Meta Ads CLI vs MCP: What's Different and How to Choose (2026)
Meta Ads CLI and Meta’s ads MCP server are two different ways to work with Meta advertising accounts. The CLI runs commands from a terminal and is designed for scripts and integrations. The MCP server connects a compatible AI client to Meta ads through a remote server, so an agent can retrieve account information or take supported actions. Meta groups both under its ads AI connectors, but its documentation gives them separate setup and authentication paths.
The CLI runs from your terminal; the MCP server connects an AI client
Meta Ads CLI is a command-line interface to the Meta Marketing API. Developers and operations teams can use it to inspect and manage campaigns, ad sets, ads, creatives, catalogs, datasets, and performance insights. Commands follow a resource-and-action pattern such as meta ads campaign list. The CLI also supports structured output for scripts and automation.
Meta’s ads MCP server is a remote Model Context Protocol server at https://mcp.facebook.com/ads. It connects an MCP-compatible AI client to supported Meta advertising tools, including reporting, campaign management, catalog tasks, signal diagnostics, help search, studies, and activity logs. Meta’s announcement presents the CLI and the MCP server together as its ads AI connectors.
| Ads CLI | Ads MCP server | |
|---|---|---|
| Where you use it | Terminal, scripts, or CI/CD workflows | A compatible AI client |
| How it connects | A local meta command calls the Marketing API | An AI client connects to Meta’s remote MCP server |
| Good fit when | You want repeatable commands, integration testing, or shell automation | You want to ask an AI client for account information or supported actions |
| Setup path | Install the CLI and configure a system-user access token | Connect through a supported client, or configure a developer app for a custom integration |
Choose based on where the work should happen. A command-line workflow is a natural fit for a developer-owned script or an integration test. MCP is the relevant path when you want a compatible AI client to use Meta’s supported tools. Do not assume the MCP server is simply a wrapper around the CLI: Meta documents them as distinct connection paths.
Install the CLI with Meta’s documented method
Meta’s CLI setup guide lists macOS and Linux requirements. Its recommended Homebrew installation is:
brew tap facebook/fb
brew trust facebook/fb
brew install facebook/fb/meta-adsMeta also documents a PyPI installation for Python 3.12 or later. After installation, run meta --version to confirm that the command is available.
The CLI authenticates with a Meta system-user access token and an ad account ID. Meta’s setup guide walks through creating the system user, assigning access to the relevant app and ad account, generating a token, and selecting permissions. For catalogs, datasets, or Pages, assign the relevant assets as well. Keep the token out of source control and chat logs. If you store configuration in a local .env file, restrict access to it and do not commit it.
Start with account and campaign reads, for example meta ads adaccount list and meta ads campaign list, before using commands that create, edit, or delete resources. Meta documents campaign and ad set creation as paused by default in the CLI command reference. Check the current command reference before relying on any default or running a write.
See Meta’s CLI setup guide, configuration guide, and command reference for current requirements and options.
Connect the MCP server through an AI client
Meta’s Business Help Center describes a connector setup for supported AI clients that does not require owning a developer app. Follow the setup instructions for the client you use, and connect to Meta’s server at https://mcp.facebook.com/ads. Meta notes that access to tools and features may vary.
Developers who need a custom integration can use Meta’s developer documentation instead. It covers creating or reusing an app, adding the ads MCP server use case, configuring authentication, and requesting permissions. The documented authentication choices include OAuth, a user access token, and a system-user access token. For this MCP server, Meta says only Employee-role system-user tokens are supported; Admin-role system-user tokens cannot access it. This is specific to the MCP server and should not be confused with the CLI’s token setup.
Start with Meta’s MCP server overview, MCP setup guide, and Business Help Center instructions.
Set account controls before enabling agent actions
An AI connection can expose tools that read account data and tools that make changes. Review which actions the connected client can request, who can authorize them, and which ad accounts or catalogs it can access.
Meta says ads created through the MCP server are paused by default until someone sets them live, and actions taken on an advertiser’s behalf require authorization through the AI agent. Where the feature is available, people with full control of a business portfolio can also manage ads MCP server rules in Meta Business Suite. Rules can block actions such as campaign creation or budget changes. Meta’s rules documentation describes the feature as being in limited availability, so it may not be open to every business yet.
For CLI workflows, confirm the account, permissions, and command before execution. The paused creation defaults documented for CLI campaigns and ad sets are useful safeguards, not a substitute for reviewing a command or checking its current behavior.
Keep the decision tied to the task
Use the CLI when the job is best expressed as a repeatable terminal command, a script, or an integration test. Use the MCP server when the job belongs in a supported AI client and its available tools cover the task. In either case, check Meta’s current documentation before granting access or running a write, because supported tools, permissions, and feature availability can change.
Sources
FAQ
- Is Meta Ads CLI the same as Meta's ads MCP server?
- No. The CLI is a terminal interface to the Meta Marketing API. The ads MCP server is a remote server that exposes supported Meta advertising tools to compatible AI clients. Meta documents separate setup paths for them.
- Does the ads MCP server run the CLI?
- Meta's public documentation describes the CLI and MCP server as distinct connectors and does not say that the MCP server runs the CLI. Treat them as separate products unless Meta documents a change.
- Do I need a developer app to use Meta's ads MCP server?
- Not for the no-app setup described in Meta's Business Help Center. A developer app is part of the custom integration route in Meta's developer documentation.
- What authentication does the CLI use?
- Meta's CLI setup guide uses a system-user access token and an ad account ID. The permissions and asset access depend on the work you need to do. Follow Meta's current setup guide rather than reusing MCP token instructions.
- Can an AI agent make live changes to my ad account?
- The MCP server exposes supported management tools, and Meta says actions taken on an advertiser's behalf require authorization through the AI agent. Meta also documents paused defaults for ads created through MCP and account rules that can restrict actions where available. Review the connected client's permissions and Meta's current controls before enabling write access.