Amazon Ads MCP server

Amazon Ads MCP Server — by SellerMate.AI

Connect Claude, ChatGPT, Cursor, or any MCP client to your live Amazon Advertising data — and, with approval guardrails, let it act: negate wasted search terms, adjust budgets and bids, launch campaigns.

Remote server. One URL, both transports (streamable HTTP and SSE). No API keys. Free on every plan.

https://api.sellermate.ai/mcp/sse

(budget/bid floors and ceilings, change-multiplier caps, per-currency limits) and can require admin approval. Full audit trail.

• Listed in Claude's connector directory — one-click add, sign in, ask.
• OAuth 2.1 + PKCE — no keys to paste or rotate; account-level access control.
• Read + write, policy-checked — every write action passes your workspace guardrails

account connected to your Amazon account is all you need.

• Read-only by default. Zero data caching.
• No Amazon Ads API credentials required — a free SellerMate

Used by 2,000+ brands and agencies in 20+ countries · $1B+ ad spend optimized · 2026 Amazon Ads Partner Awards — Technology Innovation Global Finalist.

Quickstart

Claude (claude.ai, free plan works)

(or start from sellermate.ai/product/mcp → Add to Claude).

1. Open Claude's connector directory and add SellerMate

sees your Claude conversation, Claude never sees your credentials).

1. Sign in with your SellerMate account when prompted (secure OAuth — SellerMate never

1. Ask: "Which search terms wasted budget with zero orders in the last 30 days?"

Claude Desktop, Claude Code, Cursor, VS Code, Windsurf, Gemini CLI

Add a remote MCP server with the URL:

https://api.sellermate.ai/mcp/sse

The client opens a browser window for the OAuth sign-in on first use. In Claude Code:

claude mcp add --transport sse sellermate https://api.sellermate.ai/mcp/sse

ChatGPT

Add SellerMate as a custom connector using the same URL. (Connector support varies by ChatGPT plan; see the setup guide.)

Full walkthroughs per client: docs/getting-started.md

What you can ask it

The server exposes 50+ tools across reporting, campaign management, targeting, negatives, automations, dayparting, DSP, vendor/retail analytics, and account organization. The complete reference: docs/capabilities.md.

Safe to hand to an AI

The AI proposes; your policies decide. Every write action is checked server-side against workspace guardrails your admin controls — minimum/maximum budgets and bids, change-multiplier caps relative to current values, per-currency floors. Sensitive changes can require explicit admin approval before anything touches Amazon, and every action lands in an audit log. The model cannot exceed the limits you set, because the platform enforces them on every call — not the prompt.

Details: docs/security.md

How is this different from Amazon's official Ads MCP server?

Shortened. The full README is on GitHub.