How to Automate Amazon Ads with Claude AI and MCP — Workflows, Tools, and Real Results

Amazon PPC management is a full-time job that never actually ends.

Between harvesting search terms from Auto campaigns, adjusting bids across hundreds of ASINs, pulling performance reports, expanding to new marketplaces, and running Amazon Marketing Cloud queries — a serious seller or agency can easily spend 15–20 hours a week on operational work that a well-configured AI agent should be handling.

That window just closed. On February 2, 2026, Amazon Ads officially launched its MCP server into open beta, giving any Amazon Ads API partner the ability to connect Claude — or any other MCP-compatible AI — directly to their live ad account data. The integration goes both ways: Claude can read performance data and act on it, creating campaigns, adjusting bids, running AMC analytics, and expanding to new marketplaces, all from a natural language conversation.

This guide covers every setup option, the complete tool surface, and five concrete workflows you can run starting today.

---

What Is the Amazon Ads MCP Server?

The Amazon Ads MCP server is Amazon's official implementation of the Model Context Protocol — the open standard that lets AI models connect to external tools and data sources with live, authenticated access.

It works as a translation layer between your AI agent and the Amazon Ads API. When you send Claude a natural language prompt, the MCP server converts it into structured API calls, executes them against your live ad account, and returns the results — all within your conversation. No CSV exports, no Seller Central tab-switching, no manual report downloads.

The key architectural distinction from most ad platform integrations: Amazon built its tools as orchestrated workflows, not just raw API wrappers. A tool like "create Sponsored Products campaign" handles the full multi-step operation — campaign, ad group, and ad creation — in a single action. The tools reduce complexity by orchestrating complete, multi-step operations that are core to everyday advertising workflows.

---

Three Server Options Compared

You have three routes to connect Claude to Amazon Ads, each with different trade-offs on setup complexity, tool depth, and authentication overhead.

1. Official Amazon Ads MCP Server

Amazon's own implementation. Covers the full ad stack: Sponsored Products, Sponsored Brands, Sponsored Display, DSP, Amazon Marketing Cloud, Amazon Attribution, and account management. Requires existing Amazon Ads API partner credentials. Available globally; region is auto-routed based on your advertiser profile (NA, EU, FE).

Best for: Agencies and brands already integrated with the Amazon Ads API who want the deepest, most complete toolset.

2. KuudoAI — KuudoAI/amazon_ads_mcp

The strongest community implementation. Two authentication options: bring your own Amazon developer app credentials (full control), or use Openbridge's pre-configured OAuth gateway (faster setup, minimal configuration). Covers the same major categories as the official server. Actively maintained with clean documentation.

Best for: Developers wanting flexibility on auth — especially teams already using Openbridge for Amazon data pipelines.

3. MarketplaceAdPros — Easiest Setup

The zero-friction path. Get a token from MarketplaceAdPros, paste it into your config — done. Covers Sponsored Products, Sponsored Brands, and Sponsored Display campaigns, ad groups, keywords, product ads, and targeting. A premium subscription unlocks recommendations, experiments, and advanced features. Hosted MCP endpoint means no local server to maintain.

Best for: Sellers and smaller agencies who want to start in under 5 minutes without setting up Amazon API credentials.

---

Full Tool Surface by Category

The Amazon Ads MCP server organizes its tools by prefix — each prefix maps to a specific area of the ad stack. Here's what Claude can access:

---

Setup Guide

Path 1 — Easiest: MarketplaceAdPros (Under 5 Minutes)

1. Go to marketplaceadpros.com and create an account
2. Connect your Amazon Ads account and receive your bearer token
3. Add to your Claude Desktop config:

{
  "mcpServers": {
    "amazon-ads": {
      "command": "npx",
      "args": ["-y", "@marketplaceadpros/amazon-ads-mcp-server"],
      "env": {
        "BEARER_TOKEN": "your_token_here"
      }
    }
  }
}

Or use the hosted endpoint directly: https://app.marketplaceadpros.com/mcp

Path 2 — KuudoAI with Openbridge (15–20 Minutes)

If you use Openbridge for Amazon data, this is the fastest path to the full tool surface including DSP and AMC.

git clone https://github.com/KuudoAI/amazon_ads_mcp
cd amazon_ads_mcp
pip install -r requirements.txt

# Add to Claude Desktop config
{
  "mcpServers": {
    "amazon-ads-mcp": {
      "command": "python",
      "args": ["/path/to/amazon_ads_mcp/server.py"],
      "env": {
        "AMAZON_AD_API_ACCESS_TOKEN": "your_openbridge_token",
        "AMAZON_AD_API_PACKAGES": "profiles,ads-api-v1-sp,amc-workflow,reporting-version-3"
      }
    }
  }
}

Path 3 — Official Amazon Server (30–60 Minutes)

Prerequisites: Active Amazon Ads API partner credentials (Login with Amazon OAuth, client ID, client secret, refresh token)

Apply for Amazon Ads API access at advertising.amazon.com/API if you don't already have credentials. Once approved:

{
  "mcpServers": {
    "amazon-ads": {
      "command": "python",
      "args": ["-m", "amazon_ads_mcp"],
      "env": {
        "AMAZON_AD_API_CLIENT_ID": "your_client_id",
        "AMAZON_AD_API_CLIENT_SECRET": "your_client_secret",
        "AMAZON_AD_API_REFRESH_TOKEN": "your_refresh_token",
        "AMAZON_AD_API_PACKAGES": "profiles,ads-api-v1-sp,ads-api-v1-sb,dsp,amc-workflow"
      }
    }
  }
}

---

Workflow 1: Full Account ACOS Audit (What Used to Take a Morning)

ACOS (Advertising Cost of Sale) analysis across a multi-ASIN account is the kind of work that eats entire mornings. You need to cross-reference campaign spend against attributed sales, segment by campaign type, identify outliers, and build a prioritized action list — all before you can make a single change.

Claude does this in under a minute.

A senior PPC manager doing this manually would need 2–3 hours. The same analysis with Claude takes 45 seconds.

---

Workflow 2: Auto-to-Manual Keyword Harvesting

Auto campaigns are Amazon's discovery engine — they find search terms you would never think to target. The problem is extracting those winners and promoting them to manual Exact Match campaigns before budget bleeds on terms you've already proven convert.

Doing this manually means downloading the search term report, filtering by conversion and spend thresholds, cross-checking against your existing exact match keyword list, and then adding the new keywords one by one. For a 50-ASIN account, that's 3–4 hours every two weeks.

One confirmation. 23 keywords harvested, added to manual campaigns, and negated from Auto campaigns — a process that previously took half a day done in 90 seconds.

---

Workflow 3: Marketplace Expansion

Expanding a successful US product to UK, DE, or CA Amazon marketplaces normally requires: translating copy, setting up new campaigns from scratch in Seller Central, configuring currency-appropriate bids, and waiting weeks to gather enough data. Most sellers delay this for months because the setup overhead is too high.

The Amazon Ads MCP server includes built-in orchestrated workflows specifically for marketplace expansion — one of its flagship capabilities at launch.

What typically takes 3–4 days of Seller Central setup across two marketplaces: completed review-ready in 2 minutes.

---

Workflow 4: Amazon Marketing Cloud Analytics — SQL-Free

Amazon Marketing Cloud (AMC) is one of the most powerful data tools in Amazon's ecosystem — a privacy-safe data clean room where you can run SQL queries across your ad exposure, purchase, and streaming data to get attribution insights unavailable anywhere else.

The problem: AMC requires SQL. Most sellers and even many agencies don't have the technical resources to write AMC queries consistently, so the tool sits underused.

With the amc_ tools, Claude writes and executes AMC workflows for you in plain English.

This is attribution intelligence that would previously require an AMC-certified analyst and a week of query development. With Claude, it's a 90-second conversation.

---

Workflow 5: Bid and Budget Optimization

Bids and budgets on Amazon drift. A keyword that deserved a $1.50 bid three months ago might now warrant $0.90 — or $2.20 — depending on what's happened to your conversion rate, competitor activity, and seasonal demand. Most accounts run on stale bids because the manual process of reviewing and adjusting hundreds of keywords across dozens of campaigns is just too slow.

---

Verified Results: What Sellers Are Reporting

These numbers come from published sources, not projections.

---

What Claude Cannot Do (Be Realistic)

---

Combining Amazon Ads MCP with Other Servers

---

Getting Started Today

If you're a seller (non-technical):

1. Go to marketplaceadpros.com — connect your Amazon account and get a token (5 min)
2. Install Claude Desktop
3. Add the config JSON above and restart Claude Desktop
4. Start with: "List all my Amazon advertiser profiles" to confirm the connection works
5. Run the ACOS audit workflow — you'll have your first action list in under 2 minutes

If you're an agency or developer:

1. Use KuudoAI with Openbridge if you want the full DSP + AMC surface with minimal credential overhead
2. Or apply for Amazon Ads API credentials at advertising.amazon.com/API for the official server
3. Start with the profiles and ads-api-v1-sp packages — add amc-workflow and dsp once the basics are working
4. Test keyword harvesting on a single ASIN first before running account-wide bid changes

Want to explore MCP before committing?
Go to MCP Playground Online — connect any MCP server URL in your browser to inspect its tools and test queries without any local setup.

---

The Shift That's Happening

Amazon's MCP server launch is not a minor product update. It's a signal that the company is betting on AI agents as the primary interface for advertising management — not dashboards, not CSV exports, not manual workflows in Seller Central.

The Amazon Ads API has always contained the full data and action surface. What's new is that MCP makes that surface directly accessible to frontier AI models without any custom integration work. A seller who connects Claude to Amazon Ads today gets the same programmatic access that enterprise brands have spent years and engineering resources building — in 5 minutes.

The early adopter window on this is genuinely narrow. When every competitor is running the same automated bid optimization and keyword harvesting workflows, the edge disappears. The brands building systematic AI-assisted advertising operations now will have structural advantages that compound over time.

---

The Amazon Ads MCP server is an open beta product. Features, tool availability, and pricing are subject to change. Always verify campaign changes before applying — write operations execute against live accounts immediately. Test with a low-spend profile before connecting to your primary ad account.

---

Resources:

• Amazon Ads MCP Server — Official Announcement
• KuudoAI/amazon_ads_mcp — Community server with Openbridge auth
• MarketplaceAdPros/amazon-ads-mcp-server — Easiest setup, hosted endpoint
• Amazon Ads MCP API Documentation — Official developer guide
• MCP Playground Online — Test MCP servers in your browser