Compare Amazon MCP servers

Kuudo vs Amazon Ads MCP Server

Amazon's official Ads MCP Server is a hosted endpoint on Amazon's own cloud: authorize your Login with Amazon application and your AI client reaches the campaign, report, billing and Amazon Marketing Cloud tools Amazon maintains, with no price published. Kuudo provisions the Amazon Ads, Selling Partner and Vendor Central servers into your own cloud, with verified Skills above the tools.

Kuudo facts come from the Kuudo MCP servers reference and the cloud deployment guides. Amazon facts come from the Amazon Ads MCP Server documentation and launch post, with the date each page was checked.

Thomas Spicer Updated Sep 12, 2026

Request beta access See what Kuudo runs How we compared

Whose infrastructure each piece sits on

Compared on Sep 12, 2026 · Amazon Ads MCP Server facts checked Sep 12, 2026

Side by side

Where the two actually differ.

20 rows. The first 10 follow the Amazon MCP connection step by step, from connecting the account through to a change going live: what Amazon Ads MCP Server does at each step and how Kuudo does it. Beyond an MCP endpoint, you get your Amazon Ads, Amazon Marketing Cloud, seller operations, Vendor Central, image and video generation, Amazon operating knowledge, and the AI clients you want to use. Each competitor cell links to the page it was checked against and the date.

Kuudo vs Amazon Ads MCP Server: 20 criteria compared side by side
Criterion	Amazon Ads MCP Server	Kuudo
The Amazon MCP connection, from connecting the account through to a change going live
Where it runs Whose cloud runs the server?	Amazon's own cloud: three regional endpoints, advertising-ai.amazon.com/mcp for North America, advertising-ai-eu.amazon.com/mcp for Europe and advertising-ai-fe.amazon.com/mcp for the Far East, which the documentation says are not region agnostic. No self-host path is published.²	Your own AWS, Google Cloud, Azure, or Cloudflare account. Kuudo provisions a private MCP server environment into the cloud account you pick, and you own the deployment, the data, and the perimeter.¹⁰
Your Amazon credentials Who holds the tokens to my Amazon account?	Your own Login with Amazon application. In token mode your client sends the application's client ID and a bearer access token on every request; in OAuth 2.1 mode the MCP client runs a browser login, a public client's access token expires after one hour, and a private client stores the refresh token in your system keychain or credentials file.³	Kuudo's authorization service receives and encrypts an Amazon refresh token, and in bring-your-own-app mode your application secret, then exchanges them for short-lived access tokens it issues to runtimes you authorize. The Amazon app can be your own, and the substantive API traffic does not route through Kuudo.¹¹
Where calls route Do my Amazon requests route through the vendor?	To Amazon directly, because the endpoint is Amazon's: each request carries your client ID and bearer token to Amazon's regional MCP endpoint, and the server calls the Amazon Ads API behind it. No third party sits between the AI client and Amazon.²	No. The DPA commits that your cloud resources connect directly to Amazon's business-data APIs, and that substantive API requests and responses do not proxy, relay, or route through Kuudo infrastructure.¹¹
Where your data sits Where does my Amazon data end up?	Not published. The MCP documentation covers account context modes and request headers; it does not say whether the server stores request or response data, or where.²	In your cloud. Agent Flow lands Amazon data in a lake in your own account, and Snowflake, Athena, DuckDB, Databricks, Power BI, and Tableau can read it downstream.¹⁴
Live or synced When my agent reads, is it calling Amazon or a copy?	From Amazon. The overview describes the server as a layer that translates Amazon Ads APIs into tools, and reports are generated on request, with skills that poll until a report completes.¹	From Amazon. The tools are Amazon's own API operations exposed over MCP, so a read is a call to Amazon. Agent Flow adds scheduled refreshes and lake tables in your own cloud when you want stored history.¹⁵
Tool surface What can my agent reach, and how is it counted?	Tool groups published under MCP Tools (beta): Account Management, Ads Account, Advertiser Product Group Eligibility, Amazon Live, AMC, Billing, Campaign Management, Eligibility, Manager Accounts, Reports, Terms Token, Test Accounts, User Invitation, User Permissions, User Roles and Users. The Campaign Management page lists 32 tools, and its query_campaign tool names Sponsored Products, Sponsored Brands, Sponsored Display, Sponsored Television and Amazon DSP; a total across groups is not published.⁴	1,055 Amazon API operations exposed as tools across three servers: 711 Amazon Ads, 304 Selling Partner, and 40 Vendor Central. At connect, the Amazon Ads MCP advertises 3 tools totaling 388 tokens, and the rest of the catalog is fetched only when a tool is needed.¹⁶
Writes Can it change my Amazon account?	Yes. Campaign Management tools create, update and delete campaigns, ad groups, ads, targets and portfolios, and delete_campaign removes a campaign with all associated ad groups, ads and targets.⁴	Yes. Every tool on every server is tagged read or write, and write tools exist across all three servers.¹⁵
What guards a write What stands between the agent and a change?	Client-side tool filtering: the get-started guide shows disabling tools such as campaign_management-delete_campaign and delete_ad in the client's configuration. A confirmation step, a policy gate, or a pre-read of current values is not described.²	Write tools are guarded: they validate input, apply your workspace policy, and may require explicit human approval before a change reaches Amazon.¹⁵
What is recorded What can I review afterward?	Not published. Neither the overview nor the tool reference describes a log of the tool calls an agent made.¹	Every tool call records the requesting user, agent, workspace, parameters, result summary, and approval state, across every server, in a run log you can use to reproduce an output or review a change.¹⁵
Above the tools What comes with it besides tool access?	Skills (beta): pre-defined prompt definitions provided by Amazon Ads that guide AI applications through Amazon Ads workflows and tool usage, in two types, workflow skills that sequence its tools and capability skills that explain them, triggered automatically in AI applications that recognize Anthropic skill definitions. Whether you can write your own is not addressed.⁷	12 verified Skills ship with the platform, each stating the job it does, the inputs it needs, the outputs it returns, and the limits it keeps, and you can write your own in the same format.¹⁷
Beyond an MCP endpoint
Amazon Ads Can it run my Amazon Ads?	Yes. The Reports group lists six tools for Amazon Ads reporting: create campaign, inventory and product reports, create and retrieve a report, and delete a report.⁵	711 Amazon Ads operations across Sponsored Products, Brands, Display, and TV, plus DSP, the unified Ads API v1, attribution, and reporting, through the Amazon Ads MCP.¹⁸
Amazon Marketing Cloud Can it query Amazon Marketing Cloud?	Yes. The AMC tool group lists workflows, data sources, ad-hoc SQL query execution, workflow execution, execution status and result download for an AMC instance. AMC audiences are not listed.⁶	Yes. Amazon Marketing Cloud queries and audiences run through the Amazon Ads MCP.¹⁵
Seller operations Can it run orders, inventory, pricing, and catalog work?	Not offered. The overview scopes the server to Amazon Ads insights such as campaigns, performance metrics, billing and account information, and no Seller Central tool group is listed.¹	304 Selling Partner operations for catalog, listings, orders, pricing, FBA, fulfillment, and finances, through the Amazon Selling Partner MCP.¹⁹
Vendor Central Can it run Vendor Central work?	Not offered. No Vendor Central tool group is listed.¹	40 Vendor Central operations: purchase orders and acknowledgements, direct fulfillment, shipments and invoices, transaction status, Retail Analytics, Data Kiosk, and A+ Content.¹⁵
Creative Can it generate listing and ad creative, images and video?	Not offered. No image or video generation tool group is listed.¹	Yes. Agent Iris generates listing images and product video through Google Gemini, OpenAI image models, and Google Veo 3, audits an image against listing rules, publishes to the listing, and confirms Amazon took it.⁹
Operating knowledge Does it bring Amazon know-how to the job?	Capability skills describe tool capabilities, explain input and output formats, and offer best practices for the server's own tools. Nothing is described as retrieving Amazon operating rules beyond tool usage.⁷	Agent Atlas supplies curated Amazon operating playbooks and rules that your agent retrieves before it acts.²⁰
Where you use it Which apps can I use it from?	Claude (Anthropic), ChatGPT (OpenAI), Amazon Kiro, Amazon Bedrock, Amazon AgentCore, and other MCP-compatible applications, per the overview; the OAuth 2.1 guide covers Kiro CLI, Claude Code, Claude Desktop, ChatGPT and Amazon Quick.¹	Claude, ChatGPT, Cursor, Copilot, n8n, or any MCP client.¹²
How it runs, and how you buy it
Pricing model How is it priced?	No price is published. The launch post says the server is available globally in open beta to Amazon Ads partners with active API credentials.⁸	Kuudo Community is free for one instance of each Amazon MCP server in your own cloud. Kuudo Pro is $0 a month plus $0.95 per 1,000 Amazon API calls, and Kuudo Scale is $249 a month plus $0.55 per 1,000 Amazon API calls. The calculator models a month for a brand owner, an agency or aggregator, and an automation-heavy seller.¹³
Seats and organizations How many people and accounts?	Not published as seats. Access follows the Login with Amazon application and the advertiser accounts it is authorized for, selected per request or fixed in headers; manager account scope works for some tools.²	Kuudo Pro: up to 25 seats across 5 organizations. Kuudo Scale: up to 50 seats across 10 organizations. Enterprise: unlimited.¹²
Free tier Can I try it free?	Not applicable. The server is in open beta for Amazon Ads partners with active API credentials, and no separate trial is published.⁸	Kuudo Community is free with no time limit: one instance of each Amazon MCP server in your own cloud, with no usage charges.¹²

Their fit

Choose Amazon Ads MCP Server if

Real situations where Amazon Ads MCP Server is the better buy. If one of them is yours, buy it.

You want the first-party endpoint your Amazon Ads API credentials already open, and Amazon publishes no separate price for it.
The launch post says the server is available globally in open beta to Amazon Ads partners with active API credentials, and no price is published anywhere in the documentation. If your team already holds an approved Login with Amazon application, the connection is a client ID and a token. Kuudo Community is free in your own cloud and Kuudo Pro is priced per Amazon API call; if the first-party endpoint with nothing on a card is the shape you want, this is it.
You want the server Amazon builds, maintains and updates itself.
Amazon's overview says Amazon Ads builds, maintains, and updates the MCP Server so that developers and AI applications can consistently access the latest functionality without re-engineering integrations, and it runs three regional endpoints for North America, Europe and the Far East. If the maintainer of the API being the maintainer of the endpoint is the guarantee you want from Amazon, it is theirs to give.
You want Amazon's own one-prompt workflows for launching a campaign or adding a country.
The Campaign Management tools include operations Amazon wrote for a whole job: create_singleshot_sp_campaign creates a complete Sponsored Products campaign, campaign, ad group and ads, across multiple marketplaces in one operation, and add_country_campaign extends a manual campaign to new countries with country-specific budget caps. Skills (beta) such as dsp-campaign-analysis chain report creation, polling and a combined analysis behind one request. If those are the shapes your day runs through, the first-party versions are here.

Our fit

Choose Kuudo if

Where Kuudo is the better buy: the servers, the credentials, and the data sit in infrastructure you own, with verified Skills above the tools. Beyond an MCP endpoint, you get your Amazon Ads, Amazon Marketing Cloud, seller operations, Vendor Central, image and video generation, Amazon operating knowledge, and the AI clients you want to use.

The whole Amazon job
You need Selling Partner, Vendor Central and Amazon Marketing Cloud from the same agent, not an advertising endpoint alone.
Amazon's server covers the Amazon Ads API. Kuudo's servers expose 711 Amazon Ads, 304 Selling Partner and 40 Vendor Central operations, Amazon Marketing Cloud queries and audiences, listing images and product video through Agent Iris, and the operating context from Agent Atlas. An advertising endpoint is one application on part of that surface. Kuudo is the surface.
Your AI client
Your AI client is already chosen, and the server should meet it there.
Kuudo connects to Claude, Claude Desktop, Claude Cowork, Claude Code, ChatGPT, Codex, Cursor, n8n and any other client that speaks Model Context Protocol, on your own Amazon app, from servers in your own cloud.
Your rules
You want playbooks that state their own limits, and the format to write your own.
12 verified Skills ship with the platform, each stating the job, the inputs, the outputs, and what it is not for, and you can write your own in the same format. Amazon's Skills (beta) are provided by Amazon Ads, and whether you can author one is not addressed.
Ownership
The deployment should be an asset you own, not an endpoint you rent.
Kuudo provisions private MCP server environments into your own AWS, Google Cloud, Azure, or Cloudflare account, and you own the deployment, the data, and the perimeter. The DPA commits that your cloud resources connect directly to Amazon and that substantive requests and responses do not route through Kuudo, and Agent Flow lands stored data in a lake in your own account. Private. Trusted. Owned.
Control
You need a record of what an agent did, and a gate you control before it does it.
Write tools apply your workspace policy and may require explicit human approval, and every call is recorded with its parameters and approval state in a run log. How much you automate, and who does the work, is your call.

The differences that matter

4 questions that decide it.

Each one opens with the answer, then what each product does, then what it means for the person doing the work.

How much of my context window does each server take before the first call?

Amazon's server advertises its whole catalog at connect: 92,678 tokens of tool definitions in a first-party measurement of the North America endpoint. Kuudo's Amazon Ads MCP advertises 388 tokens and fetches the rest of the catalog only when a tool is needed.

Amazon Ads MCP Server

110 tools advertised at connect, 92,678 tokens measured with the cl100k_base tokenizer on the North America endpoint: 46.34 percent of a 200k context window and 9.27 percent of a 1M window before the first prompt. Amazon publishes no tool count or token figure of its own.

Kuudo

3 tools advertised at connect totaling 388 tokens, 0.19 percent of a 200k window, with the rest of the catalog reachable on demand. A full discovery exercise across 100 tools, searches and schema fetches included, cost 25,605 tokens, 72.4 percent below Amazon's default.

What it means for you. Tool definitions are the one context cost that precedes every prompt, and on clients that re-send the catalog each turn it is paid on every turn. The measurement, its method and its caveats are in the guide.

Read the measurement

Whose cloud runs the server?

Amazon runs three regional endpoints on its own cloud, and no self-host path is published. Kuudo provisions private MCP server environments into the cloud account you choose.

Amazon Ads MCP Server

Amazon's cloud: advertising-ai.amazon.com/mcp for North America, with EU and FE endpoints, run by Amazon Ads, which builds, maintains and updates the server. No self-host path is published.

Kuudo

Kuudo provisions the servers into your own AWS, Google Cloud, Azure, or Cloudflare account, from the app, with a per-provider activation guide. You own the deployment, the data, and the perimeter.

What it means for you. Hosted by Amazon means the maintainer of the API runs the endpoint. Owned means the environment, the credentials and the data stay inside a perimeter you control, and the same deployment serves the Selling Partner and Vendor Central work beside the ads.

Cloud deployment destinations

What comes with it besides tool access?

Amazon ships Skills (beta): pre-defined prompt definitions that sequence its tools or explain them. Kuudo ships 12 verified Skills that state their limits, and the format to write your own.

Amazon Ads MCP Server

Workflow skills that orchestrate multi-step processes by combining MCP tools, such as dsp-campaign-analysis, and capability skills that describe tool capabilities and best practices, triggered automatically in AI applications that recognize Anthropic skill definitions, with availability varying by client. Whether you write your own is not addressed.

Kuudo

12 verified Skills ship with the platform, each publishing the job it does, its inputs and outputs, and what it is not for, and Agent Atlas supplies curated Amazon operating playbooks and rules that your agent retrieves before it acts. You can write your own Skills in the same format.

What it means for you. Prompt definitions that drive the vendor's own tools are the fastest path to the vendor's own workflows. Playbooks that publish their inputs, outputs and limits are what make a workflow safe to run unattended, and a format you can write in is what makes it yours.

Amazon Agent Skills reference

What stands between the agent and a change?

Amazon's guide shows removing tools such as delete_campaign from the client's configuration. Kuudo applies the workspace policy you wrote and may require a person's approval before a change reaches Amazon.

Amazon Ads MCP Server

Client-side tool filtering: the get-started guide disables campaign_management-delete_campaign and delete_ad through the client's configuration. A confirmation step, policy gate, pre-read of current values or change log is not described on the tool pages.

Kuudo

Write tools validate input, apply your workspace policy, and may require explicit human approval before a change reaches Amazon, and every call is recorded with its parameters and approval state in a run log you can replay.

What it means for you. Removing a tool from the client is a real control, and a blunt one: the tool is gone for every prompt or present for every prompt. A policy gate answers the other question, who signs off when the agent is allowed to write and is about to.

Read and write tools

Running both

Run both from one client.

MCP clients connect to many servers at once, so nothing stops you from having both. The split that makes sense is where each job belongs: Amazon's endpoint for its guided campaign launches and locale expansion, Kuudo for Selling Partner and Vendor Central work, Amazon Marketing Cloud queries and audiences beside the ads data, creative through Agent Iris, and anything that writes under your own approval policy from your own cloud.

Step 1
Add both servers to your MCP client; clients hold several connections at once.
Step 2
Let Amazon's endpoint run its guided campaign launches and country expansion when that is the job.
Step 3
Route Selling Partner and Vendor Central work, Amazon Marketing Cloud queries and creative through Kuudo's servers in your cloud.
Step 4
Put writes behind the workspace policy you set on Kuudo, and read every call back in the run log.

Methodology

How we compared

Every fact on this page links to the page that states it, with the date it was checked. What we could not verify is listed, not guessed.

Competitor facts come from pages the competitor publishes. Kuudo facts come from Amazon Agent Iris documentation, Cloud deployment destinations, Kuudo data processing addendum, Kuudo pricing, Kuudo pricing calculator, Amazon Agent Flow documentation, Kuudo MCP servers reference, Kuudo MCP servers reference: what connecting costs, Amazon Agent Skills reference, Amazon Ads MCP tools reference, Amazon Selling Partner MCP tools reference, Amazon Agent Atlas documentation. Planned work is cited to the roadmap and worded as planned. This page compares tools. For the job itself, see what your AI does once it is connected; for Kuudo's side of it, see Kuudo MCP and the Amazon Selling Partner MCP.

Sources

1 Amazon Ads MCP Server overview checked Sep 12, 2026
2 Connecting to the Amazon Ads MCP Server checked Sep 12, 2026
3 Connecting to the Ads MCP Server using OAuth 2.1 checked Sep 12, 2026
4 Amazon Ads MCP tools: Campaign Management checked Sep 12, 2026
5 Amazon Ads MCP Server: Reports tools checked Sep 12, 2026
6 Amazon Ads MCP tools: AMC checked Sep 12, 2026
7 Amazon Ads MCP Skills overview checked Sep 12, 2026
8 Introducing the Amazon Ads MCP Server checked Sep 12, 2026
9 Amazon Agent Iris documentation checked Sep 12, 2026
10 Cloud deployment destinations checked Sep 12, 2026
11 Kuudo data processing addendum checked Sep 12, 2026
12 Kuudo pricing checked Sep 12, 2026
13 Kuudo pricing calculator checked Sep 12, 2026
14 Amazon Agent Flow documentation checked Sep 12, 2026
15 Kuudo MCP servers reference checked Sep 12, 2026
16 Kuudo MCP servers reference: what connecting costs checked Sep 12, 2026
17 Amazon Agent Skills reference checked Sep 12, 2026
18 Amazon Ads MCP tools reference checked Sep 12, 2026
19 Amazon Selling Partner MCP tools reference checked Sep 12, 2026
20 Amazon Agent Atlas documentation checked Sep 12, 2026

What we could not verify

A total tool count for Amazon's server: the documentation lists tool groups and the tools in each, and publishes no total. The count and token figure on this page are a first-party measurement of the North America endpoint, dated in the guide.
Whether Amazon's server stores request or response data, and where; the MCP documentation does not say.
Whether Amazon's server offers AMC audience tools; the AMC page lists workflow, data source, query and execution tools only.
Whether Amazon's server creates Sponsored Display and Sponsored Television campaigns; create_campaign names Sponsored Products, Sponsored Brands and Amazon DSP, and query_campaign names all five.
Rate limits and quotas on the MCP endpoints; not published.
Whether a customer can author an Amazon Skill; the Skills overview says they are provided by Amazon Ads.
Whether advertisers without an Amazon Ads partner application can use the open beta; the launch post names partners with active API credentials, and the OAuth guide requires an approved application with access to the Amazon Ads API.

If you work at Amazon Ads and something here is wrong or out of date, tell us and we will fix it. Contact us.

Frequently asked questions

Questions advertisers ask before choosing between Amazon's Ads MCP Server and Kuudo.

Hosting, credentials, what each one covers, what connecting costs, what ships above the tools, and whether you need both.

Is the Amazon Ads MCP Server better than Kuudo?

For a first-party endpoint that Amazon builds, maintains and updates, with no price published and guided one-prompt workflows for launching a campaign or adding a country, Amazon's server is the better fit. For Selling Partner and Vendor Central work from the same agent, Amazon Marketing Cloud queries and audiences, servers in your own cloud on your own Amazon app, and Skills that state their limits, Kuudo is. The rows on this page are how to decide.

Can I use Amazon's Ads MCP Server and Kuudo together?

Yes. MCP clients hold several server connections at once. Let Amazon's endpoint run its guided campaign launches and locale expansion, and route Selling Partner and Vendor Central operations, Amazon Marketing Cloud queries, creative through Agent Iris, and writes under your own approval policy through Kuudo's servers in your cloud.

How much of my context window does each server use before the first prompt?

In a first-party measurement of the North America endpoint on 2026-09-12, Amazon's server advertised 110 tools totaling 92,678 tokens of tool definitions at connect, about 46 percent of a 200k window. Kuudo's Amazon Ads MCP advertised 388 tokens and fetches the rest of its catalog only when a tool is needed. The method and caveats are in the linked guide.

Does Amazon's Ads MCP Server cover Seller Central or Vendor Central?

No. Amazon's overview scopes the server to Amazon Ads insights such as campaigns, performance metrics, billing and account information, and no Seller Central or Vendor Central tool group is listed. Kuudo runs seller operations through the Amazon Selling Partner MCP and Vendor Central through its own MCP server, with purchase orders, direct fulfillment, shipments, invoices and Retail Analytics.

Does Amazon's Ads MCP Server cover Amazon Marketing Cloud?

Yes. Its AMC tool group lists workflows, data sources, ad-hoc SQL query execution, workflow execution and result download for an AMC instance; AMC audiences are not listed. Kuudo runs Amazon Marketing Cloud queries and audiences through the Amazon Ads MCP.

Which one writes to my Amazon account, and what guards it?

Both write. Amazon's Campaign Management tools create, update and delete campaigns, ad groups, ads, targets and portfolios, and the get-started guide shows removing tools such as delete_campaign from the client's configuration. Kuudo's write tools validate the input, apply the workspace policy you set, may require a person's approval, and record every call in a run log.

What does each one cost?

Amazon publishes no price; the launch post says the server is available globally in open beta to Amazon Ads partners with active API credentials. Kuudo Community is free for one instance of each Amazon MCP server in your own cloud; Kuudo Pro and Scale are priced per 1,000 Amazon API calls.

More comparisons

All comparisons

Other comparisons.

Same rows, same doctrine, a different competitor. Every page says where the other tool is the better fit before it says where Kuudo is.

Kuudo vs agentcentral for Amazon MCP agentcentral hosts one Amazon MCP endpoint and syncs your data to its cloud. Kuudo provisions the servers into your own. Sourced facts, dated checks. Read the comparison Kuudo vs Marketplace Ad Pros for Amazon MCP Marketplace Ad Pros hosts an Amazon MCP endpoint and syncs your data to its cloud. Kuudo provisions the servers into your own cloud. Sourced, dated. Read the comparison

Guides

The reasoning behind the rows.

How an agent decides whether to read live or from stored data, waits for a person before an audience activates, and runs an Amazon Marketing Cloud workflow end to end.

View all guides

Agents · Decision rules

Amazon Ads MCP Server: What Connecting Costs

Amazon's official Ads MCP Server hands an agent 92,678 tokens of tool definitions at connect. Kuudo's Amazon Ads MCP hands it 388. Measured and dated.

Human Approval for AMC Activation

Gate AMC audience activation behind human approval: size the seed, review the exact payload, then submit. 2000-user floor, 24 to 36 hour DSP lag.

Turn AMC Workflows into Agent Skills

Package repeatable AMC analysis into agent Skills with Atlas retrieval, sufficiency checks, human approval, and run logging already built in.

Thomas Spicer

4 min read

May 9, 2026

The bottom line

Own the deployment your agent works through.

Kuudo provisions the Amazon Ads, Selling Partner, and Vendor Central MCP servers into your own cloud account, on your own Amazon app, with verified Skills above the tools. Community is free.

Request beta access See pricing