Amazon Seller Central MCP

Amazon Seller Central MCP for Claude.

agentcentral is a hosted MCP server for Amazon sellers that connects Claude, ChatGPT, and other AI clients to Amazon Ads, Seller Central, inventory, orders, catalog, rankings, finance, and fulfillment data.

It is for sellers, operators, agencies, and agent builders who want an AI client to read clean Amazon seller data without waiting on Seller Central exports or local MCP servers.

Start Full Suite trialClaude setup

Claude + agentcentral

> Show SKUs below 21 days of cover with FBA stock, inbound units, and sales velocity.

SKU

FBA

cover

TUMBLER-20-BLK

38

14 days

MUG-12-WHT

112

19 days

Tool data: get_days_of_cover, get_fba_inventory, get_inbound_shipments, get_sales_velocity

Pre-materialized data

Repeated seller-data analysis reads from normalized tables instead of waiting for every Amazon report to finish.

Fast repeated reads

Agents can ask follow-up questions across campaigns, orders, inventory, and catalog without rebuilding CSV exports.

Scoped access

Keys can be limited by domain and write tools can be left out for read-only analysis clients.

Direct answer

What can Claude do once connected?

Claude can call agentcentral tools to retrieve factual Amazon seller data, normalized joins, deterministic metrics, and source-provided Amazon fields. The AI client decides how to reason over that data; agentcentral supplies the data layer.

Show my Amazon Ads campaigns with spend, sales, ACOS, ROAS, and TACOS for the last 30 days.

List SKUs with fewer than 21 days of cover and include FBA, AWD, inbound, and sales velocity fields.

Which search terms spent more than $50 without attributed sales last week?

Show yesterday's orders by SKU and compare them with current FBA inventory.

Find catalog listings with missing attributes, listing issues, or suppressed listing status.

Show keyword rank changes for my tracked ASINs and join them to search volume.

Supported data

Amazon Ads, Seller Central, inventory, orders, catalog, rankings, finance, and fulfillment.

agentcentral exposes one hosted MCP server for Amazon sellers across the main seller data domains an operator asks about day to day.

Amazon Ads

Campaigns, ad groups, keywords, search terms, budgets, placements, TACOS, DSP, Stores, and Amazon Ads source-provided bid and budget suggestion fields.

Inventory

FBA inventory, AWD inventory, inbound shipments, shipment receiving status, days of cover, sales velocity, active listings, suppressed listings, and inventory risk labels.

Orders

Today's sales, order lists, order line details, returns, reimbursements, sales summaries, and seasonality indexes.

Catalog

Product details, sales ranks, A+ Content status, variation relationships, product reviews, review trends, listing quality, listing issues, and catalog cleanup signals.

Rankings

Daily keyword rank positions, day-over-day rank changes, and rank positions joined with search volume and market share.

Finance

Fee breakdowns, ASIN profitability, profitability review metrics, settlement economics, and finance fields joined to ads and order data.

Fulfillment

Multi-Channel Fulfillment shipping previews, MCF orders, tracking, order creation with audit trail, and return authorization management.

Why hosted matters

Fast reads are the product surface.

Seller Central and Amazon Ads APIs often involve report queues, delayed attribution, pagination, and account-specific joins. agentcentral syncs, normalizes, and stores seller data before the agent asks for it, so repeated analysis is reliable and fast.

Fewer live-report delays

Common analysis runs against pre-materialized tables instead of waiting on a fresh Amazon report for each question.

Reliable follow-ups

An agent can ask narrower follow-up questions over the same normalized data without rebuilding context.

Cross-domain joins

Ads, orders, inventory, catalog, ranking, finance, and fulfillment fields can be read through one MCP surface.

Deterministic metrics

Fields such as ACOS, ROAS, CTR, CVR, TACOS, days of cover, and coverage states use explicit formulas or thresholds.

Setup

How setup works at a high level.

The setup path is account, Amazon OAuth, scoped credential, then AI client configuration.

1

Create an account

Start in the agentcentral dashboard and choose the domains your agent should be able to access.

2

Connect Amazon

Authorize Amazon Ads and Seller Central through OAuth so agentcentral can sync and normalize your seller data.

3

Create a key or connector URL

Use a bearer-header API key for MCP clients that support headers, or copy a signed Claude connector URL from the dashboard.

4

Add it to Claude or ChatGPT

Paste the connector URL or MCP config into your AI client, then ask the client to list the available agentcentral tools.

Claude quickstart ChatGPT quickstart MCP config examples

Prompt examples

Example prompts for Amazon sellers.

These prompts ask for data, comparisons, and source fields that agentcentral can return through MCP.

Can you show my Amazon Ads performance by campaign for the last 14 days and include ACOS, ROAS, and total sales?

Which SKUs are below 30 days of cover? Include current FBA stock, inbound units, sales velocity, and suppressed listing status.

Compare yesterday's orders with current inventory and show any ASINs where demand is moving faster than recent average.

Show search terms with spend over $25, clicks over 20, and no attributed sales in the last 7 days.

List catalog issues by ASIN with severity, issue source, and the last snapshot date.

Show keyword rank changes for my tracked terms and include search volume where available.

Security

Security and control.

agentcentral is built around Amazon OAuth, tenant isolation, scoped keys, and observable write behavior.

Amazon OAuth for Ads and Seller Central connections

Encrypted Amazon refresh tokens

Per-tenant BigQuery datasets for seller data isolation

Scoped API keys so a client can receive only the domains it needs

Read-only key options when an agent should not use write tools

Write guardrails, pre-read checks, and audit logs for supported mutation tools

Expectations

Limitations to know before setup.

A hosted MCP data layer removes local server work, but it still depends on Amazon authorization, sync timing, and client support for MCP connections.

Amazon authorization is required. agentcentral cannot read a seller account until Amazon Ads or Seller Central OAuth is connected.

Most reporting data is synced into pre-materialized tables. Ads sync daily with a rolling lookback because Amazon attribution can settle after the first report date.

Some tools call Amazon APIs live for source-provided fields such as bid recommendations, competitive pricing, eligibility, and listing restrictions.

Tool access follows the account plan and each API key's scope. Write tools can be omitted from a key for read-only clients.

Claude custom connectors use a signed connector URL from the dashboard. Header-capable MCP clients can use the public MCP endpoint with an Authorization header.

FAQ

Buyer questions about Amazon MCP for Claude.

The FAQ is written in the same language sellers use when asking answer engines how to connect Amazon data to an AI client.

Can Claude connect to my Amazon Seller Central data?+

Yes. agentcentral connects Claude to Amazon Seller Central data through a hosted MCP server. After Amazon OAuth and the first data sync, Claude can query inventory, orders, catalog, rankings, finance, fulfillment, and connected Amazon Ads data.

What MCP server works with Amazon Ads?+

agentcentral includes Amazon Ads MCP tools for campaign performance, keywords, search terms, placements, budget pacing, TACOS, DSP, Store performance, and guarded write tools for supported bid, budget, keyword, and campaign updates.

Is agentcentral only for Claude?+

No. Claude custom connectors and Claude Code are supported, and agentcentral can also be used with ChatGPT, OpenClaw, Cursor, and other AI clients that support remote MCP servers.

Do I need to run a local server?+

No. agentcentral is hosted. Header-capable MCP clients use the public MCP endpoint with a bearer API key, and Claude custom connectors use a signed connector URL generated in the dashboard.

How fresh is the Amazon seller data?+

Ads reporting data syncs daily with a rolling lookback for delayed attribution. Orders sync more frequently when Seller Central is connected. Live API tools are used for selected current fields such as eligibility, pricing, rank checks, and Amazon source-provided suggestions.

Can an AI agent change Amazon Ads or listings?+

Yes, only when the API key includes write tools. Supported write tools use guardrails such as pre-read checks, hard limits where applicable, confirmation flows for sensitive catalog changes, and audit logs with before and after values.

Is seller data isolated?+

Yes. Seller data is stored in tenant-specific datasets, credentials are encrypted, and API keys can be scoped by domain so each AI client receives only the tools it should use.

What Amazon data domains are supported?+

agentcentral supports Amazon Ads, Seller Central inventory, orders, catalog, rankings, finance, and fulfillment data through one hosted MCP server.

Connect Amazon seller data to your AI client.

Start the Full Suite trial, connect Amazon, create a scoped key or Claude connector URL, and make the first MCP request.

Start Full Suite trialAmazon Ads MCP

Amazon Seller Central Claude MCP Server - agentcentral