Google Ads MCP tools compared: official server, open source and hosted connectors

As of 2026 there are three main ways to connect Google Ads to an assistant: the official MCP server published by Google, open-source servers built by the community, and hosted connectors such as Tecrube. All three carry the name “MCP”, but their capabilities and the division of responsibility differ sharply.

The wrong choice is expensive in two ways: trying to manage an account through a read-only server, or setting up a write-enabled connection when all you wanted was reports.

We use five criteria. Write capability: can the server pause a campaign, change a budget, add negatives? Safety layer: if it writes, is there a preview, an approval and a log? Setup burden: does it need a Google Cloud project, a developer token, a local process? Scope: Google Ads only, or other platforms too? Maintenance: who is responsible for updates and API changes?

Write capability: none on the official server; present on some open-source servers; present in Tecrube. Safety layer: not needed on the official server (read-only); usually absent or left to the assistant's confirmation in open source; preview, approval, budget cap and log in Tecrube. Setup burden: a Cloud project and local process for official and open source; one URL and OAuth for Tecrube.

Scope: Google Ads only on the official server; varies by server in open source; five ad platforms plus Google tools in Tecrube. Maintenance: Google for the official server, you for open source, Tecrube for Tecrube. Price: official and open source are free (excluding infrastructure cost); Tecrube is a subscription after 75 free credits.

Google's server is designed for querying ad data and is read-only by design: reports, performance queries, account structure. It does not pause campaigns, move budgets or change bids. Setup requires your own Google Cloud project, API access and running it locally.

Its strength is that it comes from the source with a complete data model. Its weakness is that it cannot manage anything and the setup is technical. It is the right choice for organisations with a data team that only want to query.

GitHub hosts many Google Ads MCP servers, some of which offer write tools as well. Flexibility is highest: you see the code, change it and run it on your own server. The cost sits there too: developer token, OAuth flow, hosting, API version updates and the safety layer are all yours.

Most community servers have no preview-approval mechanism, or leave it to the assistant's own confirmation; know that difference before enabling write tools. They suit developers who can run their own infrastructure.

Tecrube is a hosted server: one URL, OAuth authorisation, setup in minutes. Read and write tools come together; every write follows preview → approve → apply, single-step budget increases above fifty percent are rejected, and every approval is logged. Scope is not limited to Google Ads: Meta, Microsoft, TikTok and LinkedIn ads plus Google tools such as Search Console and Tag Manager sit in the same connector.

The cost is a subscription: after 75 free credits, Pro is $49 a month, Agency $99 a month and Enterprise from $299. Maintenance and API changes are Tecrube's responsibility. It is built for businesses and agencies that want to manage accounts without building infrastructure.

If you will only read data and a Cloud project is no obstacle, Google's official server. If you want to write and run your own code and build the safety layer yourself, open source. If you want to manage the account from chat, keep several platforms on one connection and make every change approved and logged, Tecrube.

Using several MCPs at once is possible: Google's server for deep data queries, Tecrube for daily management, for instance. The assistant routes each request to the right tool. When granting write access, one rule never changes: do not approve without a preview.