What’s in this article
- What Meta Ads AI Connectors are — The new way to manage campaigns with plain English.
- 5 Starter Prompts to Copy — The exact prompts I use for a 2-minute ad account audit.
- How I set this up — A step-by-step guide to connect Claude to your Meta Ads account.
- What this changes for agencies — How this shifts ad management from clicking to conversing.
- FAQ — Answers on security, cost, and which AI clients work.
🚀 Plug this into Claude Code or Claude Desktop
This post gives you the prompts. The downloadable spec file gives Claude the full context: the 29 tools available, their categories, and how to structure requests to the Meta Ads MCP server. It’s the fastest way to get started.
Want hands-on help wiring this into your agency’s workflow? That’s what the Talk-to-Build community is for.
I’m going to show you how to run your Meta ad account from a chat window. You will get the five starter prompts I use to do a full account audit in about two minutes — something that used to take 30 minutes of clicking around in Ads Manager.
Meta recently launched Ads AI Connectors, a free MCP server that gives AI assistants like Claude and ChatGPT direct access to 29 different ad management tools. You can create campaigns, check pixel health, and fix catalog errors in plain English. Here is exactly how I’m using it.
What are Meta Ads AI Connectors?
Meta Ads AI Connectors are a set of tools that allow AI assistants like Claude and ChatGPT to directly manage a Meta Ads account using natural language. The system provides access to 29 distinct functions across campaign creation, catalog management, and performance insights through an MCP server. This eliminates the need for manual work in the Ads Manager interface for many common tasks.
The 5 Meta Ad Prompts I Start With — Copy Them Right Now
These are the first five prompts I run to get a complete, high-level audit of any ad account. This sequence checks performance, diagnostics, product data, and audience health before I even think about creating a new campaign. You can copy and paste these directly into any MCP-compliant client like Claude Desktop once you’re connected.
# 1. Performance Audit
List all active campaigns sorted by highest CPA in the last 30 days. Include spend, conversions, and ROAS for each.
# 2. Pixel Health Check
Check my pixel events and flag any mismatches or drops in quality between browser and server events in the last 7 days.
# 3. Catalog Quality Scan
Find all product catalog items with quality issues. For each, describe the specific error and list the affected product IDs.
# 4. Audience Overlap
Show all audiences across active ad sets and flag any with a significant overlap that could be causing self-competition.
# 5. New Campaign Draft
Create a new campaign for my 'Summer Sale' collection, targeting 'Lookalike Audience 1%', with a $50/day budget. Use https://mystore.com/summer-sale as the landing page. Leave the campaign in paused status for my review.The most important part is the safety measure: any campaign, ad set, or ad you create via the connectors is automatically set to PAUSED. Nothing goes live until you manually review and activate it in Ads Manager.
┌────────┐ ┌────────────┐ ┌──────────┐ │ Claude │────>│ Meta's MCP │────>│ Meta Ads │ └────────┘ │ Server │ │ (Paused) │ (Prompt) └────────────┘ └──────────┘
Here’s exactly how I’d do this
Connecting your ad account to an AI assistant is a one-time setup that takes about five minutes. The process uses a standard OAuth flow, so you don’t need to generate API keys or create a developer app. My setup uses Claude Desktop, but the steps are nearly identical for ChatGPT Plus or Cursor.
- Open your AI client. In Claude Desktop or on Claude.ai, start a new conversation and look for the option to add a tool or connection. This is usually a small plus icon or a ‘Tools’ menu.
- Find the Meta Ads AI Connector. Search for ‘Meta’ in the tool directory. You should see the official ‘Meta Ads AI Connectors’ tool. Select it to begin the authentication process.
- Authorize with Meta Business. You’ll be redirected to a Meta login screen. Log in to the Meta Business account that owns the ad account you want to manage. Grant the requested permissions for the AI to access and manage your ads.
- Run your first prompt. Once connected, you’re ready to go. Copy the first prompt from the asset block above (“List all active campaigns…”) and paste it into the chat. The AI will now use the MCP tool to fetch the data directly from your account.
What this changes for designer-run agency work
This tool changes ad management from a series of clicks in a complex interface to a conversation about outcomes. For agencies, this shift impacts how we audit accounts, manage campaigns, and report to clients. The core task moves from manual data pulling to strategic question asking, which is a higher-value service.
| Dimension | Old way (Manual Ads Manager) | New way (Conversational MCP) |
|---|---|---|
| Account Audits | 30-60 minutes of clicking through 10 different reports. | 2 minutes running 5 standard prompts for a complete overview. |
| Campaign Creation | Manual setup through a multi-step wizard, prone to user error. | Describe the campaign in one sentence; the AI builds the draft. |
| Troubleshooting | Digging through Pixel Helper and diagnostics tabs to find errors. | Ask, “Are there any pixel or catalog issues?” and get a plain-English list. |
| Client Reporting | Exporting CSVs and building custom dashboards or slide decks. | Ask the AI to summarize performance and generate the report text. |
This is a big deal for smaller, design-led agencies that don’t have a dedicated media buyer. It makes sophisticated ad management accessible without needing to master the entire Ads Manager UI. For more industry intel like this, I keep a running log of how these tools are changing agency work.
My $0.02 — How I’d roll this out
If I were running an agency, I wouldn’t just turn this on for everyone overnight. I’d use a phased, three-day rollout to test the workflow, build trust in the tool, and integrate it safely into my client management process. The goal is to make my team faster and more accurate, not to replace them.
Day 1 — Internal test on one account. I’d connect the tool to my own agency’s ad account first. I’d have my team spend the day running the 29 available tools. The goal is to learn the system’s capabilities and limitations on a non-client account. I’d have them document which prompts give the most useful outputs.
Day 2 — Shadow-mode on a friendly client. I’d pick one trusted client and use the MCP prompts to generate my next weekly report for them. I would still do the manual pull from Ads Manager to verify the AI’s data is 100% accurate. This side-by-side comparison builds confidence that the tool is reliable.
Day 3 — Build standardized prompt library. Based on the first two days, I’d create a library of 10-15 standard prompts for my agency. These would cover our core services: weekly performance reports, new campaign drafts, and technical health checks. This turns the tool from a novelty into a repeatable, scalable part of our standard operating procedure.
FAQ
Is this free to use?
Yes, the Meta Ads AI Connectors are free during the open beta period, which started on April 29, 2026. Meta has not yet announced if there will be a cost or what the pricing will be after the beta ends.
Is it safe to give an AI access to my ad account?
It is designed with a key safety feature: any new campaign, ad set, or ad created via the connectors is automatically placed in a PAUSED status. Nothing goes live without a human manually reviewing and activating it inside the standard Meta Ads Manager.
Which AI assistants are compatible?
The connectors work with any MCP-compliant client. This includes Claude (Desktop and claude.ai), ChatGPT (Plus or higher), and developer-focused editors like Cursor. The authentication uses a standard Meta Business OAuth flow.
How many tools are included?
There are 29 tools available in total. They are grouped into five categories: Campaign Creation & Management (5 tools), Product Catalog (10), Accounts/Pages/Assets (3), Dataset Quality & Diagnostics (4), and Insights & Performance (7).
Can I manage multiple client ad accounts with this?
The documentation isn’t explicit on multi-account management in a single session. The current workflow seems optimized for connecting to one ad account at a time. Agencies would likely need to re-authenticate to switch between different client accounts.
Does this replace the need for a human media buyer?
No. It replaces the tedious, manual parts of the job like pulling reports and setting up campaign shells. The strategy—choosing audiences, writing copy, analyzing performance, and making decisions—still requires a human expert. This tool makes that expert faster.
Want help applying this?
Four ways to go deeper:
- Build with Builders. Join the Talk-to-Build community to learn to build AI-native websites, cinematic AI video, and agent-driven workflows you can sell.
- 1-on-1 working session. Book a screen-share with me — bring a real problem, leave with a working piece of it.
- Done-for-you. MK-Way builds AEO-ready websites, apps, and AI agent workflows.
- Quick question. DM me on Instagram or LinkedIn. I read every message.
Part of the AI Pulse series. If you commented “META-MCP” on one of my videos — this is the breakdown.
Last updated: 2026-07-13.