Amazon MCP: AWS, Ads, or Seller Central?
“Amazon MCP” is three different products, and most people searching for it want the one they didn’t mean. The term collapses AWS cloud infrastructure, Amazon advertising, and Amazon seller data into one phrase — and the answer is completely different for each. There is an official AWS MCP Server (generally available since May 2026). There is an official Amazon Ads MCP server (open beta since February 2026). And for Seller Central orders and inventory — the thing most e-commerce people assume “Amazon MCP” means — there is no official Amazon MCP server, only third-party builds sitting behind a genuinely difficult API gate.
This post sorts out which one you’re after, what each actually does, and the limit all three share.
Amazon is one of many — the complete list of MCP servers covers every app that ships one, grouped by category.
Which “Amazon MCP” you’re looking for
Match your job to the row:
- You want an AI coding agent to read your AWS account, call AWS APIs, or look up AWS docs → the AWS MCP Server. Official, GA, free. Most search traffic for “amazon mcp” lands here.
- You want to create or optimize Amazon advertising campaigns by chat → the Amazon Ads MCP server. Official, open beta, requires Amazon Ads API credentials.
- You want orders, FBA inventory, listings, or sales numbers out of Seller Central → no official MCP server exists. Third-party servers do, but you supply your own SP-API access.
- You want to look up or buy products on amazon.com → community projects only, and none of them are sanctioned by Amazon.
The AWS MCP Server
AWS shipped a managed, remote MCP server and made it generally available on May 6, 2026. It’s hosted by AWS at regional endpoints — https://aws-mcp.us-east-1.api.aws/mcp for US East, with Europe (Frankfurt) also available — and it can make API calls into any region.
What’s interesting is how few tools it exposes for how much surface it covers:
call_aws— executes any of the 15,000+ AWS API operations using your existing IAM credentials. One tool, the whole API.search_documentationandread_documentation— pull current AWS docs at query time, so the agent isn’t reasoning from a stale training snapshot. These work without authentication.run_script— runs Python in a sandbox that inherits your IAM permissions but has no network access and no reach into your local filesystem or shell.
Auth is IAM SigV4 rather than OAuth, which MCP doesn’t natively speak, so AWS publishes an open-source MCP Proxy that bridges the two. Access is expressed as a normal IAM policy with context keys, CloudTrail captures the calls, and CloudWatch publishes metrics under an AWS-MCP namespace so you can tell agent traffic apart from human traffic. There’s no charge for the server itself — you pay for the resources it creates.
Separately, AWS maintains awslabs/mcp, an open-source collection of 40+ purpose-built servers for individual services: DynamoDB, Bedrock, Lambda, EKS, CloudFormation, Redshift, Neptune, SageMaker and more. Most install via uvx with one-click buttons for Claude Code, Cursor, VS Code, Kiro, and Windsurf. The managed AWS MCP Server is the general-purpose front door; the awslabs collection is what you reach for when you want deep, service-specific tooling.
The Amazon Ads MCP server
Amazon Ads runs its own MCP server, which moved to open beta on February 2, 2026 and is available globally to Amazon Ads partners with active API credentials. It works with Claude, ChatGPT, Gemini, and custom agents.
It’s more opinionated than a thin API wrapper. Rather than exposing one tool per endpoint, it ships orchestrated workflows that bundle multi-step jobs — launching an end-to-end Sponsored Products campaign, expanding an existing campaign into new countries, generating reports, handling account setup and financial data — into single operations. That’s a sensible design choice for an API as fiddly as Amazon Ads.
The catch is the entry requirement: you need active Amazon Ads API credentials, which is a partner-level thing, not something a solo seller flips on in their account settings. And it’s ads data only. Your inventory, orders, and profitability aren’t in it.
Seller Central has no official MCP server
This is the honest part, and it’s the one aggregator blogs blur. Amazon has not published an MCP server for the Selling Partner API. Every “Amazon Seller MCP” you’ll find on GitHub or an MCP directory is third-party — some from vendors like Seller Labs or Zapier, some from individual developers — and all of them require you to bring SP-API access.
Getting that access is not a five-minute job. Amazon’s own registration flow runs through the Solution Provider Portal and involves administrator identity verification, business and contact information, and a data access assessment. You register an application, retrieve Login with Amazon (LWA) client credentials, and then wait for role approval — Amazon documents processing as taking up to 10 business days. Roles covering personally identifiable information get scrutinized harder.
None of that is impossible, but it is a real gate, and anyone telling you to “just point Claude at your Seller Central data” is skipping the part that takes two weeks.
The limit all three share
Set the auth differences aside and every one of these servers behaves identically in the way that matters most: it answers when spoken to and does nothing otherwise.
MCP has no trigger mechanism and no scheduler. Its own Triggers and Events working group opens by conceding that clients learn about server-side changes by polling, and that work is still listed as on the horizon rather than shipped.
So an MCP connection to AWS can tell you which EC2 instances are oversized — when you ask. The Ads MCP can adjust a bid — when you ask. A seller MCP can report FBA stock — when you ask. None of them notice a spike in ad spend at 2am, a stockout on your best ASIN, or a CloudWatch alarm, because noticing isn’t something the protocol does.
The unattended-work gap closed in 2026: ChatGPT Scheduled Tasks and Claude Cowork both run on their own. The remaining gap is what starts them. These run on a timer, not in response to an event in your apps, so "when this happens, do that" is still outside what they do.
There’s a second limit worth naming: one server, one app. Getting an inventory alert into Slack and a Google Sheet means standing up and authorizing a separate MCP server for each, then hoping your client can hold them all in one turn. If you’re evaluating this across clients, ChatGPT’s MCP support and Codex’s have their own quirks on top of these shared ones.
Amazon work that runs without a chat window
This is where Carly fits, and it’s worth being precise about which Amazon it covers: Carly connects natively to Amazon Seller Central through SP-API — orders, FBA inventory, product listings, and sales analytics. Not AWS infrastructure, not Amazon Ads. If your job is the seller-data one that has no official MCP server, that’s the gap Carly fills.
The difference from MCP is triggers and schedules. Carly’s workflows fire on events and clocks, so the work happens whether or not anyone has a conversation open:
- When FBA stock for an ASIN drops below your threshold → post to
#opsin Slack, add a row to the reorder sheet, and email the supplier. - Every morning at 7 → summarize yesterday’s orders and revenue and send it before the standup.
- When an order ships → log it to your accounting tool and update the customer record in your CRM.
Because it reaches ~260 apps natively plus anything with a public API through your own key, the Amazon signal doesn’t dead-end in a chat reply — it moves into the tools where the follow-up work actually happens. Pricing: free, unlimited Zapier-style workflows; AI agents from $35/month.
FAQ
Does Amazon have an official MCP server? It depends which Amazon. AWS publishes an official managed MCP Server, generally available since May 2026, plus 40+ open-source servers under awslabs/mcp. Amazon Ads publishes an official MCP server in open beta. Amazon has not published an official MCP server for Seller Central or the Selling Partner API — those are all third-party.
Is the AWS MCP Server free? Yes, there’s no additional charge for the server itself. You pay only for the AWS resources it creates and any applicable data transfer costs. It authenticates with your existing IAM credentials, so your normal IAM policies control what it can touch.
Can I connect Amazon Seller Central to Claude or ChatGPT with MCP? Only through a third-party server, and only if you obtain your own SP-API access first — which means registering an application in Amazon’s Solution Provider Portal, getting LWA credentials, and waiting for role approval that Amazon says can take up to 10 business days.
Can an Amazon MCP server alert me when I’m about to stock out? No. MCP has no triggers and no scheduler — a server responds to requests inside a chat and can’t notice anything on its own. Stockout alerts need an event-driven workflow tool watching the account, not an MCP connection waiting to be asked.
What’s the difference between the AWS MCP Server and awslabs/mcp?
The AWS MCP Server is one managed, AWS-hosted remote server that covers the whole AWS API through a call_aws tool. The awslabs/mcp repository is a collection of 40+ individual open-source servers you run yourself, each built for a specific service like DynamoDB, Bedrock, or EKS with deeper service-specific tooling.
Ready to automate your busywork?
Carly schedules, researches, and briefs you—so you can focus on what matters.
See what people say
"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.
Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.
On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."
