What Happened
On April 29, 2026, Meta did something that marks a milestone in digital advertising history: opened Ads MCP (Model Context Protocol) connector, allowing AI assistants to directly operate Facebook and Instagram ad accounts.
This isn’t some minor “AI helps you read reports” thing. Meta opened 29 tools at once, covering everything from creating ad campaigns to industry benchmarking. All you need to do is authorize your Meta ad account in Claude, and you can manage ads using natural language.
Put more plainly: everything agencies used to charge you 20% service fees to do, now a skilled advertiser with AI can do on their own.
What Can These 29 Tools Actually Do?
Meta grouped these 29 tools into five categories, each directly corresponding to core ad operations:
Ad Creation & Management (5 tools)
This is the most sensitive part, because every action involves real spending:
- Create ad campaign: Set objectives and budget
- Create ad set: Set audience targeting and scheduling
- Create ad: Link creatives and placements
- Edit existing settings: Adjust budget, audience, scheduling
- Restart paused ads: One-click resume
In the past, you had to click through Ads Manager one by one. Now you just tell Claude “create a conversion campaign with a $500 daily budget, targeting women aged 25-35 interested in fitness,” and it’s done.
Product Catalog Management (10 tools)
If you run e-commerce, this set of tools has the most value:
- Create and manage product catalogs
- Diagnose feed errors
- Check product data details
- Manage product collections
This directly maps to Advantage+ Shopping and dynamic product ads. In the past, feed errors required engineer troubleshooting. Now Claude can directly tell you what’s wrong and how to fix it.
Account & Assets (3 tools)
Basic but essential:
- List accessible ad accounts
- List all campaigns, ad sets, and ads
- View connected Pages
Claude almost always calls these tools first when starting a new conversation, just like an advertiser sitting at their desk and opening Ads Manager to take a look.
Tracking & Data Quality (4 tools)
This set of tools solves a pain point that agencies and brands have argued about for years:
- Pixel and CAPI info: One-click view tracking settings
- Match quality score: Understand event data match rates
- Event stats and deduplication: Confirm there’s no double counting
- Recent tracking errors: Find which events are broken
The industry’s reaction was “we’ve been asking Meta for this access for years.” In the past, diagnosing tracking issues might require submitting a support ticket and waiting for a response. Now you can check directly in Claude.
Performance Insights & Benchmarking (7 tools)
This is the most valuable intelligence category:
- Industry benchmarks: Where your CPM and CTR rank in your industry
- KPI anomaly detection: Automatically spot metrics that deviate from normal
- Bid ranking: Compare performance with other advertisers
- Historical trend analysis: Track metric changes over time
- Opportunity score: Meta’s proprietary metric showing how much room for improvement remains
Real Test: 90 Minutes of Work Down to Under 1 Minute
Based on current real-world tests, the most direct impact is on daily report reviews.
An advertiser managing multiple accounts used to spend 90 to 120 minutes every morning: opening each account, checking ROAS, examining frequency, identifying anomalies, writing summary reports. Now with MCP, Claude can complete the same work in under a minute.
A fitness brand case is even more compelling: after automating creative monitoring and variant launch flow with MCP, monthly revenue increased by 15%. Not because AI’s strategy is better than humans, but because AI executes faster and more intensely—able to test more variants before humans can even react.
Why the Agency’s 20% Model Can’t Hold
Traditional agency pricing is built on a premise: Ad operations are complex, brands can’t do it themselves.
The services agencies sell:
- Account setup, tracking configuration → Now OAuth one-click authorization
- Creative upload, campaign creation → Now done with natural language
- Report reading, weekly summaries → Now Claude finishes in under a minute
- Budget adjustments, optimization → Now AI can execute more frequently
When the barrier to doing these jobs drops to near zero, the justification for charging 20% service fees disappears.
This doesn’t mean agencies will disappear. But if agencies are still selling “I’ll set up your account, upload creatives, read your reports,” it’s like a copy shop trying to charge high prices after laser printers became ubiquitous.
But What AI Can’t Do: Strategic Judgment
What Meta Ads MCP can do is fundamentally execution. It can help you build campaigns, pull reports, compare benchmarks—but it can’t answer the most critical questions:
- Why this audience? — You chose women aged 25-35 because the data told you to, or because you understand your brand’s mental positioning in that demographic?
- Why this message? — Using “limited time” or “exclusive” in your copy is backed by judgment about market timing
- Why this timing? — Increasing spend when a competitor just messed up big, or avoiding during everyone’s fighting for Black Friday?
These judgments require not data (AI beats you at data), but understanding of brands, markets, and human nature. This understanding comes from years of running campaigns, observing your industry, and intuition about consumer behavior.
The co-founder of Markacy put it directly: “Meta’s AI and algorithms will always be at the core of ad optimization.” Tools help with execution, but picking the direction still requires human judgment.
One Level Deeper: It’s Actually Meta’s Lock-in Strategy
eMarketer analyst Jacob Bourne’s observation is spot on: On the surface, Meta opening MCP appears to embrace an open ecosystem, but it’s actually a “clever lock-in strategy.”
The logic goes:
- Let you use Claude to manage Meta ads → Your workflow becomes more dependent on Meta
- Give you better tools → You spend more budget on Meta
- More AI tools integrate → Meta’s data moat gets deeper
Meta won’t open what truly makes a decisive difference (like core algorithm logic). What it opens are tools that make it easier for you to spend money on Meta.
Impact on Different Roles
If you’re a brand advertiser
Good news: You can reduce reliance on agencies, at least on the execution side. Bad news: You need to handle (or hire someone for) strategic judgment yourself.
If you’re an agency
Bad news: The execution-based pricing model won’t last long. Good news: If you actually have strategic capability, your value will stand out instead of being buried under execution-level grunt work.
If you’re an experienced advertiser
You’re the biggest winner. You understand strategy and know how to use tools. Now AI amplifies your execution capability 10x. Build an ad placement skill, let Claude execute—all you need to do is take a final look to confirm the direction.
If you’re a new advertiser
The barrier has actually gotten higher. In the past, you could learn operations first, then strategy. Now AI has taken over operations—you need to build strategic thinking faster to have value.
How to Get Started
Using Meta Ads MCP doesn’t require applying for a Developer App, waiting for review, or manually managing tokens:
- Prepare your AI tool: Claude Pro/Max or ChatGPT Plus subscription
- Authorize account: Standard Meta Business OAuth flow (as simple as authorizing Shopify)
- Start operating: Manage ads directly with natural language in Claude conversations
- Terminal users: Developers using Claude Code can install Meta CLI and use commands like
meta campaigns list,meta insights run
Currently free during Open Beta. Official pricing hasn’t been announced yet.
History Always Repeats Itself
New tools appear → Some people transition → Some people get eliminated.
When Google Ads appeared, traditional media buyers got eliminated. When Facebook Ads Manager appeared, agencies that didn’t go digital got eliminated. Now that AI can directly operate ad accounts, pure execution agencies and advertisers could be next.
But at every wave of elimination, some people become stronger by embracing tools. The difference isn’t the tool itself—whether you see it as an amplifier or a replacement.
Want to learn more about how AI tools are changing how various industries work? Subscribe to JudyAI Lab Newsletter for the latest articles.