Doe gives you the complete Amazon data layer — every metric, clean and AI-ready — so your AI runs on your terms.
Two MCP servers. Same prompts. Same Claude account. One returned charts and insights. The other returned zeros and burned the token limit.
Two MCP servers. One Claude Pro account. Same Amazon seller, same questions. The official Amazon Ads MCP and DataDoe MCP, head to head.
The official Amazon Ads MCP Server connects to ads data only. No sales, no orders, no inventory, no profitability, no Buy Box. Any question that crosses ads and operations dies on first contact.
The architecture doesn't support how teams use AI.
Three structural issues, each with a workaround in DataDoe.
Amazon's MCP exposes the Ads API as dozens of granular tools — find profile, create report, list reports, fetch report, parse rows. To answer a single question Claude has to call several of them. Each call triggers a permission prompt in the chat.
"We need to approve some operations in order to continue… okay, we also have to approve this one… okay, we need to approve it."— from our side-by-side test
The question requires joining ad data with order data. Amazon's MCP only sees ads. Without SP-API, it has no way to answer — and instead of saying so, it returned zeros and looked like it succeeded.
"It seems like the Amazon is returning zeros, meaning it actually couldn't do the job."— from our side-by-side test
Amazon's reporting is asynchronous: ask anything beyond a real-time metric, and Claude has to create a report, poll, then download. Each step burns tokens. Failures retry. Pro plans hit the wall fast.
"I'm actually afraid we'll run out of the tokens soon… And it hit the limit."— from our side-by-side test, on Claude Pro
Both MCP servers connect AI tools to Amazon. That's where the similarity ends.
Whether you use Claude, ChatGPT, Cursor, Codex, or any MCP-compatible client — DataDoe MCP plugs in the same way. Add it alongside Amazon's, or instead of it, and try yourself.
Authorize SP-API and Ads API in three clicks at app.datadoe.com. Multi-marketplace, multi-account, no engineering required.
Generate the token in DataDoe settings. One token covers all your Amazon data — sales, ads, inventory, vendor.
Add to your AI tool's MCP config, restart, and ask anything. 7-day free trial, no credit card.
The Amazon Ads MCP server is the official Model Context Protocol server from Amazon Ads. It lets AI clients like Claude, ChatGPT, and Cursor query Amazon advertising data through a defined set of tools — campaigns, keywords, reports, billing. It launched in open beta in early 2026 and is the first official MCP integration Amazon has shipped for sellers and agencies.
Only Amazon Advertising API data — Sponsored Products, Sponsored Brands, Sponsored Display, Amazon Marketing Cloud, account settings, and billing. It does not connect to the Selling Partner API, so orders, FBA inventory, listings, reviews, vendor data, and Buy Box status are out of scope.
Yes. The MCP server itself is free for any seller or agency with an Amazon Ads account. You only pay for the AI tool you connect it to — Claude Pro, ChatGPT Plus, Cursor, or another client — and for the tokens those tools consume during conversations.
You authorize OAuth against Amazon Ads, point your MCP-compatible client at the endpoint, and approve the connection. Each ad profile is authenticated separately, so multi-marketplace sellers go through the flow once per region (US, EU, UK, JP, etc.).
Sponsored Products data goes back 95 days. Sponsored Brands and Sponsored Display data go back 60 days. Anything older than that isn't retrievable through the MCP server. For longer history you either export through the Ads API directly or use a tool that maintains its own time-series store.
No. The Amazon Ads MCP only reads advertising data. Questions like "which SKUs ran out of stock last week" or "what's my real margin per ASIN" require the Selling Partner API, which the official MCP doesn't expose. Tools that bundle SP-API and Ads API in a single MCP — like DataDoe MCP — fill that gap.
Each tool the server exposes — create_report, poll_status, fetch_data, parse_rows — triggers its own permission prompt in clients like Claude Desktop. Because most business questions require several tools in sequence, you'll click "approve" several times before the answer arrives. It's a property of how the tool surface is designed, not a Claude limitation.
Most Amazon Ads reports are asynchronous. The AI has to create a report, poll until it's ready, download it, then parse it — each step round-trips through the conversation and consumes tokens. Cross-source questions and retries on failed jobs can exhaust a Claude Pro plan on a single prompt.
Amazon's MCP covers only the Ads API and uses live async reporting. DataDoe MCP covers the full Selling Partner API plus the Ads API, runs on pre-aggregated reports, supports cross-source queries (ads × orders × inventory × search terms), and answers in seconds with charts and written insights. Both work with Claude, ChatGPT, Cursor, Codex, Copilot, and any other MCP-compatible client.
Yes. Any MCP-compatible AI client can connect to multiple servers at once. Some teams keep Amazon's MCP for direct campaign management actions and use DataDoe MCP for analytics, cross-source questions, and historical reporting beyond Amazon's 60–95 day windows.
Every integration. Full onboarding support. If it’s not the best decision you made in 2026, you can cancel anytime.
Know what makes you money
Catch problems instantly
Connect anything with API & MCP
Replace tools with your own apps
Access Amazon-audited infrastructure
Every integration. Full onboarding support. If it’s not the best decision you made in 2026, you can cancel anytime.
We use cookies to improve your experience and analyze traffic. By clicking "Accept", you agree to our use of cookies. Read the Privacy Policy.