Amazon Ads MCP Server Open Beta

What's Happening

Amazon just opened the doors to its Ads MCP (Model Context Protocol) Server. In practical terms, this means any AI assistant that supports MCP can now plug directly into your Amazon advertising account. Claude, ChatGPT, Gemini, custom-built agents: they all get the same access. One prompt can create a full Sponsored Products campaign, adjust bids, or pull performance data.

This isn't Amazon's own AI managing your ads (that's the Ads Agent announced at unBoxed 2025). The MCP Server is the open protocol that lets any third-party AI connect. Think of it as the API layer between your ad account and whatever AI tool you prefer. It's a fundamental shift in how PPC management works.

For the bigger picture on where MCP is heading for seller analytics, beyond just the ad layer, see our POV on why the data model is the real moat and the dedicated AI agent for Amazon sellers Page.

What AI Can Do Through MCP (and How Fast)

The MCP Server exposes campaign management and reporting capabilities, but they don't perform equally. Campaign creation is near-instant. Reporting is async and slow. Here's the breakdown:

The speed gap is significant. You can spin up 20 campaigns in under a minute, but pulling last month's ACoS data requires an async request that could take several minutes to resolve. For sellers managing high-volume PPC operations, the reporting bottleneck matters.

Data Limits You Should Know

If you're used to pulling 12 months of keyword-level data for seasonal analysis, the MCP Server won't help. It's designed for operational execution, not deep historical analytics. For that kind of analysis, you still need a dedicated Amazon analytics platform that stores your data long-term.

What the MCP Server Can't See

This is where it gets interesting. The MCP Server operates in an advertising-only silo. It has zero visibility into the rest of your Amazon business:

• ✗Inventory levels: Your AI can bid aggressively on a product that's about to go out of stock
• ✗Profit margins: No COGS, FBA fees, or profit and loss data to inform bid decisions
• ✗FBA fees: Can't factor in fulfillment costs, storage fees, or fee changes when optimizing bids
• ✗Organic rankings: No insight into organic vs. PPC sales split
• ✗Return rates: Your AI doesn't know if a product has a 30% return rate eating into profits

Who Is Affected

PPC agencies Face the biggest disruption. When any AI can create and manage campaigns, the value shifts from execution to strategy and profit optimization. Agencies that already offer product-level ROI tracking are better positioned than those selling campaign management hours.

What You Should Do Now

1. 1.

   Don't hand AI your ad account without guardrails

   Set budget caps, bid ceilings, and campaign naming conventions before connecting any AI tool via MCP. An AI that can create campaigns instantly can also waste budget instantly.

2. 2.

   Audit your TACoS baseline before and after

   Track your Total ACoS before enabling any MCP-connected AI. If TACoS rises after AI takes over campaign creation, the AI is optimizing for the wrong metrics.

3. 3.

   Keep profit data separate from AI execution

   Since MCP can't see your margins, maintain a P&L analytics layer that feeds context back into your bidding decisions. Let AI execute, but let your analytics decide what to execute.

4. 4.

   Watch for SP-API rate limit Issues

   MCP requests count against your standard API throttling limits. If you're already running other integrations, adding AI-driven MCP calls could trigger rate limiting on your reporting.