are off limits.
5. Keep approval on for anything that changes a listing. Titles, bullets, and images affect conversion and compliance. Keep a human in the loop, especially given how many listing changes Amazon is already making on its own, as we covered in our look at Seller Assistant and the Featured Offer.
6. Set an internal rule for plugin access. Decide who may connect the plugin, to which AI accounts, and where you will review the audit trail. If you connect Claude, use an account your company controls.
7. Review the audit trail on a schedule. Amazon says every action is logged. A weekly 15-minute review is the difference between automation and unattended automation.
8. Keep your own numbers. Continue to track margin, ACOS, and inventory in your own system so you can tell whether the assistant's recommendations actually improve results.

Where This Fits in Amazon's Broader Seller Push

This launch lands the day before Amazon's multichannel announcement. Read together, they show a clear direction: Seller Central becoming the operating layer for the whole business, with AI on top. We covered the other half in Amazon's new multichannel selling tools. Amazon also recently announced an Enhance My Listing tool for generative AI listing improvements and a Seller Wallet for managing when and how much funds convert between USD and EUR, which round out the same push.

Frequently Asked Questions

What is the difference between Seller Assistant and Seller Assistant workflows?

Seller Assistant answers questions and acts when you ask, with your permission. Workflows are automations you define once that keep running, monitor conditions, and respond even when you are not logged in to Seller Central.

Can a workflow change my prices or listings without my approval?

For any workflow you choose whether it only surfaces recommendations or takes actions. Amazon says that when it takes actions, you review and approve them before they are carried out, and every action is logged with an audit trail.

Which AI tools does the plugin work with?

The Amazon Selling Partner plugin launches in Amazon Quick and in beta with Anthropic's Claude, with more integrations planned. It is available in beta to sellers in Amazon's U.S. stores, with international expansion to follow.

Who qualifies for the free Amazon Quick Plus subscription?

Amazon says every primary account holder globally can sign up for a free 12-month Amazon Quick Plus subscription, available through December 31, 2026. Primary account holders can also designate two additional co-workers to receive a free year.

Does using the Claude plugin send my data outside Amazon?

Amazon says Seller Assistant keeps seller data within Amazon's infrastructure and does not share it outside Amazon, and that plugin connections are protected by access boundaries, human approval on actions, and audit trails. Review the plugin's permissions when you connect it, since you are deciding what a third-party AI account can see and do.

Takeaways

• Amazon announced Seller Assistant workflows that monitor and act around the clock, persistent memory across Seller Central, Amazon Quick, and Claude, and a selling partner plugin in Quick and in beta with Claude.
• You choose whether each workflow only recommends or also acts, and Amazon says actions require your approval and are logged in a full audit trail.
• The plugin beta is available to U.S. sellers, and a free 12-month Amazon Quick Plus offer runs through December 31, 2026.
• The real risk is loose guardrails and widened access, not the feature itself.
• Start with recommend-only workflows, hard price floors, approval on listing changes, an internal plugin access policy, and a weekly audit trail review.

Ongoing coverage of Amazon policy and platform changes runs in our newsletter.

Sources

• Amazon: Amazon gives sellers an even smarter Seller Assistant and a new plugin for Amazon Quick and Anthropic's Claude
• Amazon: Agentic AI across the seller experience
• Amazon: Amazon Quick
• Amazon: Seller Assistant canvas
• Amazon: AWS and Anthropic
• Amazon: Enhance My Listing
• Amazon: Seller Wallet