Meta Ads MCP sees your ad account, not your competitors

Meta now hosts its own Model Context Protocol server for ads at https://mcp.facebook.com/ads. Connect it to Claude, ChatGPT or another MCP client and an agent can read and change the ad accounts you have access to. That answers one question well: how are my ads doing. It does not answer the question most ad research starts with: what are competitors running.

What the Meta Ads MCP server covers

Meta's developer overview lists permission-scoped tools in two groups. As of October 18, 2026:

• Reading: campaign performance reporting, signal health and quality, an ad account's activity log, existing A/B tests and conversion lift studies, and Help Center search.
• Writing: creating and editing campaigns, ad sets and ads; creating catalogs and adding product data; creating and managing A/B tests and lift studies.

Every tool is framed around an ad account you are permitted to use. The overview lists no tool for the Ad Library, and none for other advertisers' ads.

What it cannot see

• Competitors' ads. Those live in the public Ad Library, a separate product with its own limits. See the Facebook Ad Library guide.
• Ads that have stopped. Even the Ad Library drops commercial ads once they stop running outside the EU, so history has to be recorded while ads are live.
• Other platforms. It is a Meta server. Google, TikTok and the rest need their own connections.

The official Ad Library API does not fill the gap either: outside the EU it returns political and social-issue ads only. What the Meta Ad Library API actually returns covers that boundary in detail.

Giving an agent both sides

A useful ads agent compares your results with what the market is running. That needs two kinds of evidence with different provenance, and they should stay labelled as different:

1. Your own performance, from the platform's reporting for accounts you control. These are measurements.
2. Competitor ads, observed in public libraries: what ran, from when to when, in which countries. These are observations, with coverage that varies by platform and region.

AdLemur MCP exposes both from one project. search_ads, get_ad and list_ad_changes return stored competitor ads with source, date and coverage attached. get_owned_performance reads connected-account reporting, and get_creative_lineage relates reviewed creative to owned results. Owned reads are read-only; the few supported writes are separately gated and capped.

What AdLemur does not do is part of the answer. LinkedIn competitor collection is switched off until a validated public contract exists, and modelled spend is never mixed with observed figures. The coverage map lists every source and its geography.

Is there an official Meta Ads MCP server?

Yes. Meta hosts one at https://mcp.facebook.com/ads as part of its ads AI connectors. It works with MCP clients such as Claude and ChatGPT.

Can the Meta Ads MCP show competitors' ads?

Its documented tools cover reporting and management for ad accounts you can access. Competitor ads are in the public Ad Library, which the overview does not list as a tool.

Can one agent use Meta's server and AdLemur together?

Yes. MCP clients can connect several servers. Meta's server can manage your campaigns while AdLemur supplies competitor evidence and read-only owned performance.