AI agents in your Meta ad account: how to set them up safely

In April 2026 Meta opened its advertising system to outside AI assistants. Through what it calls ads AI connectors, an AI agent can read an ad account's results, build and edit campaigns, repair catalogue feeds and set up tests, all from instructions in plain language. The connectors are in open beta, and in July Meta gave business admins controls over them. Whether your team runs ads itself or an agency does it for you, this is a genuinely useful development, and one to set rules for before anyone switches it on.

Two connectors, two kinds of user

The ads MCP server is built on the Model Context Protocol, a standard way for AI assistants to work with outside tools. Meta's developer documentation shows how to set it up with Claude Code and ChatGPT, and says it works with any assistant that supports the protocol. The Ads CLI is a command line tool for developers and automated scripts; it runs on Python and authenticates with a system user token.

Signing in to the MCP server uses Facebook Login for Business and needs no developer setup, which means the agent acts with the permissions of whoever connected it. Since July 2026, any AI app can also connect through a business's own developer app. Meta's documentation lists what a connection can be allowed to do, from reading and managing ads to managing catalogues and business assets.

What an agent can be asked to do

• Reporting: pull results for any campaign, ad set or ad, and explain them.
• Building and managing ads: create and edit campaigns, ad sets and ads, manage custom audiences (deleting them included) and boost Instagram posts.
• Catalogues: build them and fix problems in product feeds.
• Signals: check the health of pixel and Conversions API data.
• Tests: set up A/B and lift tests.
• Help and activity logs: troubleshoot, and see what has changed.

Meta says partnership ads are coming to the MCP connector over the rest of 2026, so agents will be able to create them and manage creators' permissions too.

The safeguards Meta has built in

The most important default is simple: everything an agent creates starts paused, and the assistant asks for confirmation before switching anything on. An agent can draft a complete campaign, but a person has to approve it going live. In July 2026 Meta added ads MCP server rules, which let anyone with full control of a business portfolio decide what AI agents may do in its ad accounts, from changing budgets to updating catalogues.

Where it can go wrong

• Learning resets. Frequent changes to budgets, audiences or creative can push ad sets back into the learning phase. An agent that never stops adjusting can leave a campaign worse off.
• Too much access. The agent can do whatever the person who connected it can do, so connect through an account with only the access the task needs.
• Lost audiences. Custom audiences can be deleted through the connector. A carefully built audience is quick to lose and slow to rebuild.
• Account health. Digiday reported anecdotal claims of accounts being restricted after heavy automated use. No pattern has been confirmed, but it is a reason to start slowly.
• Your data. Reports and customer lists pass through whichever AI service is connected. If an agency runs your account, ask which service it uses and whether that sits within what your own agreements allow.

A sensible way to begin

• Start with reporting only. Let the agent read and summarise results before it is allowed to change anything.
• Use MCP server rules to limit what agents may do in each account, beginning with budget changes.
• Keep the paused-by-default workflow, and make approving an agent's drafts part of someone's daily routine.
• Keep a record of which accounts have an agent connected, who connected it and what it may do.
• If an agency manages your ads, expect to be asked, and to give your agreement in writing, before an AI agent works in your account.

Set up this way, an agent becomes a quick analyst and a patient drafter. It is good at pulling the numbers each morning, spotting a broken feed or preparing a test, and it never spends your money on its own. That is where most of the value sits, and where the risk stays small. However the campaigns get built, keep UTM tags consistent on every link so results can still be traced in your analytics.

Workflow automation and custom AI tools are part of what we do at ALXO. If you are deciding whether to let an AI agent into your ad accounts, and how to fence it in, we would be glad to think it through with you.