Meta ads MCP server: manage ads from AI tools, with agent rules
Meta’s ads MCP server lets an AI assistant read and manage a Meta ad account directly: pull reports, create and edit campaigns, ad sets and ads in natural language, manage catalogues and check signal health. It launched in open beta on 29 April 2026 with no developer setup required, gained portfolio-level rules on 16 July that let the person with full control decide what agents may do, and on 6 October 2026 Meta said agent-prepared work now lands as a draft in Ads Manager, with explicit approval before anything publishes.
| Platform | Meta Ads |
|---|---|
| Feature | Meta ads MCP server |
| Announced | Oct 6, 2026 |
| Status | Beta |
| Where available | Open beta since 29 April 2026 for businesses of all sizes across regions. Rules for governing agents added 16 July 2026. Draft handoff into Ads Manager and expanded workflows announced 6 October 2026. Endpoint https://mcp.facebook.com/ads. |
| Who is affected | Agencies and in-house teams that use AI assistants, anyone with full control of a Meta business portfolio, and advertisers whose partners may connect agents to their ad accounts. |
| Action needed | Recommended for portfolio owners. Decide what AI agents may do on your ad accounts and set the rules before anyone connects a tool. Keep publishing approval with a named person. |
| Sources | Meta for Business: Introducing Meta ads AI connectors (updated 16 July 2026) · Meta for Developers: Ads MCP server overview · Meta Newsroom (Latin America): Advertising Week New York 2026 |
What changed
The Model Context Protocol, MCP, is an open standard that lets AI assistants call external tools. Meta’s implementation is a hosted server at mcp.facebook.com/ads. The announcement describes a Meta-authenticated connection from the ad account to the agent, so every answer reflects real campaign data rather than generic advice, and says that for MCP no developer credentials, API setup or coding is needed. Meta’s developer overview lists seven tool categories:
- Comprehensive reporting.
- Ad creation and management for ads, ad sets and campaigns.
- Catalogue creation and management.
- Signals and datasets, including signal health and quality.
- Help and troubleshooting, which searches Meta Business Help Centre articles.
- A/B tests and conversion lift studies.
- Activity logs, mirroring the Ads Manager campaign history page.
The 16 July update added two things. Developers can connect any AI application through their own developer app, for deeper integrations. And ads MCP server rules give anyone with full control of a business portfolio the ability to govern what AI agents can do on the ad account, from budget changes to catalogue updates. Meta also ships an ads command-line interface that authenticates with a system user token and calls the Marketing API.
On 6 October, in its Advertising Week release, Meta said the server now supports more workflows, including uploading creatives, previewing ads as they will appear on Facebook and Instagram, and tracking performance with trends. Agent-prepared work arrives as a draft in Ads Manager, people and agents can build the same campaign together and publish from either surface, every action is governed by advertiser-defined permissions, and explicit approval is required before anything is published.
Why it matters
For agencies this is the most consequential Meta change of the year, and it has had almost no coverage because it is not a feature in Ads Manager. It means the AI assistant your team already uses can become the interface to the ad account, with the same data the UI shows. Reporting is the obvious first use: pulling a cross-account weekly summary without exports. The second is draft creation, where the agent builds and the human reviews, which is exactly the handoff Meta formalised in October.
The risk sits in the permissions. The server can change budgets and edit live ads, and by default an agent connected by someone with the right access can do whatever that person can do. The July rules exist so that the portfolio owner, not the individual who connected a tool, decides the ceiling. Our view: treat the rules as mandatory before the first connection, start every agent in read-only reporting, and only extend to drafts once you have seen what it produces. Do not give an agent publish rights on a client account at all; the October draft-and-approve flow means you do not need to.
What to do
- In Business Manager, confirm who has full control of each business portfolio; only those people can set ads MCP server rules, and they should set them before any tool is connected.
- Define the rules per account: reporting only, reporting plus draft creation, or full edit. Keep budget changes and catalogue updates out of agent scope on client accounts.
- Connect one agent to one low-stakes account first, and run reporting questions you can verify against Ads Manager before trusting any figure in a client deck.
- When you move to campaign creation, have the agent produce drafts only, review them in Ads Manager, and keep publishing approval with a named person.
- Check the activity log tool, or the campaign history page it mirrors, weekly to see every change an agent made and who approved it.
Still unknown
Meta’s setup, authentication and full tool-list pages are behind a developer login, so we have not verified which AI clients are supported, how the rules are configured or whether rules can be scoped per agent. Meta has not published rate limits or regional restrictions for the beta. We have not connected an agent to a live account; the read-only-first approach above is how we would start.
Frequently asked questions
What is the Meta ads MCP server?
A Meta-hosted Model Context Protocol server at mcp.facebook.com/ads that lets any MCP-compatible AI agent read and manage a Meta ad account: reporting, creating and editing campaigns, catalogues, signals, A/B tests and lift studies, help-centre search and activity logs. Meta launched it in open beta on 29 April 2026 as part of Meta ads AI connectors.
Can an AI agent change my Meta ad budgets?
It can if it is allowed to. Since 16 July 2026, anyone with full control of a business portfolio can set ads MCP server rules governing what agents may do on an ad account, which Meta says covers budget changes through to catalogue updates. Meta’s 6 October update says agent-prepared work arrives as a draft in Ads Manager and publishing needs explicit approval.
Do I need developer credentials to use the Meta ads MCP server?
No. Meta says the MCP connector needs no developer credentials, API setup or coding; it uses a Meta-authenticated connection from the ad account to the agent. A separate option added in July lets developers connect their own AI application through their own developer app.
Which AI tools work with the Meta ads MCP server?
Meta’s overview says any MCP-compatible AI agent. Meta also offers an ads command-line interface that authenticates with a system user token. Meta’s setup and tool-list pages are behind a developer login, so we have not verified a client-by-client list.
Official sources
- Meta for Business: Introducing Meta ads AI connectors (updated 16 July 2026) ↗ · Jul 16, 2026
- Meta for Developers: Ads MCP server overview ↗
- Meta Newsroom (Latin America): Advertising Week New York 2026 ↗ · Oct 6, 2026
Related coverage
- Meta Customer Lifecycle Strategy and creative AI: what is live
- Meta Business Agent: AI agent for WhatsApp, Messenger, Instagram
- Meta engage-through attribution: link clicks only, 5-second views
Update history
- — First coverage, prompted by the 6 October expansion. Includes the April open-beta launch and the July rules update as background.
Suggested checks are editorial guidance, not platform requirements. Every claim links to its official source. Report a correction →