How to Connect Meta Ads to Claude

What connecting Meta Ads to Claude actually means

Claude cannot reach the Meta Marketing API on its own. MCP (the Model Context Protocol) is the open standard that lets an AI assistant talk to external tools through a defined server. The VigilDog Ads MCP is that server: it holds the authenticated connection to your Meta ad account and exposes a set of well-defined actions Claude can call, like pulling campaign performance or drafting a new ad set.

So 'connecting Meta Ads to Claude' really means authorizing the VigilDog Ads MCP to act on your ad account, then adding that connector to Claude. Claude never sees your Meta password; it only sees the specific, scoped actions the server offers.

Before you start

You need three things. First, a Meta ad account you have access to, ideally through Meta Business Manager, with a role that permits what you intend to do (viewing reports needs read access; making changes needs advertiser or admin access). Second, a VigilDog account. Third, a Claude plan that supports custom connectors.

It helps to know your ad account ID ahead of time, the act_ followed by a long number you see in Ads Manager, since you may be asked to pick the right account during setup if you manage several.

Connect the account

The flow is a standard OAuth handoff, which is why you never hand Claude a password.

• In VigilDog, add the Ads MCP and start the Meta connection. You'll be sent to Meta's own login and consent screen.
• Sign in to Meta and grant the ads scopes (read access for reporting, and management access if you want Claude to draft or push changes). Meta shows you exactly which permissions are requested.
• Select the ad account (act_...) you want to connect. If you manage a Business Manager with many accounts, pick the specific one here.
• Back in Claude, add the VigilDog Ads MCP as a connector and complete its sign-in so Claude is authorized to reach the server.
• Confirm the link by asking a simple read-only question, like a summary of last week's spend, before you attempt anything that changes the account.

What you can do, and the guardrails that keep it safe

Once connected, reads are the easy win: ask for spend by campaign, which ad sets are underperforming their target CPA, week-over-week ROAS, or a breakdown by placement, all in plain language. Claude fetches the numbers through the MCP and explains them.

Writes are where discipline matters, and this is deliberately not a system that quietly rewrites your account. VigilDog's Ads MCP is approval-gated: when you ask Claude to create a campaign, adjust a budget, or pause an ad set, it produces a preview of the exact change first and waits for your explicit confirmation before anything touches Meta. Dry-run by default means a misread instruction becomes a discarded draft, not a live budget change. That safety model, and how it works across Google and LinkedIn too, is covered on the Ads MCP page and mirrors the approach in run Google Ads from Claude.

Where this fits

Connecting Meta Ads to Claude does not replace your judgment; it removes the clicking between a question and an answer, and the guessing between a decision and its execution. You stay the approver on every change, which is what makes it safe to use on real budgets.

If you want to try it, see plans and pricing or sign up to link your first ad account. Read-heavy workflows work immediately, and you can decide later whether to grant management access for approval-gated changes.