Connect Claude or ChatGPT to Amazon Ads with Xnurta MCP

What do you need before you connect?

You need an Xnurta login, an AI assistant that speaks MCP, and, on Claude Team or Enterprise, an Owner who adds the connector once. Xnurta MCP connects Claude, ChatGPT and other MCP clients to your Xnurta data. You sign in with your own login, so the AI sees only the ad accounts that login already carries, main accounts and sub-accounts included.

How do you connect Claude desktop, Cowork or claude.ai?

You add Xnurta as a custom connector and sign in through your browser. These three clients cannot configure MCP themselves, and local MCP config files do not exist in Cowork or claude.ai, so the connector is the only road in. On a Free, Pro or Max account, you do it yourself:

1. Open Customize, then Connectors, click +, and choose Add custom connector.
2. Enter https://mcp.xnurta.com/mcp and click Add.
3. Sign in to Xnurta in the browser window that opens. Review the permissions and ad accounts, approve, and return to Claude.
4. Install the skills in a new chat (two sections down).

How does a Claude Team or Enterprise organization set it up?

It takes two stages, because only an Owner can add a connector for the whole organization. If you are a member and cannot find "Add custom connector," that is expected. Ask your Claude admin to do stage one.

1. The Owner or Primary Owner adds it once. Open Organization settings, then Connectors, then Add. Hover Custom, choose Web, enter https://mcp.xnurta.com/mcp and click Add. Leave Advanced settings empty. No OAuth Client ID or Secret is needed.
2. Each member connects on their own. Open Customize, then Connectors, find the Xnurta connector and click Connect. Sign in with your own Xnurta login.

Stage two stays personal on purpose. Adding the connector for the organization does not widen anyone's access, because every member still sees only what their own Xnurta account allows. Anthropic's help center covers who can add connectors in Get started with custom connectors using remote MCP.

How do you install the Xnurta skills?

You ask your assistant to install them, in a new chat, after the connector is in. Adding a connector never installs skills. The connector gives the assistant access to Xnurta. The skills give it the routine for Xnurta tasks, such as what to pull for a weekly report or how to preview a change before it happens.

Please read this GitHub repository and follow the instructions to install the Xnurta Skills: https://github.com/Xnurta/Xnurta-MCP

If the assistant connects MCP but skips the skills, send the repository link again and ask it to install them.

How do you connect Claude Code, Codex or another agent?

You send the agent one prompt and it configures MCP itself. This works for any client that can set up MCP on its own, including Claude Code and ChatGPT Codex.

Please read the following GitHub repository and follow its instructions to configure Xnurta MCP and install the Skills: https://github.com/Xnurta/Xnurta-MCP

When authorization is required, open the sign-in page so I can complete the authorization.

After the configuration is complete, call get_user_authorized_context to verify the connection.

When the client asks, choose OAuth, Sign in or Authenticate. Your browser opens the Xnurta authorization page. Sign in, review the permissions and ad accounts, approve, and go back to the client.

How do you connect ChatGPT?

Codex takes the same one-prompt setup shown above. For ChatGPT apps, add Xnurta MCP with the same address, https://mcp.xnurta.com/mcp, and sign in when the Xnurta authorization page opens. What ChatGPT lets you add depends on your plan and workspace, so read OpenAI's Help Center article "Developer mode and MCP apps in ChatGPT" first.

Can you connect with a token instead?

Yes, in clients that can send a bearer token. Claude desktop, Cowork and claude.ai have no field for one, so use the connector there. In Xnurta:

1. Open MCP & Skill and click New token.
2. Create it with the narrowest scope that covers your first use.
3. Copy the token and store it somewhere safe. You cannot view it again after you leave the page.
4. Send the agent prompt above with one extra line: Token: [paste your token].

How do you check that it worked?

You ask the assistant to call get_user_authorized_context. If it returns your user ID and your authorized ad accounts, you are connected. Then ask something real, and name the account, a time window and a comparison. Comparisons turn numbers into conclusions.

Call get_user_authorized_context and list the ad accounts I'm authorized to access.

For ad account [account name], rank campaigns by ACOS for the last 7 days vs the prior 7 days. Show a table with spend, sales and ACOS for both periods.

Can the AI change your campaigns once it's connected?

Only if write access is on for your Xnurta account. Reading never changes anything. Writing comes in three kinds: AI-managed groups, edits to ads and campaigns, and creating Sponsored Products campaigns. Each kind has its own permission scope, so a token that can edit ads cannot necessarily create campaigns.

Changes to money, bids or archiving never run straight away. The assistant shows a preview first: which objects, and from which value to which value. It submits only after you confirm. Archiving cannot be undone, so pause a campaign if you only want it to stop. Read-only vs read-write MCP for Amazon advertisers explains which access to start with.

What if it doesn't connect?

Most failures trace back to an expired authorization, a missing permission or a skipped step. Match what you see to the row below. Xnurta MCP troubleshooting: 12 errors and fixes covers more errors.