Amazon Ads runs an official MCP server, in open beta globally since February 2, 2026. It creates, updates and deletes campaigns, runs reporting queries, and reaches account settings and billing. Access is limited to Amazon Ads partners with active API credentials.
Last updated September 21, 2026. Figures are from Amazon Ads's own documentation. New to the protocol? Start with what MCP is.
What the Amazon Ads MCP server exposes
| Surface | What you get |
|---|---|
| Campaigns | Create, update and delete |
| Workflow tools | Multi-step jobs in one call, such as a whole Sponsored Products campaign |
| Reporting | Performance and reporting queries |
| Account | Account-level settings |
| Billing | Billing and financial data |
The endpoint is https://advertising-ai.amazon.com/mcp. You sign in with Login with Amazon. That URL serves North America; accounts elsewhere should use the regional endpoint in Amazon’s documentation. Amazon names Claude, ChatGPT, Gemini and custom-built agents as clients.
Connecting it
1. Open your client's connector settings
In Claude, go to Settings, then Customize, then Connectors. In ChatGPT, turn on Developer mode under Settings, then Security and login, and add it under Settings, then Plugins. If Amazon Ads is already listed in the client's directory, add it from there instead.
2. Paste the endpoint
Add a custom connector with the URL https://advertising-ai.amazon.com/mcp and leave the advanced fields empty. There is no client ID or secret to enter; the server registers itself when you sign in.
3. Sign in and read the scopes
The consent screen is the one place you see what you are granting. Read it, and choose the narrower scopes wherever you are offered a choice.
4. Confirm the tools arrived
Ask the assistant what the connector can do. An empty list after a successful sign-in means the token is not being accepted: disconnect and reconnect to clear the stale grant.
The exact path for thirteen clients is in the MCP client directory.
What it costs
- Amazon publishes no separate price for the server.
- The gate is access, not price: you need to be an Amazon Ads partner with active API credentials.
- It is an open beta, so tools and behaviour can change.
Workflow tools: its best idea
Most platform servers mirror their API one endpoint per tool, which leaves the model to chain calls in the right order. Amazon added tools that do the chaining. Its Sponsored Products tool creates the campaign, the ad groups and the ads from one prompt — work that would otherwise take three or more operations. Fewer calls, fewer chances for the model to get one wrong.
How it compares with the other platforms
Amazon is one of four ad platforms whose official server can change campaigns, alongside Meta, TikTok and X. Meta and X create new campaigns paused; Amazon has not said the same. The full map is in the best MCP servers for paid ads.
The limits worth knowing
- Partners only. A brand without API credentials cannot connect directly. Most will reach it through an agency or a tool that is already a partner.
- Deletes and billing are in scope. A server that can delete campaigns and read financial data deserves the narrowest grant you can give it, and a person reviewing every write.
- It is a beta. Open beta means tool names and behaviour can change. A workflow built on it today should be re-checked after updates.
- Amazon only. It sees Amazon placements and nothing else, so it cannot weigh an Amazon dollar against a Meta or Google one.
Where it stops for marketing
Amazon’s server manages Amazon. Deciding how much of the budget Amazon should get is a cross-channel question. Sprites is an AI marketing agent that covers Google, Meta, LinkedIn, TikTok and Reddit through one MCP server, not Amazon, and holds every change for approval.
That is the edge of what one vendor's data can answer, not a flaw in the server. It is why a marketing stack usually runs two or three: this one for what it is good at, and one for the channels that spend money. For the other half: Every ad platform’s MCP server. If you would rather hand the marketing half to one agent, Sprites runs Google Ads, Meta, LinkedIn and SEO behind one connector, with every write staged for your approval.
Before you grant write access
MCP has no built-in approval step: a write tool applies the moment the model calls it, unless the server itself holds the change. Check three things before you connect: which scopes the consent screen asks for, whether writes are staged or applied, and how to revoke access. The rest is in is MCP safe.
Related
- The best MCP servers for paid ads
- Google Ads over MCP
- Meta Ads over MCP
- Every MCP server worth connecting for marketing
- The Claude connector shortlist
Amazon Ads MCP questions
Does Amazon Ads have an MCP server?
Yes. Amazon Ads has run an official MCP server in open beta since February 2, 2026, available globally to Amazon Ads partners with active API credentials.
What can the Amazon Ads MCP server do?
Create, update and delete campaigns, run performance and reporting queries, manage account settings and read billing data. Workflow tools handle multi-step jobs, such as building a Sponsored Products campaign with its ad groups and ads in one call.
Who can use the Amazon Ads MCP server?
Amazon Ads partners with active API credentials. Advertisers without them typically reach it through an agency or tool partner.
Which AI tools work with Amazon Ads MCP?
Amazon names Claude, ChatGPT, Gemini and custom-built agents.
Is the Amazon Ads MCP server free?
Amazon has published no separate price for it. The requirement is partner access with API credentials.