Yes, connecting Supermetrics to Claude is safe. Claude never connects to your ad platforms directly, only ever pulls your data when you ask, and makes changes to your campaigns in a paused state (you have to manually approve them before they go live). On a business plan, Anthropic doesn’t use your data to train its models by default — and you can disconnect the integration any time.

With Supermetrics for Claude, you can turn Claude into your marketing command center. Analyze cross-channel performance data, generate reports and dashboards, and manage campaigns — all from within a Claude chat.

But as exciting as this is, you might be wondering if it’s safe.

You might feel slightly concerned about connecting an AI tool to your live ad accounts and your clients’ data. How exactly does it work? What guardrails are in place? And just how much access does Claude have to your customer data?

These are all questions that came up during our webinar with Layer, the marketing agency that cut reporting time by 30x with Supermetrics for Claude, so we thought we should answer them in detail.

The short answer is that yes, connecting Supermetrics to Claude, ChatGPT, or any other AI tool is safe — and this article explains how.

Key takeaways

  • Claude doesn’t connect to your ad platforms directly. It only talks to Supermetrics, which sits in the middle, so your platform logins never go near Claude.
  • Claude only pulls your data when you ask. It doesn’t have always-on access to your accounts, and it only ever gets the specific data your question needs.
  • You never hand over a password. You authorize a secure, revocable connection through each platform’s own login screen — and you can cut it off whenever you want.
  • On a business plan (Claude for Work or the API), Anthropic doesn’t use your data to train its models by default. And even if you send feedback, your raw marketing data is never included — only the visible chat, with your identity stripped out.
  • The AI can’t change your live campaigns without your approval. New campaigns start paused, every change is logged, and you can undo anything in a click.
  • You can disconnect the integration in a few clicks, at any time.

What actually happens when I connect Supermetrics to Claude?

Connecting Supermetrics to Claude means linking your Supermetrics account to your Claude account. Supermetrics is a marketing intelligence platform that centralizes your marketing data and helps you turn insights into action.

The Supermetrics for Claude integration lets you ask Claude questions about your data. For example, if you ask: “Compare our Meta ads performance in Q2 versus Q3”, it will pull the relevant figures straight from your ad platforms.

However, this connection is bidirectional, which means Claude can also interact with your ad platforms. For example, you can ask Claude to set up a new campaign or pause an underperformer without leaving the AI chat.

When we say “interact with your ad platforms”, we need to make a distinction here.

The Supermetrics for Claude integration is actually a three-way connection, with Supermetrics sitting in the middle. Claude talks to Supermetrics, while Supermetrics talks to your ad platforms.

Claude doesn’t plug into Google Ads, Meta, or any of your other platforms directly. It connects to Supermetrics — and Supermetrics connects to your platforms.

Claude only ever deals with one connection. Your individual ad platform logins stay with Supermetrics.

How does the Supermetrics for Claude integration work?

The Supermetrics for Claude integration process works like this.

Claude connects to Supermetrics using the Supermetrics MCP (Model Context Protocol) server. MCP is a language that lets AI tools and other software talk to each other.

Once that connection’s in place, here’s what happens each time you ask Claude something:

  1. You ask Claude a question, as if you were speaking to a colleague. E.g. “Which of my current campaigns has the lowest CPC?”
  2. Claude sends the request to Supermetrics through the MCP server.
  3. Supermetrics pulls the data from your connected platforms (Google Ads, Meta, Google Analytics, and so on) and cleans it up.
  4. The data travels back through the MCP server to Claude.
  5. Claude uses it to answer your question.

It also works the other way. If you ask Claude to make a change — like pause a campaign or update one — the request follows the same path back through Supermetrics to the right platform.

Note: You still have to approve these changes manually before Supermetrics makes the changes. We’ll touch on this approval step in more detail in question 6.

Does Claude have access to my marketing data?

Yes, but only in a limited and on-demand way.

When you ask Claude a question, it gets the specific data it needs to answer that question. Nothing more.

It doesn’t have the keys to your ad accounts. It can’t log in, look around, or pull whatever it likes. It only gets the specific piece of data you’ve asked about, fetched from the connectors you’ve already approved in Supermetrics.

And this only happens when you ask. Claude doesn’t have a standing, always-on line into your data. Between questions, nothing’s flowing. Supermetrics only hands over data when you’ve actually asked for it.

Claude doesn’t have free rein over your accounts, and it can’t go looking for anything you haven’t asked for.

For agencies, each client's platforms stay connected under their own Supermetrics authorization, so pulling one client's data never exposes another's. Claude only ever sees the specific client account behind the question you asked.

Is this riskier than pulling data manually?

No, the opposite — it's usually safer than exporting CSVs by hand. There's no spreadsheet of client data sitting in someone's downloads, and access runs on a revocable token that logs every change.

Does Claude use my Supermetrics data to train its AI models?

No. The raw data Supermetrics pulls to answer your question is never used to train Claude’s models — on any plan.

This came up a few times in our webinar with Layer. Agencies worried that hooking up Supermetrics for Claude would mean their customer data feeding Anthropic’s next model. It won't.

Anthropic’s model training policy is explicit that data brought in through connectors and MCP servers is left out of training.

“It does not include raw content from connectors (e.g. Google Drive), including remote and local MCP servers, though data may be included if it’s directly copied into your conversation with Claude.”

So the full datasets behind every answer stay out of it. The one thing that can be used is the visible conversation — your questions, and whatever appears in Claude’s replies (e.g. a table of your Q3 Meta ads performance). And even that depends on your plan:

  • On a business plan (Claude for Work or the API), your conversations aren’t used for training by default. The only exception is if you actively rate a reply with thumbs up or down — a deliberate choice to send that conversation to Anthropic. When you do, it’s de-identified first, so it’s not linked back to you or your company.
  • On a personal plan (Free, Pro, or Max), you decide. Whether your conversations are used for training comes down to whether you’ve enabled the setting. To check, or to turn it off, head to Settings → Privacy → Help improve our AI models and toggle the button.

If you’re still unsure what will and won’t be used to train Anthropic’s models, this table clarifies things.

Data typeBusiness plan (default) Business plan, if you rate a reply Personal plan (Free / Pro / Max)
Raw data pulled through the connector / MCP: The full underlying datasets behind every answer Never usedNever usedNever used
The visible conversation: Your questions, Claude's replies (including any figures or tables it shows), and anything you paste inNot usedCan be usedUsed only if you've turned the training setting on

How does Supermetrics keep my data secure?

As well as being SOC 2 Type II, GDPR, and CCPA-compliant, Supermetrics also encrypts your data ((TLS 1.2+) as it travels between systems so it can’t be read if it’s intercepted. The platform is also penetration-tested annually. We hire ethical hackers to try and break in, then fix whatever they find before a real attacker could.

If you’re on an Enterprise plan, single sign-on (SSO) gives your admins central control over who gets access. Your team logs in with your existing company identity — using Google, Microsoft, or Okta — which means that when someone leaves, you cut their access in one place. This means there are no stray logins for potential hackers to exploit.

If you want to find out more about our security documents, check out the Supermetrics Trust Center.

Do I have to hand over my ad platform logins?

No, Supermetrics for Claude doesn’t need you to hand over your passwords.

When you connect a platform to Supermetrics, you authorize the connection yourself, under your own credentials and your own data governance rules. If you ask Claude a question about your data, Supermetrics pulls it from your ad platforms and passes it over to Claude to analyze.

But what about when you make a change to one of your campaigns?

Here’s how that process works:

  1. You connect the platform once: Log in through Google’s own login screen and approve the connection. Supermetrics gets a secure OAuth token to act on your behalf — it never sees or stores your password.
  2. You ask Claude to make a change: For example: “Pause the Summer Sale campaign.”
  3. Claude passes the instruction to Supermetrics: Through the MCP server.
  4. Supermetrics makes the change on the real platform: It uses that token to pause the campaign in Google Ads.

To reiterate: You never hand a password to Claude or Supermetrics — you grant a secure connection through each platform’s own front door. Because it’s a token and not a login, you can revoke it from the platform whenever you want.

Claude never touches your ad platforms directly. It asks Supermetrics, and Supermetrics makes the changes for you.

Can the AI make changes to my live campaigns without my permission?

No, Supermetrics for Claude doesn’t make any changes to your live campaigns without your permission. You need to approve any changes it makes to your Google Ads, Meta, Microsoft Advertising, and TikTok ads before they go live.

When you ask Claude to create a new campaign, it builds it in a paused state. You won’t spend anything until you switch it on yourself. When you ask it to change an existing campaign, it shows you the proposed change first and waits for you to confirm before it touches anything.

Plus, it tracks everything it changes. Every action shows up in your Campaign History and you can undo it in a single click. So even if something goes out that you didn’t mean to, you can roll it back.

However, you can’t undo any budget that you already spent.

Can I disconnect or revoke access whenever I want?

Yes, you can disconnect or revoke Claude’s access to Supermetrics with just a few clicks. To revoke Claude’s access to Supermetrics, all you have to do is:

  1. In Claude, go to Settings → Connectors.
  2. Find Supermetrics in your list.
  3. Click Disconnect.

The moment you disconnect the integration, Claude will no longer be able to access your Supermetrics data.

Get started with Supermetrics for Claude

It’s only right to be cautious about your customer data. But as this article has explained, there’s only one thing to keep in mind when using Supermetrics for Claude: on a personal plan, the training toggle is yours to manage, so check that it's set the way you want. On a business plan it's off by default.

Claude will have limited, on-demand access to your data. It only pulls what

you want, when you want. And it can’t make any changes to your campaigns without you manually approving them first.

Keen to test out the integration for yourself?

If you’re already a Supermetrics customer, you can link your account to Claude and start asking questions straight away. If you’re not yet, you can book a demo and see it in action first.

FAQs