Amazon seller data layer

Amazon seller data layer for AI agents.

agentcentral is an Amazon seller data layer for AI agents. It syncs and normalizes Amazon Ads, Seller Central, inventory, orders, catalog, ranking, finance, and fulfillment data, then exposes that data through a hosted MCP server so Claude, ChatGPT, and custom agents can query factual seller records.

The platform is intentionally not a recommendation engine. It provides source fields, normalized joins, deterministic metrics, classifications, and audited write results for agents and workflows to reason over.

Last updated May 7, 2026.

Data layer
BigQuery-backed Amazon seller data exposed through MCP
Tool surface
91 domain-scoped MCP tools plus five utility tools
Tenant isolation
Per-tenant datasets with encrypted Amazon refresh tokens
Primary clients
Claude, ChatGPT, OpenClaw, Cursor, and custom agents
Metrics
Explicit formulas for ACOS, ROAS, CTR, CVR, TACOS, and days of cover
Boundary
Returns data and labels; the agent decides how to use them

Hosted MCP endpoint

Header-capable MCP clients use https://mcp.agentcentral.to/mcp with an agentcentral bearer key.

Scoped access

API keys can be read-only, domain-scoped, or include supported write tools for a specific workflow.

Data-layer boundary

agentcentral returns source fields, normalized joins, deterministic metrics, classifications, and audit results.

Data coverage

What the MCP server exposes.

The hosted server is built around factual seller data that an AI client can query repeatedly. The client does the reasoning; the agentcentral tools provide the structured data surface.

Normalized joins

Seller data is modeled so ads, orders, inventory, catalog, finance, and ranking fields can be read through one MCP surface.

Pre-materialized reads

Common analysis uses stored tables instead of forcing every follow-up question through a fresh Amazon report queue.

Deterministic labels

Fields such as days of cover, coverage state, risk level, and mapping confidence use explicit formulas or thresholds.

Audited writes

Supported write tools can be omitted, scoped, or audited with before and after values when enabled.

Access model

How AI clients connect.

Connect Amazon once, let agentcentral sync the seller data, then give the AI client a scoped MCP credential or connector URL.

OAuth connection

Amazon Ads and Seller Central are connected from the dashboard with OAuth.

Tenant dataset

agentcentral provisions an isolated BigQuery dataset for each tenant.

Scoped credentials

Each AI workflow gets its own key scope and can stay read-only when needed.

MCP client setup

The same hosted MCP server can be used from Claude, ChatGPT, or a custom HTTP MCP client.

FAQ

Common questions.

What is an Amazon seller data layer?+

It is a structured data surface that normalizes Amazon seller data and exposes it to tools or AI clients without requiring manual Seller Central exports.

Why use MCP for seller data?+

MCP gives AI clients a standard way to call authenticated tools. agentcentral uses MCP to expose Amazon seller facts, metrics, and source fields to supported clients.

Does agentcentral tell sellers what to do?+

No. agentcentral returns data, deterministic metrics, classifications, source-provided fields, and audited write results. Customer-side agents decide what to do with that data.

Can the data layer support writes?+

Yes, for supported Amazon Ads, catalog, price, inventory quantity, and fulfillment operations when a scoped key includes those write tools.

Use agentcentral as the Amazon seller data layer.

Start a 7-day free trial, connect Amazon, and create a scoped MCP key or connector URL for the AI client you use.

Amazon Seller Data Layer for AI Agents - agentcentral