Official Google Ads MCP server vs a hosted connector
Google publishes an open-source MCP server for the Google Ads API. It is a solid way to query your account from an AI client if you are comfortable running it yourself. A hosted connector like Ads Connector takes care of setup and adds tools that change the account. Here is how they compare, based on the official repository as of October 2026.
What each option is
The official server lives at github.com/googleads/google-ads-mcp under the Apache 2.0 license. According to its README, it exposes three tools: search, which retrieves information about the account with GAQL queries; get_resource_metadata, which describes the fields of a resource such as campaign; and list_accessible_customers. It also offers resources with the API discovery document, metrics, segments and release notes. None of the listed tools change the account, so in practice it is a reporting and analysis server.
Ads Connector is a hosted remote MCP server at https://ads-connector.com/mcp. It works with Claude, Claude Code, Codex, ChatGPT (developer mode), Cursor and Gemini CLI. Besides reading reports, it can change budgets, bids, keywords, negatives, RSA ads, sitelinks and callouts, create Search campaigns, manage conversion goals and use Keyword Planner. The same connection covers Meta Ads, ChatGPT Ads, GA4, Search Console and Tag Manager.
Side by side
The official server is free software you operate. The hosted connector is a service you sign into.
- Setup (official): install pipx and Python, enable the Google Ads API in a Google Cloud project, configure credentials (Application Default Credentials, a client library config file or an OAuth proxy) and, for production accounts, a developer token with at least Explorer access. Then add it to your MCP client.
- Setup (hosted): sign in with Google, connect Google Ads with OAuth and paste https://ads-connector.com/mcp into your AI client. No Cloud project or developer token of your own.
- Write actions: the official README lists only read tools. Ads Connector supports the common write tasks listed above.
- Platforms: the official server covers Google Ads. Ads Connector adds Meta Ads, ChatGPT Ads, GA4, Search Console, Tag Manager and a knowledge base.
- Safety: with the official server, data stays between your machine or server, Google and your AI client, and since it only reads, there is nothing to undo. Ads Connector creates new campaigns paused, previews and confirms risky changes, logs every change, encrypts credentials with AES-256-GCM and has read-only mode per platform.
- Cost: the official server is free to use; you pay for any hosting you run it on. Ads Connector is $50 per month for up to 20,000 tool calls after a 14-day trial with no card.
- Clients: the README shows configuration for Claude Code, Cursor, VS Code with Copilot and other MCP clients. A hosted URL also works in clients that only accept remote connectors, such as ChatGPT developer mode and Claude on the web.
When the official server is the better choice
Choose Google's server if you are a developer or analyst who wants to self-host, keep everything inside your own Google Cloud project and only needs read access. It is free, open source and maintained by the Google Ads API team, and GAQL access gives you a lot of reporting flexibility. It is also a good base if you plan to build your own tooling on top.
When a hosted connector is the better choice
Choose a hosted connector if you want the AI to act on the account, not just report on it: add negatives from a search term review, shift budget, pause a weak keyword, write new RSA headlines or draft a Search campaign. It also saves the setup steps, which matters for marketers who do not want to manage a Cloud project, credentials and a developer token application.
For agencies, Ads Connector adds read-only sharing limited to specific ad accounts, so a client or teammate can ask questions about their account without write access.
How to try Ads Connector
Sign in at ads-connector.com with Google, start the 14-day trial, connect Google Ads and add https://ads-connector.com/mcp to your client. You can run it next to the official server and compare answers to the same question. If you want to keep things read-only at first, turn on read-only mode for Google Ads and switch it off once you trust the workflow.
FAQ
Does Google have an official Google Ads MCP server?
Yes. Google publishes an open-source server at github.com/googleads/google-ads-mcp under the Apache 2.0 license.
Can the official Google Ads MCP server change campaigns?
Its README lists three tools, search, get_resource_metadata and list_accessible_customers, and none of them modify the account. It is suited to reporting and analysis.
Do I need a developer token?
For the official server, the README says a developer token with at least Explorer access is needed to query production accounts in setups that require one. With Ads Connector you connect through OAuth and do not need your own token.
Does the official server work with ChatGPT?
It runs locally over stdio by default and can run as an HTTP service with an OAuth proxy. Clients that only accept hosted remote connectors need it deployed at a public URL. Ads Connector is already hosted.
Is the official server free?
The software is free and open source. You cover your own hosting and setup time. Ads Connector costs $50 per month after a 14-day trial.
Can I keep Ads Connector read-only?
Yes. You can set read-only mode per platform, and turn on write access later when you are ready.
Connect your ad accounts
Sign in with Google, connect your accounts and paste one URL into your AI client.