Skip to main content

AI-assisted advertising workflows with controlled access.

Natural-language interaction can simplify complex advertising work, but authorization, scope and deterministic controls remain essential.

The Amazon Ads MCP ecosystem

MCP — the Model Context Protocol — is an open way for AI clients to use tools. Amazon and Terra PPC each provide their own MCP layer for their own products. They are separate, and they are operated by different companies.

Amazon Ads MCP Server

Amazon's technology

Amazon Ads has announced an official Amazon Ads MCP Server that connects AI agents with Amazon Ads API functionality. Amazon describes it as available to eligible Amazon Ads partners with active API credentials.

Availability and use are subject to Amazon Ads eligibility, credentials, authorization and applicable terms.

Operated by Amazon. Terra PPC does not own, operate or represent it.

Terra PPC AI/MCP

Terra PPC's product layer

Terra PPC's own interface for AI-assisted workflows inside Terra PPC, designed for MCP-compatible AI clients chosen by the customer.

Designed to use approved and authorized data and actions only, through Terra PPC authentication, authorization, scope, policy and mode controls.

Operated by ĐT ORGANIC COMPANY LIMITED.

Terra PPC AI/MCP is not the Amazon Ads MCP Server and is not presented as being powered by it.

Terra PPC AI/MCP architecture

An AI client never talks to data directly. It talks to Terra PPC, and Terra PPC applies the same checks it applies to a person using the product.

The order matters: identity and permission come first, scope limits what context exists, the decision engine decides what may be proposed, the execution planner and the guardrails decide whether and in what order it can proceed, and the mode decides whether anything changes.

Conceptual Terra PPC AI/MCP workflow. The request “Show campaigns above modeled break-even ACoS.” passes, in order, through: User request, Terra PPC AI / MCP, Authentication, Authorization, Tenant / account / marketplace scope, Approved decision context, Decision engine, Proposal, Execution planner, Guardrails, Manual / Review / Auto, Audit history. The read-only answer: 1 of 7 campaigns is above its modeled break-even ACoS. SP | Competitor Product Targeting has an ACoS of 52.1% against a modeled break-even of 37.0%. Sources: advertising data, customer-configured assumptions and derived values. Nothing was changed.

ConceptualConceptual architecture. It does not describe internal infrastructure.

What AI does not receive by design

The safest access is the access that was never granted. These exclusions are part of the design, not settings a request can change.

  • No direct unrestricted database access. AI clients call Terra PPC tools; they never query underlying tables.
  • No customer password. Neither Amazon passwords nor Terra PPC credentials are passed to an AI client.
  • No cross-tenant global context. A request only sees the selected organization, account, profile and marketplace.
  • No automatic bypass of authorization. Every request is authenticated and authorized like any other user action.
  • No unrestricted action execution. Changes follow the configured mode, policy and guardrails — or they do not happen.

Agent-assisted examples

Conceptual requests and what Terra PPC is designed to do with each. They show the pattern, not live output.

  • Request

    “Show campaigns above modeled break-even ACoS.”

    Uses
    Campaign performance and configured product economics in the selected scope
    Returns
    1 of 7 campaigns is above its modeled break-even ACoS. SP | Competitor Product Targeting: 52.1% vs 37.0%. Nothing changes.

    Read-only

    ADS: advertising data (Amazon Ads API context)ASSUMPTION: customer-configured input, not provided by AmazonDERIVED: calculated by Terra PPC from the values shown
    Ask it in the demo
  • Request

    “Why is scaling held for DEMO-LG-001?”

    Uses
    Cost versions, modeled break-even and campaign ACoS in the selected scope
    Returns
    How the product's cost reaches automation, step by step — and whether the Cost Increase Guard holds scaling because a recent cost increase pushed the campaign above break-even. Nothing changes.

    Read-only

    ADS: advertising data (Amazon Ads API context)ASSUMPTION: customer-configured input, not provided by AmazonDERIVED: calculated by Terra PPC from the values shown
  • Request

    “Show my modeled profit for this marketplace.”

    Uses
    Seller revenue, actual Amazon fees and refunds, ad spend and customer-configured costs — where authorized
    Returns
    Modeled Net Result $6,331.46 (29.5%), TACoS 9.6%, each line with its source.

    Read-only

    SELLER: seller-business data (SP-API context, where authorized)FINANCE: financial events (SP-API finance context, where authorized)ADS: advertising data (Amazon Ads API context)ASSUMPTION: customer-configured input, not provided by AmazonDERIVED: calculated by Terra PPC from the values shown
  • Request

    “Which SKUs have less than 7 days of cover?”

    Uses
    FBA inventory states and recent sales velocity
    Returns
    SKUs with their days of cover and inventory risk. If FBA inventory is not authorized, the answer says the context is unavailable.

    Read-only

    SELLER: seller-business data (SP-API context, where authorized)DERIVED: calculated by Terra PPC from the values shown
  • Request

    “Preview what Profit Guard would change.”

    Uses
    Rule R-02 evaluated as a simulation
    Returns
    Proposed bid changes with each guardrail check shown. Nothing is applied.

    Preview

    ADS: advertising data (Amazon Ads API context)DERIVED: calculated by Terra PPC from the values shown
  • Request

    “Increase all bids 50%.”

    Uses
    Bid targets in the selected scope, safety limits and business guardrails
    Returns
    Limited to the +20% maximum step; targets above break-even or low on inventory are held; each campaign's mode decides what happens next.

    Governed proposal → Review

    ADS: advertising data (Amazon Ads API context)SELLER: seller-business data (SP-API context, where authorized)ASSUMPTION: customer-configured input, not provided by Amazon
    Ask it in the demo

ConceptualConceptual examples with synthetic values from the demo account. No customer Amazon data is shown on this website. Every other request here is also a suggested prompt in the demo.

Action Center

One prioritized list of what needs attention in the selected account and marketplace — risk, efficiency, growth and maintenance items drawn from every module, each linked to the page that acts on it.

The AI answer to “What needs attention in this account?” reads the same list. Amounts are measured values of the period — never projected savings — and nothing in the list changes a campaign.

Work, ordered by severity

  • Risk

    Stock, listings, emergency controls and the circuit breaker — money or sales at risk now

  • Efficiency

    Cost-cutting recommendations, weak hours, placements and spend pacing above plan

  • Growth

    Budget-limited winners, winning word patterns, funnel promotions and scale recommendations

  • Maintenance

    Rules not evaluated, scheduled cost changes, plan gaps and returning snoozes

Each item keeps its sources ADS: advertising data (Amazon Ads API context)SELLER: seller-business data (SP-API context, where authorized)FINANCE: financial events (SP-API finance context, where authorized)ASSUMPTION: customer-configured input, not provided by AmazonDERIVED: calculated by Terra PPC from the values shown

IllustrativeSynthetic values from the Terra Home & Packaging demo account — not live Amazon data.Explore in the demo

Human control

AI-assisted requests follow the same Manual, Review and Auto modes as rules. The customer decides how far they may go.

How the modes work
  1. Manual

    AI-assisted requests can analyze, explain and preview. They cannot apply changes.

  2. Review

    AI-assisted requests can draft recommendations. A permitted user approves or rejects each one.

  3. Auto

    Only eligible actions that pass policy, guardrails and business guardrails may apply, and each one is recorded.

Amazon independence

Terra PPC is an independent software product operated by ĐT ORGANIC COMPANY LIMITED. It is not affiliated with or endorsed by Amazon. The Amazon Ads MCP Server, the Amazon Ads API and the Selling Partner API are Amazon technologies; references to them describe compatibility and ecosystem context only.

MCPModel Context Protocol — an open protocol for connecting AI clients to tools.

Request access to Terra PPC.

Tell us which accounts and marketplaces you manage and how much control you want to delegate. We'll follow up with access details and a quote.

Terra PPC is a software product operated by ĐT ORGANIC COMPANY LIMITED.

  1. Tell us about your accounts

    Advertising and seller accounts, marketplaces, and the workflows you want to control.

  2. We review scope

    Which modules, integrations and support your team needs.

  3. You receive a quote

    Access details and pricing, with no obligation to subscribe.