An MCP (Model Context Protocol) server for ad platforms. It lets an AI assistant like Claude or ChatGPT read and manage your Google, Meta or LinkedIn ad accounts through their official APIs, driven by plain-language requests.

Is it safe to let AI manage a live ad account?

With the right guardrails, yes. A good Ads MCP is dry-run by default (it previews every change), gates write actions behind human approval, uses scoped revocable OAuth access, and is policy-aware to avoid suspensions.

Which platforms can I run from Claude?

The VigilDog Ads MCP supports Google Ads, Meta and LinkedIn today, with more platforms rolling out. You connect each account once and control them all conversationally.

How to run Google Ads from Claude with an Ads MCP

An Ads MCP lets you manage real ad accounts by chatting with Claude or ChatGPT: "pause the underperformers", "shift 20% of budget to the top campaign", "pull last week's numbers". Here is what MCP is, how an Ads MCP works, and how it stays safe on live accounts.

What is an MCP — and an "Ads MCP"?

MCP (Model Context Protocol) is an open standard that lets an AI assistant securely use external tools. An MCP server exposes a set of actions the model can call; the assistant decides which to call from your natural-language request.

An Ads MCP is an MCP server for advertising platforms. Connect your Google Ads, Meta or LinkedIn account once, and the assistant can read performance and make changes through the platform's official API — no dashboards, no export-to-spreadsheet, no clicking through five screens to pause a campaign.

What you can do from chat

Anything the platform API supports becomes a sentence:

Reporting: "how did search spend do last week vs the week before?" / "which campaigns are over target CPA?"
Optimisation: "pause ad groups below a 1% CTR" / "raise the budget on the top ROAS campaign by 15%".
Housekeeping: "add these negative keywords across the account" / "pull the disapproved ads and tell me why".
Cross-platform: the same conversational control across Google, Meta and LinkedIn from one place.

How it stays safe on a live account

Letting an AI touch a live ad account only works with guardrails. A well-built Ads MCP is scoped, reversible and explicit:

Dry-run by default: it shows you exactly what it will change — the campaign, the budget, the before-and-after — before anything touches the account.
Approval gate: write actions require a confirmation step, so nothing changes without a human "yes".
Scoped and revocable access: you connect through the platform's OAuth and can revoke it any time; your passwords are never shared.
Policy awareness: it knows the platform rules and refuses actions that could get an account suspended.

Getting started

The VigilDog Ads MCP connects Google, Meta and LinkedIn to Claude and ChatGPT with all of the guardrails above, dry-run first. If you also want your infrastructure watched, it shares one account with VigilDog's domain, SSL, DNS and email monitoring. Start free and connect your first platform.

Frequently asked

What is an Ads MCP?: An MCP (Model Context Protocol) server for ad platforms. It lets an AI assistant like Claude or ChatGPT read and manage your Google, Meta or LinkedIn ad accounts through their official APIs, driven by plain-language requests.
Is it safe to let AI manage a live ad account?: With the right guardrails, yes. A good Ads MCP is dry-run by default (it previews every change), gates write actions behind human approval, uses scoped revocable OAuth access, and is policy-aware to avoid suspensions.
Which platforms can I run from Claude?: The VigilDog Ads MCP supports Google Ads, Meta and LinkedIn today, with more platforms rolling out. You connect each account once and control them all conversationally.