Amazon runs its own Ads MCP server and charges no additional fees for its Ads API. What the $39–$100-a-month MCP services add, the two things to check before you give an AI your ad account, and why we don't offer one.
What an MCP actually is
MCP stands for Model Context Protocol. It's a plug that lets an AI app such as Claude, ChatGPT or Gemini reach into another system, read from it, and, if you allow it, act on it. For a seller, that means typing "which keywords lost money last week?" into an AI chat and getting the answer straight from your ad account. Or typing "pause them" and having it happen.
That's genuinely useful, and it's why MCP is being sold as the next big thing.
Amazon already built one
On February 2, 2026, Amazon Ads put its own MCP server into open beta. In Amazon's words, it connects "custom-built agents or AI platforms such as Claude, ChatGPT, or Gemini" to the Amazon Ads API. It can create, update and delete campaigns, run performance reports, manage account settings, and access billing and financial data.
On cost, Amazon's API page says: "There are no additional fees from Amazon Ads to use the API." The catch isn't money, it's access. The server is "available globally in open beta to Amazon Ads partners with active API credentials," and API access "requires an application and approval process." That's paperwork most sellers have never done.
What the $39 to $100 versions sell
Several companies now sell MCP access to Amazon sellers as a subscription. The ones I checked charge between $39 and $100 a month, some of them on top of their main product. What you're paying for is mostly two things: they've done the API paperwork for you, and they connect data that Amazon's ads server doesn't — Seller Central sales, inventory, fees and cash flow.
That can be worth paying for. Just know what you're buying: convenience and extra data, on top of a kind of connection Amazon now offers at no extra cost.
Before you give an AI your ad account
Two questions matter more than the price.
Where does your data go? When an AI app reads your account through an MCP, your sales, ad spend and financial numbers become part of that app's conversation. Whether the app keeps them, for how long, and whether they can be used to train its models depends on the app and your settings, not on the MCP. Check those settings before you connect anything. If you can't find the answer, that is an answer.
What can it change? Amazon's own server can create, update and delete campaigns. Some MCP services ask your approval before acting. Some AI apps ask before each action and let you click "always allow." Decide in advance what an AI may do without you. A bid can be put back tomorrow. An archived campaign can't be brought back.
Why we don't offer one
People ask, so here's the answer. We hold your Amazon data under Amazon's data-protection rules for developers, and every company that processes it for us is under contract. The AI inside RedHen Labs runs on Anthropic's commercial API. Their terms bar them from training models on that data, it receives keyword and performance numbers rather than your buyers' details, and it keeps them for no longer than 30 days.
An MCP server would hand the same data to whichever AI app you plug in, under settings we can't see and can't vouch for. We're not willing to make promises about your data on someone else's behalf, so we don't offer one. If you want an MCP, use Amazon's. It's Amazon's own connection to your own account, and it costs nothing extra.
What I use instead
I don't run an MCP on my own account. I use the AI built into RedHen Labs. I ask it the same questions — what lost money last week, which search terms to cut, which bids to raise — and when it wants to change something, the change lands in my approval queue. It can't send anything to Amazon on its own.
That isn't because AI can't be trusted with a bid. It's because I've had automated bidding run away with my account before, and I want to see every change before it ships.