Connect Xnurta to Claude, ChatGPT, or any MCP-compatible assistant. Sign in with OAuth and ask in plain language. Get the answer where you're working.
Every analysis today starts the same way: log in, filter, export, open somewhere else. Xnurta MCP removes those four steps. Your assistant queries your live Xnurta data directly and answers in the conversation you're having.
Spend, sales, ACOS, ROAS, CTR, CVR, CPC by campaign, ad group, target or product ad. TACOS, sessions and Buy Box % at ASIN, parent-ASIN and product-line level.
Campaigns, ad groups, targets, product ads, ASINs, managed groups, product lines. What happened, and how it's all set up.
A full operation log of human and AI actions, filterable by operator, action type, entity, and time. "Who changed this budget, and when?" now has an answer.
Create and update managed groups, set objectives and targets, and configure which levers the AI may use. Your assistant proposes, you confirm, and the optimizer keeps running.
Xnurta knows what happened in your ads account. It doesn't know your unit cost, your margin, or the number you're trying to hit, that data was never ours.
Give your assistant those numbers alongside your ad data, and ask the question you've never been able to ask a dashboard:
"Based on my real margin, which campaigns are losing money, cut them or scale them?"
Two properties built into the design rather than promised in a policy.
Every action on your account, whether by a person, by Xnurta's optimizer, or by your assistant through MCP, lands in the same operation log, filterable by operator, action type, entity, and time. That includes the connection you're using right now: ask "what changed on this account last week, and who changed it?" and the MCP audits itself.
The margin question works because your costs and targets stay on your side. Your assistant holds them, reasons with them alongside your live ad data, and sends Xnurta only questions and instructions, never the spreadsheet. Your most sensitive numbers inform every decision without entering our systems.
They only ever reach your own assistant.
We'd rather you know the boundaries than discover them.
We're publishing all four phases with dates, before we've built all of them, because a roadmap you can see is a roadmap you can hold us to.
Beyond that: AMC, DSP, and expansion beyond Amazon. Coming, and we're not going to put a date on it until we can hold it.
These are closed-beta dates from our engineering roadmap, and each phase enters beta only after it passes its evaluation gate. We haven't set general-availability dates yet, and we'd rather say that than invent one. If a date moves, we'll say so here before you have to ask.
Xnurta MCP is client-agnostic by design. If it speaks MCP, it works. Including whichever assistant your compliance team has cleared.
Prefer a token? For clients that don't support OAuth, generate an API token from MCP & Skill and configure manually.
Read the full setup guide →Access runs through a secure OAuth sign-in with your Xnurta account (or a scoped token, for clients that need one) and it only reaches the stores your account has permission for. You can revoke access at any time from MCP & Skill. Data queries are read-only, and any managed-group change your assistant proposes is executed only after you confirm it.
Insight Agent is our analyst, running inside the platform on the same query engine, ask it for an analysis and it hands one back. MCP is the data line into your assistant, for when you want to combine Xnurta data with your own or work inside your own tools. Both draw on the same query engine, so the underlying figures reconcile. Analysis and interpretation are produced separately, so wording and conclusions will differ.
Talk to your CSM or account owner. Open beta is available to every Xnurta customer, and there's no waitlist or application form to fill in.
Nothing during the beta. We haven't finalised pricing, and we'll tell you before that changes rather than after.
Campaigns, not yet. Bulk campaign edits arrive with Phase 3 (beta from 8 September). Managed groups, yes: Phase 2 is now in beta, so your assistant can create and edit managed groups and configure action spaces, confirming each change with you before it applies. The rest follows the schedule above.