
TL;DR
This glossary defines the 30 terms you meet when you connect Claude, ChatGPT or another MCP client to Amazon Ads through Xnurta MCP, from MCP and OAuth to managed groups and preview and confirm. It is for Amazon advertisers and agency teams who already use an AI assistant for ad analysis and want the setup docs to make sense. The terms are grouped by what you are doing: connecting, reading, changing, or teaching the assistant a routine.
An MCP connection has a server that offers capabilities, a client that uses them, and tools that name what the assistant can do. Xnurta MCP connects Claude, ChatGPT and other MCP clients to your Xnurta data. Every term below shows up in the setup steps or in what the assistant tells you, and where Xnurta has an exact address, screen label or tool name, it is spelled the way you will see it.
| Term | What it means in Xnurta MCP |
|---|---|
| MCP | A shared standard for connecting AI assistants to systems, often described as a USB-C port for AI. A system builds one connector and any MCP-capable assistant plugs into it. The name stands for Model Context Protocol. |
| API | The way a system opens its capabilities to developers and software. It hands back raw data, not a finished answer, and was never designed for an AI to use directly. |
| MCP server | The system side of the connection, the part that offers capabilities. Xnurta's is added to a client as a remote or Streamable HTTP MCP server. |
| Amazon Ads MCP server | Amazon's own MCP server for Amazon Ads. It is a separate thing from Xnurta MCP, which is what the rest of this page defines. |
| MCP client | The AI assistant that connects to a server. Claude, Claude Code, ChatGPT, ChatGPT Codex, Cherry Studio, WorkBuddy and other MCP-compatible assistants all count. |
| Tool | One specific capability the assistant can call, such as querying ad performance, settings or the operation log. Tools answer what the AI can query. |
| Endpoint | The address of Xnurta's MCP server: https://mcp.xnurta.com/mcp. Every client you connect points at this one address. |
The API row is the one people trip on. MCP does not replace APIs. It standardizes the same capabilities so any AI can use them, and that matters once you count the wiring: three assistants and three systems the old way means nine custom connections. MCP vs API vs dashboard for retail media teams compares the three approaches, and What is an Amazon Ads MCP? covers the idea from the top.
You point your client at the endpoint and authorize with OAuth or an MCP token. Claude desktop, Cowork and claude.ai have no field for a token, so they use a custom connector and browser sign-in. Claude Code and ChatGPT Codex can configure themselves from one prompt that points at the skills repository.
| Term | What it means in Xnurta MCP |
|---|---|
| Connector | The link between an AI assistant and an MCP server, as Claude names it. Adding one gives the assistant access to Xnurta. It never installs skills. |
| Custom connector | A connector you add yourself by entering an address. On Free, Pro and Max: Customize, Connectors, +, Add custom connector, enter the endpoint, Add. On Claude Team or Enterprise, an Owner or Primary Owner adds it once under Organization settings, Connectors, and each member then connects on their own. |
| OAuth | Browser sign-in with your own Xnurta login. There is no token to create or store, and you see only the ad accounts and permissions your login already carries. |
| MCP token | A credential you create in Xnurta under MCP & Skill, New token, with the scope you choose. Copy it once, because it cannot be viewed again. |
| Permission scope | The permission that decides what a token or authorization may do. Each of the three kinds of write has its own, so a scope that edits ads does not necessarily create campaigns. |
| get_user_authorized_context | The tool that checks your connection. It returns your user ID and your authorized profile IDs. A 401 with a token means the token is wrong or lacks the scope. A 401 with OAuth means the authorization expired or lacks the new scope, so disconnect and reconnect. |
On Claude Team or Enterprise, a member will not find "Add custom connector" until an Owner has added it, and that is expected. Anthropic documents who can add connectors in Get started with custom connectors using remote MCP. The full walk-through is in Connect Claude or ChatGPT to Xnurta MCP.
It can read ads performance, entity metadata and the operation log, and all three are read only. That is the safe half of the glossary: you can ask as many questions as you like and your account stays exactly as it was.
| Term | What it means in Xnurta MCP |
|---|---|
| Ad account | The advertising account the assistant works on. You see only the ad accounts your own Xnurta login carries, main accounts and sub-accounts included. |
| Query | A request that reads data. Queries on ads performance, entity metadata and the operation log never modify your account. |
| Ads performance | Numbers for a time range you name, by campaign, ad group, target, search term, placement, product ad or ASIN. Daily by default, hourly for SP, SB and SD campaigns and SP keywords. Campaign performance can also be broken out by audience. |
| Entity metadata | The settings of what sits in your account, with no time range needed: campaigns, ad groups, targets, product ads, ASINs with title, inventory and ad-type eligibility, managed groups, and which campaigns have an audience bound and at what bid uplift. |
| Operation log | The record of human and AI actions on campaigns, ad groups, targets, bids, budgets and managed groups. Filter it by entity type, operator, action type, ad type, targeting type, placement type or Amazon entity ID. |
| Operator | The operation log filter for who acted: the AI, a human, an automation rule, or one user's email. |
History reaches back about 15 months across 22 marketplaces. Freshness follows the Xnurta platform's update cadence, so treat an answer as current to the last refresh, not to the second. If the question is "who touched this budget?", Who changed my Amazon campaign? shows the operator filter at work.
It can change three kinds of thing, each behind its own permission scope: AI managed groups, the ads and campaigns themselves, and new Sponsored Products campaigns. Reading needs none of this. Writing needs write access, and write access is something your CSM switches on.
| Term | What it means in Xnurta MCP |
|---|---|
| Write access | Permission for the assistant to change your account instead of reading it. It covers all three kinds of write. See the access rule below. |
| Managed group | A group of campaigns that Xnurta's AI manages under shared settings: AI status, personality, goal, target ACOS, total budget and action space. Through MCP you can create, edit, schedule and delete them. |
| Action space | The list of things the AI may do inside a managed group, such as placement bid adjustment, dayparting bid adjustment, budget dynamics, negative targeting and target harvesting. Each one switches on or off. |
| Ad-entity editing | Changing the ads themselves: campaign state, daily budget, bidding strategy, products, negatives, placement and audience bid adjustments, and keyword and product targets with their state and bid. Bulk edits are supported. |
| Campaign creation | Building SP campaigns, with ad groups, products, targeting, negatives, placement bid adjustments and an optional AMC audience, in one request. It is atomic: if any campaign fails validation, none are created. |
| Preview and confirm | The rule for money, bids and archiving. The assistant shows which objects change and from which value to which value, and submits only after you confirm. |
| Archive | Retiring an ad entity for good. It is irreversible. Deleting a managed group archives it, and campaigns cannot be deleted at all, so pause when you only want delivery to stop. |
The access rule
Generally available from 27 Oct for Sponsored Ads. Write access is turned on per account by your CSM.
Some managed-group settings stay in the platform UI. Through MCP you can write the name, AI status, personality, goal, target ACOS, total budget, campaign markers and membership, action space on or off, schedule and template. Brand, non-brand and competitor mode, and every word-list feature, are platform UI only. Read-only vs read-write MCP for Amazon advertisers explains which access to start with, and Read and change Xnurta's AI levers through MCP goes deeper on the managed-group levers.
A tool is something the assistant can call. A skill is the routine that tells it how to use tools for one type of task. Take a weekly report: the tools can pull the numbers, but the skill decides what to pull, in what order, and what the finished report looks like.
| Term | What it means in Xnurta MCP |
|---|---|
| Skill | A routine that teaches the assistant how to handle a type of task: what data to query, in what order to analyze it and what format to return. Tools answer what the AI can query. Skills answer how. |
| Official skill | A skill Xnurta provides. Insights and reports: Search Term Traffic Diagnosis, Weekly Ads Report, Product Diagnosis, Monthly Ads Report, Ads Structure Analysis. AI managed groups: Create AI Managed Group, Edit AI Managed Group, Delete AI Managed Group. Campaign tools: Promo Adjustment Reset, Amazon Ads Performance Query Tool, Amazon Ads Configuration Query Tool, Amazon Ads Operation Log Query, Amazon SA Ads Edit Tool, Amazon SA Campaign Creation. |
| Custom skill | A skill you build yourself by describing the scenario, setting the rules and defining the output. You keep it private or share it with your organization. |
| Skills repository | Xnurta's GitHub repository at https://github.com/Xnurta/Xnurta-MCP. Its instructions let an assistant configure MCP and install the skills. |
Adding a connector never installs skills. In a new chat, ask your assistant to follow the instructions in the Xnurta MCP skills repository, and if it connects MCP but skips the skills, send the link again. How an SOP becomes something an AI assistant can run shows how a written procedure becomes a skill your team can run.
A few pairs of terms blur together, and each blur has a short answer. A connector is access, a token is a credential, a skill is a routine. Connecting does not install a skill, and a token does not work in Claude desktop, Cowork or claude.ai.
A scope is not the same as write access. Write access is switched on for your account by your CSM, while a scope is the specific permission a token or authorization carries, and each of the three kinds of write has its own. After new access is switched on, remove the connector and add it again, or create a new token, because signing in again on the old connector keeps the old permissions. Then update your skills.
Archive is not pause. Pause stops delivery and you can turn it back on. Archive cannot be undone. And a query is not a write: asking never changes anything, while anything touching money, bids or archiving shows a preview and waits for your yes.
FAQ
A connector gives your assistant access to Xnurta. A skill teaches it the routine for a type of task, such as a weekly report. Adding a connector never installs skills.
OAuth signs you in through your browser with your own Xnurta login, so there is no token to create or store. An MCP token is a credential you create in Xnurta and add to your client by hand, and Claude desktop, Cowork and claude.ai have no field for one.
No. Queries on ads performance, entity metadata and the operation log are read only. Only the three kinds of write change your account, and they need write access.
A scope is the specific permission a token or authorization carries. Each of the three kinds of write, managed group management, ad-entity editing and SP campaign creation, has its own, so a scope that edits ads does not necessarily create campaigns.
It is https://mcp.xnurta.com/mcp. You enter it as a custom connector in Claude, or an MCP client uses it when it configures Xnurta MCP.
Xnurta MCP
Xnurta MCP is for Xnurta customers only. Create your MCP connection directly in the Xnurta platform. New to Xnurta? Book a demo and we'll walk through your account with you.