Top Amazon MCP servers for Ads & Seller data (2026)
What is the best Amazon MCP server in 2026?
An Amazon MCP server connects your Amazon data to Claude, ChatGPT, or Gemini through the open Model Context Protocol, so you can ask questions in plain language instead of exporting reports. Amazon's own official Amazon Ads MCP Server is the anchor for first-party Ads API access, and since September 2026 a separate Amazon Selling Partner MCP covers Seller Central in the US. Marketplace Ad Pros spans Ads, Selling Partner, DSP, and AMC. Helium 10, Ad Badger, and Atom11 ship connectors over their own platforms. DataDoe, Nova, Agent Central, Seller Labs, and PPC Prophet are hosted connectors, and openbridge plus community SP-API servers are open source.
Marketplace Ad Pros publishes this guide and ships its own Amazon MCP server, so it is listed with that disclosed. Servers are compared on features, hosting, security, and public documentation.
Last updated: November 4, 2026
What changed in November 2026
- Added Amazon's Selling Partner MCP (launched September 23, 2026): Seller Central data with approval-gated changes, in beta in Claude and Amazon Quick, US stores only.
- Updated the Amazon Ads MCP Server: still open beta, now with a three-tool Lite endpoint and closed-beta "expert tools" (starting with an AMC SQL generator), and expert-tool general availability with pay-per-use pricing planned for early 2027. Amazon's sample SP-API MCP servers on npm are now noted under DIY.
- Added Nova, Agent Central, Seller Labs, and Ad Badger.
- Added Atom11, whose MCP server is now live on its own site (November 4).
- Corrected Pacvue to read-only, and Teikametrics now documents its server: 12 read-only tools, waitlist access.
- Updated DataDoe (reads by default, opt-in writes, $97/mo; DSP and AMC removed because DataDoe's own pages don't list them) and Helium 10 (also sold as an add-on from $50/mo; ad changes need your approval).
What is an Amazon MCP server?
The Model Context Protocol (MCP) is an open standard for connecting AI assistants to live external data and tools. An Amazon MCP server implements that standard for Amazon: it authenticates to your account and exposes a set of tools the AI can call -- pull a campaign report, look up a product, list orders -- so the assistant answers with your real numbers instead of guessing.
The servers differ on four things that matter: what data they cover (Ads, Selling Partner, Vendor, DSP, AMC), whether they can write or only read, where they run (hosted vs self-hosted), and what sits on top of the raw data (nothing, or recommendations and automation).
How to choose one
- Data coverage. Ads only, or Ads plus Selling Partner / Vendor / DSP / AMC? Amazon splits its first-party servers into Ads-only and Seller-Central-only, and PPC-focused servers are Ads only; broader coverage comes from optimization platforms and multi-source connectors.
- Read vs write. Read-only is the safest default. If a server can change campaigns, check whether writes are opt-in and confirmed per change.
- Hosting. Hosted means no infrastructure to run; self-host gives you control and data residency but you operate it.
- Security & auth. Prefer Login with Amazon (OAuth) over pasted API keys, and server-side token storage.
- What's on top. Raw API passthrough vs an optimization layer (recommendations, experiments, automation) an operator can actually act on.
The servers
Optimization-native
- Marketplace Ad Pros -- An AI-native Amazon optimization layer over MCP. Launched in May 2025, six months before Amazon's own Ads MCP entered closed beta in November 2025, it covers Amazon Ads + Selling Partner + DSP + AMC in one server, with one connection for every region -- and adds daily optimization recommendations, experiment tracking, direct report SQL, and multi-brand organization on top. Read-only by default; you can opt in to an Edit Mode where the assistant confirms each change before making it. Works with Claude, ChatGPT, Claude Code, n8n, Codex, and OpenClaw, and it ships two more servers -- a Keepa MCP server for price/BSR history and a Seller Central MCP server -- so third-party market data sits next to your own campaign data in one conversation. Plans from $10/week (AI Connect, data access) to $999/mo (Dominion). Amazon Ads Verified Partner. Where it sits: the pick for operators who want analysis and action, not raw API calls. Wrong choice if all you need is raw Ads API access or quick Seller Central answers for a US store -- Amazon's own servers do that without a subscription. See the MCP server.
First-party (Amazon)
Amazon Ads MCP Server (official) -- Amazon's own MCP server, in open beta since February 2026. It turns natural-language prompts into Amazon Ads API calls: create, update, and delete campaigns, run reporting queries, manage account settings, and access billing; its AMC workflow tools work across AMC instances. Read and write. Each region (North America, Europe, and the Far East) has its own endpoint and needs its own connection, so a brand selling in the US, Germany, and Japan connects three times. Amazon's docs name Claude, ChatGPT, Amazon Kiro, Amazon Bedrock and AgentCore, and custom agents as clients, and browser sign-in over OAuth 2.1 arrived in July 2026. In September 2026 Amazon added a Lite endpoint (
/mcp/lite) that exposes three tools -- search, describe, and invoke -- instead of the full tool list, and a closed-beta, US-only tier of expert tools, starting with an AMC SQL generator. Amazon plans general availability for the expert tools in early 2027, priced pay-per-use through AWS Marketplace; it has published no price for the core server. Where it sits: the first-party anchor for Ads -- raw-API access with no Selling Partner data and no optimization intelligence. Access runs through Amazon Ads API credentials, so most advertisers reach it through a partner or developer. Many advertisers will use it alongside a broader server rather than instead of one. Wrong choice if you have no Ads API access, or you want Seller Central data in the same conversation.Amazon Selling Partner MCP (official) -- Amazon's first-party Seller Central server, launched September 23, 2026 at
sellingpartner-ai.amazon.com/mcp. It connects Claude (beta, with a waitlist) and Amazon Quick to your listings, inventory, orders, and performance data, and it drafts changes -- listing fixes, price changes, FBA inbound plans -- that run only once you approve them. US stores only; no advertising or Vendor data. Free during beta. Where it sits: the cheapest way to ask Seller Central questions in Claude. It does not join that data to your ads, so it complements an Ads server rather than replacing one. Wrong choice if you sell outside the US, sell through Vendor Central, or need ad spend next to sales.
Amazon's other September 2026 launch, Amazon Ads Agent (the console that replaces the separate DSP and Ads consoles, with DSP now sold as DVA and DVA+), is an AI assistant inside Amazon's own interface rather than an MCP server, so it is not listed here.
Suite-native and platform-native
Helium 10 MCP -- Helium 10's own connector, released July 2026 at
mcp.helium10.com/mcp. Unlike most servers here it does not front the Amazon APIs directly; it exposes the data Helium 10 already holds, across eight toolsets: keyword research, product research, Brand Analytics, keyword rank tracking, listing analysis and scoring, profit and P&L, inventory and sales velocity, and ads. Their product page now advertises 40+ tools (get_keywords_by_asin,search_amazon_brand_analytics,get_keywords_rank_history,get_listing_score,get_search_query_performance,execute_ads_query, and others). Included with the Diamond plan ($359/mo billed monthly, $279/mo billed yearly) for up to 1,000 calls, and now also sold as an add-on from $50/mo; Helium 10's pricing page doesn't say which plans can buy the add-on. Three things worth knowing. It reads and writes -- campaign changes, bid adjustments, tracked keywords, and listing generation and section rewrites (generate_listing,rewrite_listing_section) -- and Helium 10 says ad changes such as bids and budgets run only with your approval; the V1 shipped read-only, so older comparisons calling it read-only, including earlier versions of this guide, are out of date. It is built for multi-account use -- agencies and multi-brand sellers can pull research, ad, and keyword data for several Amazon accounts through one connection. And it spans 24 marketplaces with nine years of history, which is deeper back-history than any first-party Amazon API will give you. Works with Claude, ChatGPT, Cursor, Codex, OpenClaw, and Muse; auth is OAuth 2.1. Where it sits: the pick if you already pay for Helium 10 -- it answers questions about the market using their proprietary dataset, where first-party servers answer questions about your own account using Amazon's. Wrong choice if you don't use Helium 10, or you need Vendor, DSP, or AMC data.Pacvue MCP -- Pacvue's own connector, marketed as "agent-native data access for retail media." It is three read-only servers -- Report, Data Query, and Share of Voice -- that pull Pacvue's data across Amazon, Walmart and other retailers into your AI client; campaign changes go through Pacvue Agent inside the platform, not over MCP. Its distinguishing property is cross-retailer breadth rather than Amazon depth. Bundled with the Pacvue platform (custom enterprise pricing). Where it sits: the pick if you already run Pacvue and your problem spans retailers -- it answers questions about your whole retail-media footprint, where an Amazon-native server answers them about Amazon in detail. Wrong choice if you aren't a Pacvue customer or want your AI client to make changes.
- Intentwise AI Gateway -- Intentwise's MCP server, exposing its analytics layer across Amazon, Walmart, Criteo, Instacart, TikTok Ads, Google Ads and Meta Ads, with a semantic layer over the top. Intentwise states that data stays on your server when connecting to external agents, which is unusual and matters for buyers with data-residency constraints. Connects to ChatGPT, Claude, Microsoft Copilot, and custom agents. Where it sits: the analytics-first, multi-retailer option, and the strongest fit where privacy review is the gating step. Wrong choice if you want write actions or aren't an Intentwise customer.
- Jungle Scout MCP -- Connects an AI client to Jungle Scout's marketplace intelligence: categories, competitors, brands, sellers, pricing, products, attributes and keywords, plus connected sales and advertising data where available. Like Helium 10's, it mostly fronts Jungle Scout's own data -- it answers "why is a competitor gaining share" better than "why did my ACOS move." Access is tied to a Jungle Scout Cobalt subscription, by demo or account-manager request. Read-only; works with Claude, ChatGPT, Gemini, and Copilot. Where it sits: market research in your AI client; a complement to a first-party server rather than a substitute. Wrong choice if you aren't on Cobalt or want to manage campaigns.
- Teikametrics MCP -- Teikametrics now documents its server in its help center: 12 tools, all read-only (write tools are planned but not available yet). It covers Amazon and Walmart ad performance, Amazon, Walmart, and TikTok Shop catalog profitability, and Smart Pages listing text, over fixed 7-, 14-, 30-, and 90-day windows. Access is by waitlist; Teikametrics publishes setup guides for Claude and ChatGPT, with no separate price. Where it sits: useful if Teikametrics is already your bidding platform. Wrong choice if you need custom date ranges, Seller Central operations data, or write access.
- Ad Badger -- Ad Badger's Amazon Ads MCP server, available since late August 2026 at
ai.adbadger.com/mcp/ads. You sign in with your Ad Badger login, and it is read-only by default: nine tools read profiles, campaigns, keywords, search terms, n-grams, targets, product ads, portfolios, and organic ranks. If you opt in, nine more tools can change Sponsored Products bids and budgets and pause or enable campaigns, keywords, and targets; the assistant asks before each change, and each one is logged in Ad Badger's change history. It cannot create or delete anything, and Sponsored Brands and Sponsored Display stay read-only. Included on every Ad Badger plan, free trials included. Works with Claude, ChatGPT, Claude Code, Cursor, and Codex. Where it sits: the obvious pick if Ad Badger already runs your bids. Wrong choice if you need Seller Central data, Sponsored Brands or Display changes, or don't use Ad Badger. - Atom11 -- Atom11's MCP server, marked live on its site, connects Claude (which Atom11 calls its official integration), ChatGPT, Cursor, and other MCP clients to the Atom11 platform through one endpoint. You sign in with your Atom11 login, the AI can reach only what that user can already see in Atom11, and removing the key in Atom11's settings ends the connection. It reads campaign performance live, explains sales and ACOS changes with Neo, Atom11's AI analyst, and gives cross-brand overviews; agencies use the same connection for several client accounts. Its MCP page says an agent can adjust bids and budgets after a confirmation step, while its pricing page still lists Task AI write access as coming soon, so check which applies to your plan. The MCP page doesn't say which ad types, or whether DSP and AMC data, it covers. Atom11 plans start at $499/mo for up to $50K in monthly ad spend, and Marketing Stream data needs the Professional plan (from $1,199/mo). Where it sits: the pick if Atom11's rules already run your ads and you want to question them from Claude or ChatGPT. Wrong choice if you don't use Atom11: the MCP comes with the platform, not on its own.
Managed hosted connectors
- DataDoe MCP -- A hosted MCP exposing Seller Central, Vendor Central (1P), and Amazon Ads data -- 130+ tables including Marketing Stream and Data Kiosk -- plus a reconciled "true profit" calculation. Login with Amazon, server-side tokens. Reads by default; opt-in writes: actions for listings, prices, A+ Content, FBM shipment confirmations, and Sponsored Products, Brands, and Display campaigns are off until you enable them, with dry runs and logging. Works with Claude, ChatGPT, Gemini, Cursor, Codex, and GitHub Copilot. $97/mo after a 14-day free trial, all accounts included, with paid add-ons for seats, data volume, and buyer PII. Where it sits: a close broad-data alternative -- Seller, Vendor, and Ads in one connection, but data/reporting-first rather than optimization-first. Wrong choice if you want recommendations, or DSP and AMC, which DataDoe's pages don't list.
- Nova -- A hosted MCP server from Nova Analytics (a French company; unrelated to Amazon's Nova AI models) for Amazon sellers, vendors, and agencies. One connection covers Seller Central, Vendor Central, Amazon Ads, and DSP across 23 marketplaces, and Nova models that data into SKU-level profit, reconciling 40+ Amazon fee types with the costs the seller enters (COGS, VAT, one-offs). Nova pitches it as a way for Claude or ChatGPT to answer which products and campaigns actually make money, not just which ones have a good ACoS. 52 tools (39 read, 13 write) at
mcp.novadata.io/api/mcp; writes cover cost inputs, notes, and listing-content edits (titles, bullets, descriptions, keywords, attributes), which need workspace-owner activation and per-change confirmation and never touch price or stock. Works with Claude, ChatGPT, Gemini, Microsoft Copilot, and Grok. From $49/mo per Seller or Vendor Central account, 14-day free trial. Where it sits: the profit-first option, and the better choice than an ads tool if your question is SKU margin after every fee. Wrong choice if you want your AI client to change campaigns -- ad data is read-only and bid/budget writes are listed as "coming soon" -- or you need AMC, which Nova doesn't cover. - Agent Central -- A hosted MCP server from Subsecond Data at
mcp.agentcentral.to/mcp. It joins Amazon Ads (Sponsored Products, Brands, Display, and DSP reporting) with Seller Central inventory, orders, catalog, rankings, finance, and fulfillment across 23 marketplaces, through about 185 tools. Keys can be read-only, scoped to certain domains, or write-enabled; supported writes -- bids, campaign state, keywords, prices, inventory quantities, listings, and MCF orders -- are previewed with current and requested values, run only when you confirm, and are logged. Agent Central describes itself as "not a recommendation engine." Priced by monthly Amazon orders, from $39/mo (0-1,000 orders) to $159/mo (up to 20,000), with extra Ads profiles at $19/mo each and a 14-day trial. Works with Claude, Claude Code, ChatGPT, Cursor, and OpenClaw. Where it sits: a low-cost way to give an agent both Ads and Seller Central data with guarded writes. Wrong choice if you need Vendor Central or AMC, neither of which it lists, or want the server to suggest what to change. - Seller Labs Amazon MCP -- A hosted connector (sold as MCP Genius) that joins Seller Central and Amazon Ads data, using the Finances API to tie settlement cash back to SKUs. It reads and writes campaign changes: every write is logged before it runs, risky moves (a budget over twice the 30-day average, a launch against under 14 days of stock) trigger warnings, and per-change confirmation is set through your AI client's tool permissions. Works with Claude, Claude Code, and ChatGPT. Priced by tool calls: free up to 100 calls a month, paid from $39/mo for 101-400 calls. Where it sits: a light, pay-as-you-go entry point for sellers who want Ads and Seller Central together. Wrong choice for heavy daily use, where per-call pricing adds up, or if you need Vendor, DSP, or AMC data, which it doesn't list.
- PPC Prophet (amazon-ads-mcp) -- A hosted Amazon Advertising MCP with 27+ tools. The free tier is read-only (one profile, 100 calls a day); Pro adds write tools, five profiles, and unlimited calls for $99/mo. Works with Claude Desktop/Code, ChatGPT, and Cursor. Its public repository has not been updated since February 2026. Where it sits: a focused Ads-PPC hosted option -- Ads-only versus broader coverage elsewhere. Wrong choice if you need Seller Central data or more than five profiles.
- Adzviser -- A US-hosted MCP that unifies 40+ marketing platforms (Amazon Ads among them) for cross-channel querying; read-only ("Claude cannot change anything in Amazon Ads"). Priced by number of connected data sources after a $0.99 14-day trial. Where it sits: broad-but-shallow across many ad platforms, rather than deep on Amazon. Wrong choice if Amazon is most of your spend.
DIY / open-source
- openbridge/amazon-ads-mcp -- An open-source, Docker-packaged Amazon Ads API MCP server for self-hosting, made by Openbridge. The Docker image is still published as
openbridge/amazon-ads-mcp; the source now lives atKuudoAI/amazon_ads_mcpon GitHub. Developer-oriented API passthrough. Where it sits: the infrastructure/self-host Ads option, with no managed layer. Wrong choice if nobody on your team will run and patch a server. - Community Selling Partner (SP-API) MCP servers -- Open-source projects on GitHub that expose orders, inventory, pricing, and reports through the Selling Partner API for self-hosters and developers. Amazon also publishes its own sample local SP-API MCP servers on npm (
@amazon-sp-api-release/sp-api-dev-mcp, first published May 2026) and a Data Kiosk sample, which it labels unsupported educational examples. Where it sits: the DIY path to Selling Partner data -- you stand up and credential SP-API access yourself. Wrong choice if you want support or a hosted endpoint.
Comparison table
| Server | Type | Data coverage | Read / write | Hosting | Optimization layer | Pricing |
|---|---|---|---|---|---|---|
| Marketplace Ad Pros | Optimization-native | Ads + SP + DSP + AMC | Read-only default, opt-in confirmed write | Hosted | Yes -- recs, experiments, SQL | $10/week to $999/mo |
| Amazon Ads MCP (official) | First-party | Ads only | Read + write | Hosted (Amazon); one endpoint per region | No | Ads API access; expert tools pay-per-use from early 2027 |
| Amazon Selling Partner MCP (official) | First-party | Seller Central, US only (beta) | Read + approval-gated write | Hosted (Amazon) | No | Free in beta |
| Helium 10 MCP | Suite-native | H10 research + Brand Analytics + rank tracking + ads + profit/inventory; 24 marketplaces, 9 yrs history | Read + write (ad changes need approval) | Hosted | Yes -- their own | Diamond, $279-359/mo; add-on from $50/mo |
| Pacvue MCP | Platform-native | Pacvue cross-retailer data (reports, data query, SOV) | Read-only | Hosted | Yes -- their own | Bundled with Pacvue (custom) |
| Intentwise AI Gateway | Platform-native | Intentwise analytics: Amazon, Walmart, Criteo, Instacart, TikTok, Google, Meta | Read-only | Hosted | Yes -- their own | Bundled with Intentwise |
| Jungle Scout MCP | Suite-native | JS marketplace intelligence + connected sales/ads where available | Read-only | Hosted | No | Bundled with Jungle Scout Cobalt |
| Teikametrics MCP | Platform-native | Teikametrics data: Amazon + Walmart ads, TikTok Shop catalog | Read-only (12 tools) | Hosted | Yes -- their own | Bundled; waitlist |
| Ad Badger MCP | Platform-native | Ads (SP, SB, SD) | Read-only default; opt-in SP bid/budget/status writes, confirmed per change | Hosted | Yes -- their own | Included on all Ad Badger plans |
| Atom11 MCP | Platform-native | Atom11 ad and retail analytics | Read; bid/budget changes after confirmation (per its MCP page) | Hosted | Yes -- their own (Neo) | Comes with Atom11, from $499/mo |
| DataDoe MCP | Hosted connector | Seller + Vendor + Ads | Read by default; opt-in writes | Hosted | No (profit calc) | $97/mo |
| Nova | Hosted connector | Seller + Vendor + Ads + DSP; 23 marketplaces | Read (ads); opt-in listing edits | Hosted | No (profit calc) | From $49/mo per account |
| Agent Central | Hosted connector | Ads + DSP + Seller Central; 23 marketplaces | Scoped keys; previewed, confirmed writes | Hosted | No | $39-159/mo by order volume |
| Seller Labs Amazon MCP | Hosted connector | Seller Central + Ads | Read + logged write | Hosted | No | Free (100 calls/mo); from $39/mo |
| PPC Prophet | Hosted connector | Ads only | Read (free) / write (Pro) | Hosted | No | $99/mo Pro |
| Adzviser | Hosted connector | Ads (+ 40 platforms) | Read-only | Hosted | No | By number of data sources |
| openbridge | DIY open-source | Ads only | Read + write | Self-host | No | Free to self-host |
| Community SP-API servers | DIY open-source | Selling Partner | Read + write | Self-host | No | Free to self-host |
Third-party pricing and features are as published by each vendor and can change -- verify on the vendor's own page before deciding.
How Marketplace Ad Pros fits in
Marketplace Ad Pros is listed first because it is the optimization-native entry, not because this is a ranking. It is honest about the trade-off: if all you want is raw Amazon Ads API access, Amazon's own official server or a DIY passthrough will do it, and some advertisers will run those alongside MAP.
Several servers here now join Ads with Seller Central data -- Agent Central, Nova, Seller Labs, and DataDoe among them. What MAP adds is coverage and action in one place: Amazon Ads, Selling Partner, DSP, and AMC in a single MCP server, next to Keepa market data through its Keepa MCP server. And instead of only returning data, it layers daily optimization recommendations, experiment tracking, and direct report SQL on top -- the things an operator actually wants from an AI client. On security, it is read-only by default: you opt in to edits, and the assistant confirms each change. Where something else fits better, use it: Amazon's free Selling Partner MCP for basic US Seller Central questions, Nova for fee-reconciled SKU profit built on your own costs, and Ad Badger, Atom11, or Teikametrics if one of them already runs your bids. Connect MAP in Claude, ChatGPT, or any MCP client via the setup guide.
How to pick
Want first-party Amazon Ads API access with nothing in the way? Amazon's official Ads MCP Server is the anchor.
Want free, first-party Seller Central answers in Claude for a US store? Amazon's Selling Partner MCP, approval-gated and in beta.
Want Selling Partner data and an optimization layer inside Claude or ChatGPT? Marketplace Ad Pros spans Ads + Selling Partner + DSP + AMC with recommendations and experiments, read-only by default.
Want managed, multi-source data (Seller + Vendor + Ads) with a profit view? DataDoe and Nova are the closest broad connectors. Agent Central and Seller Labs join Ads and Seller Central with guarded writes, and PPC Prophet is the focused Ads-only hosted option.
Already paying for Helium 10 and want keyword, competitor, and Brand Analytics research in your AI client? Their own MCP is included on Diamond -- it answers market questions rather than account questions, so it complements a first-party server more than it replaces one. If Ad Badger, Atom11, Pacvue, or Teikametrics already runs your ads, start with that platform's own MCP.
Want to self-host and control the infrastructure? openbridge (Ads), Amazon's sample SP-API servers, and community SP-API servers (Selling Partner) are open source.
Frequently asked questions
What is an Amazon MCP server?
MCP (Model Context Protocol) is an open standard that lets AI assistants like Claude, ChatGPT, and Gemini connect to live external data. An Amazon MCP server exposes your Amazon Advertising and/or Selling Partner data -- and sometimes DSP and AMC -- to those AI clients, so you can query performance, get recommendations, and take action in plain language instead of exporting spreadsheets.
Does Amazon have an official MCP server?
Yes, two. Amazon's official Amazon Ads MCP Server has been in open beta since February 2026. It turns natural-language prompts into Amazon Ads API calls -- creating and updating campaigns, running reports, and accessing billing -- and works with clients such as Claude, ChatGPT, and Amazon's own Kiro; it covers Amazon Advertising only. On September 23, 2026 Amazon added a separate Amazon Selling Partner MCP for Seller Central data, with changes made only after you approve them. It is in beta in Claude and Amazon Quick, for US stores only, and it does not include advertising data.
Can I connect Amazon Ads to Claude or ChatGPT?
Yes, through an Amazon Ads MCP server. Options include Amazon's own official server, Marketplace Ad Pros (which also adds Selling Partner, DSP, and AMC data plus optimization recommendations), hosted connectors like DataDoe, Agent Central, Seller Labs, Nova, and PPC Prophet, and platform connectors from Ad Badger, Atom11, and Teikametrics for their own users. Helium 10 also ships an MCP server on its Diamond plan, which reaches ad performance through their own ads layer alongside their keyword and competitor research.
What is the difference between Amazon's official MCP server and third-party ones?
Amazon's servers give raw, first-party API access, split in two: the Ads MCP covers Amazon Ads only, and the newer Selling Partner MCP covers Seller Central only (US stores, beta). Third-party servers add things the first-party servers do not: Ads and Selling Partner data in one place, Vendor data, one connection across every region, cross-account organization, and -- in Marketplace Ad Pros' case -- an optimization layer with daily recommendations, experiment tracking, and automation on top of the data.
Is an Amazon MCP server safe to use?
It depends on the server's design. Look for Login with Amazon (OAuth) rather than pasted credentials, and check whether the server is read-only or can write to your account. Marketplace Ad Pros is read-only by default; edits are opt-in and confirmed per change. Amazon's official servers and several hosted options support writes, so review each server's permissions before connecting.
Which Amazon MCP server covers Selling Partner data, not just ads?
Amazon's own Selling Partner MCP (launched September 2026, beta, US stores only) covers Seller Central data with approval-gated changes, but no ads data. To see Selling Partner data next to your ads, Marketplace Ad Pros covers Ads plus Selling Partner, DSP, and AMC; DataDoe and Nova cover Seller, Vendor, and Ads; and Agent Central and Seller Labs cover Seller Central and Ads -- all as managed servers. There are also open-source SP-API MCP servers you can self-host, including Amazon's own unsupported samples.
Can I self-host an Amazon MCP server?
Yes. Community projects like openbridge's Ads server and various Selling Partner API servers on GitHub are open source and self-hostable. Amazon also publishes sample SP-API MCP servers on npm, which it labels unsupported educational examples. Most managed servers also offer a hosted endpoint so you do not have to run infrastructure yourself.
Do MCP servers change my campaigns automatically?
Only if you let them, and the answer changes as servers ship new versions. Amazon's official Ads server and PPC Prophet Pro can write directly. Amazon's Selling Partner MCP drafts changes for your approval, Helium 10 runs ad changes such as bids and budgets only with your approval, Ad Badger and Agent Central keep writes opt-in with a confirmation or preview step, DataDoe's writes are off until you turn them on, and Seller Labs logs every write before it runs. Pacvue, Intentwise, Jungle Scout, Teikametrics, and Adzviser are read-only, and Nova reads ad data only, editing listing content once you enable it. Marketplace Ad Pros is read-only by default and confirms each change before making it. Always confirm a server's current read/write scope before connecting; this is the fastest-moving detail in the category.
Related guides
- Best Amazon ads management tools in 2026
- Amazon advertising automation tools, compared (2026)
- Best Amazon market intelligence & data tools in 2026
- MCP setup guide
Sources
- Amazon Ads MCP Server (official) -- Amazon Advertising
- Amazon Ads API release notes (Lite endpoint, OAuth 2.1) -- Amazon Advertising
- Expert tools bring AMC SQL generation to agents -- Amazon Ads
- Amazon Selling Partner plugin for Amazon Quick and Claude -- About Amazon
- Amazon Selling Partner MCP -- Claude connectors directory
- Amazon SP-API sample MCP servers -- amzn/selling-partner-api-samples
- Model Context Protocol -- introduction
- Helium 10 MCP
- Helium 10 pricing
- helium10/Helium10-MCP -- Helium 10's public MCP repository
- Pacvue MCP -- agent-native data access for retail media
- Intentwise AI Gateway (MCP)
- Jungle Scout MCP
- Teikametrics MCP overview -- Teikametrics help center
- Ad Badger -- Amazon Ads MCP
- DataDoe -- Amazon MCP
- DataDoe pricing
- Nova -- Amazon MCP server
- Agent Central -- hosted Amazon MCP server
- Seller Labs -- Amazon MCP
- PPC Prophet -- amazon-ads-mcp
- Adzviser -- Amazon Ads to Claude
- openbridge -- amazon-ads-mcp (Docker Hub)
- KuudoAI/amazon_ads_mcp -- source repository
- Atom11 MCP
- Atom11 pricing
Server capabilities and vendor MCP availability above were first checked against each vendor's own site on August 26, 2026. On November 1, 2026 every entry was re-checked against the vendor's own pages -- Amazon's Ads and Selling Partner MCPs and SP-API samples, Helium 10, Pacvue, Intentwise, Jungle Scout, Teikametrics, DataDoe, PPC Prophet, Adzviser, and openbridge -- and Nova, Agent Central, Seller Labs, and Ad Badger were added from their own documentation. Atom11 was added from its own MCP and pricing pages on November 4, 2026. Perpetua and Quartile had published no MCP server as of November 1, 2026.
Which half of your ad spend is wasted?
Stop guessing. Connect your Amazon account and ask Claude or ChatGPT — your own numbers, in the AI client you already use.
Connect freeFree Amazon Ads Skills
Ready-made Claude skills for wasted ad spend, search-term harvest, title optimization and more. Free to download.
Browse the skillsWhy Sellers Choose MAP
-
Easy AI Integration via MCP
-
Slash analysis time by 80%
-
Know which bids & keywords to adjust
-
Track experiments without spreadsheets
Have Questions?
Our team is ready to help you optimize your Amazon ad campaigns.
Contact UsSetup Guides
MCP Server Setup
Claude Integration
ChatGPT Integration
Microsoft Copilot Integration
Google Gemini Integration
Cursor Integration
mcp.json entry & sign-in