TikTok Ads is coming to Adrails

Google Ads MCP server: official vs hosted, reads and actions

Founder of Adrails

Published · Updated

Google publishes an official Google Ads MCP server, and it is read-only: three tools that run on infrastructure you manage. A hosted server such as Adrails holds the connection for you and adds keyword research, campaign builds and edits, each applied after a review. Compare setup, data access and actions below, then choose the server that fits your workflow.

Side by side

Google's server and Adrails, side by side

One is a program you run against the Google Ads API with your own credentials. The other is a hosted server you sign in to.

Google's server and Adrails, side by side
CriterionGoogle's serverAdrails
Who runs itYou: locally over stdio, or on Cloud Run you deployAdrails, hosted
Sign-inAn OAuth client or a service account you create in Google CloudOAuth in the browser, no Google Cloud project
Developer tokenNot required since 9 September 2026Not required
Tools3: list accounts, search with GAQL, resource metadataReads, Keyword Planner, build, edit, status and budget
ReadsAnything a GAQL query expressesSpend, results, revenue, ROAS, CPA and structure, as of the last sync
Keyword volumesNoKeyword Planner volumes and ideas
WritesNone in the current releasePrepared, then applied in Adrails
Campaign types builtNoneSearch, Display, Demand Gen, Performance Max
Other platformsNoMeta Ads through the same connection
PriceFree, Apache-2.0, plus your Cloud Run bill if hostedFrom $49 a month
Google figures from the Google Ads API MCP documentation, last updated 16 October 2026 and re-read on 21 October 2026, and from the googleads/google-ads-mcp repository. Adrails figures from the MCP tools reference on 21 October 2026.

What a Google Ads MCP server does

The Google Ads API has existed for years. MCP, the Model Context Protocol, is what lets Claude, ChatGPT, Claude Code or Cursor call it from a sentence.

A Google Ads MCP server exposes tools; the client lists them, the assistant picks one from your question, the server calls the API and the answer comes back in the conversation.

Two kinds exist. Google's own server is open source and runs where you put it. A hosted server, such as Adrails, runs at a fixed address and you sign in to it.

The tools, the setup and what happens to a change differ. In short, Google's server reads any field the API exposes and writes nothing, while a hosted server trades open queries for packaged tools and reviewed writes.

Google's official server: three tools and GAQL

Google's server was announced on the Google Ads Developer Blog in October 2025 and is maintained by the Google Ads API team under the Apache-2.0 license. It exposes three tools.

It is read-only in its current release. Google's documentation, last updated 16 October 2026, says this implementation cannot modify bids, pause campaigns or create new assets.

The same page points people who need mutate code to the Google Ads API Developer Assistant, which generates client-library code. That assistant does not add write tools to the MCP server.

What the server reads is wide: any resource and any field the API exposes, including ones no packaged tool covers, such as auction insights.

  • list_accessible_customers returns the account IDs your credentials reach
  • search runs a query in GAQL, the Google Ads Query Language, written by the assistant from your question
  • get_resource_metadata tells the assistant which fields a resource has, so it can write a correct query

What Google's server needs from you

A Google Cloud project, an OAuth client or a service account, and credentials on the machine that runs the server.

Locally it runs over stdio from pipx, for a client such as Claude Code, Cursor or VS Code. For a team, you build the image, deploy it on Cloud Run and add Firestore or Redis so OAuth tokens survive a restart.

Developer tokens, once the gate to the API, are no longer required since 9 September 2026 according to Google's documentation.

Each person who uses it locally repeats the setup, or the team shares the Cloud Run address. Someone owns the deployment, its credentials and its updates.

What a hosted server like Adrails adds

A hosted server moves the setup to a sign-in. You add one address to the client, it registers itself, and a browser page asks which workspace and which permissions.

Nobody creates a Cloud project or keeps a service running. The same address works in Claude, ChatGPT, Claude Code and Codex, for everyone on the team.

It also changes what a tool can do. Because the server holds the connection to Google Ads, it can expose writes and decide how a write happens: straight away, or as a proposal that someone applies.

Adrails exposes 78 tools across Meta Ads and Google Ads. On Google Ads, performance_period reads one account over an exact date range at the account, campaign, ad group or ad level, with an optional comparison period, and campaigns_structure returns how the account is built as of the last sync.

google_keyword_planner returns monthly searches and top-of-page bids for a keyword list, or new ideas from a few words or a page, for the countries and languages you name. google_ads_library and targeting_options look up the account's images, audiences and the real IDs of locations and languages.

google_ads_build prepares a new Search, Display, Demand Gen, Performance Max or Shopping campaign with its ad groups, keywords and ads, or its asset groups. Video campaigns are read and edited, not created.

google_ads_edit changes ad texts, keywords, names, dates, target CPA or ROAS and asset group texts. campaigns_set_status pauses or resumes, and campaigns_change_budget moves a campaign budget by at most 30% in one change.

A new workspace connects Google Ads under Ad accounts, then connects its AI client. The tools reach every Google Ads account of the workspace.

Worked example: a weekly read, then a budget

This example is illustrative. You ask what each campaign spent last week against the week before, then ask to raise the best one from $100 to $125 a day.

On Google's server, the assistant looks up the fields with get_resource_metadata if it needs them and writes a GAQL query for search. It works for any field, and a wrong field name or a wrong date filter returns a wrong figure that looks right.

The budget request stops there: the server is read-only, so the answer is a figure, not a change.

On Adrails, performance_period reads those exact dates at the campaign level, with the week before as its comparison period, and puts store revenue from Shopify or Stripe next to spend when the workspace has it.

The budget is a 25% move, inside the 30% one change allows, so campaigns_change_budget prepares it and returns a link to apply it in Adrails, which reads Google Ads again first. A proposal expires after 24 hours, and an applied change can be undone.

When Google's own server is the better pick

This is the real trade. GAQL answers any question on any field, including the ones nobody packaged, and the assistant writes the query for you. Plain inputs answer the recurring questions in a fixed shape, with the figures the agents use, and do not answer a question nobody wrote a tool for.

Pick Google's server when a developer is at hand and only reads are needed, above all reads no packaged tool covers, such as auction insights.

Pick it when cost decides: it is free under Apache-2.0, plus your Cloud Run bill if you host it, while Adrails is a paid plan.

A hosted server is the better pick for a team that wants Keyword Planner, writes with a review step, Meta in the same connection or no setup at all. Claude Code accepts both at once.

Pick by the job, or run both

Google's server is the open, read-only option: every field, your infrastructure, no writes. A hosted server is the managed one: packaged reads, Keyword Planner and changes you review before they apply.

If the team needs both, add both, and let the question pick the server.

Tool by tool

The same job on each server

The same job on each server
JobGoogle's serverAdrails
List the accounts you can reachlist_accessible_customersaccounts_list
Spend and results over a periodsearch, with a GAQL queryperformance_period
How the account is builtsearch on campaign, ad_group and ad_group_adcampaigns_structure
Field names and resourcesget_resource_metadataNot needed: tools take plain inputs
Monthly searches for a keyword listNonegoogle_keyword_planner
A new Search or Performance Max campaignNonegoogle_ads_build
New ad texts, keywords or a target CPANonegoogle_ads_edit
Pause, resume or move a budgetNonecampaigns_set_status, campaigns_change_budget

Common questions

Is there a Google Ads MCP server?

Yes. Google publishes one. Its documentation, last updated 16 October 2026, calls the current release read-only, with three tools: list_accessible_customers, search with GAQL, and get_resource_metadata. Adrails runs a separate hosted server at https://adrails.ai/api/mcp. It reads the account, looks up Keyword Planner volumes, and prepares builds and edits that an owner applies in Adrails.

Is Google's Google Ads MCP server read-only?

Yes in its current release, per Google's documentation last updated 16 October 2026. It lists accounts, runs GAQL queries and returns resource metadata. It does not create, edit or pause anything. The Google Ads API Developer Assistant can generate client-library code for mutations. It is not a write tool on the MCP server.

Do I need a Google Ads developer token?

Not for Google's server: its documentation says developer tokens were sunset on 9 September 2026 and are no longer required. Not for Adrails either: Google Ads is connected once in the workspace, and the assistant signs in to Adrails.

Can Claude Code use both servers?

Yes. Add Google's server with its stdio command and Adrails with claude mcp add --transport http adrails https://adrails.ai/api/mcp. Claude Code lists the tools of both and picks from your question.

Can I connect Google Ads to Adrails today?

Yes. Connect Google Ads in Adrails under Ad accounts, then connect your AI client to the MCP server.

Further reading

Sources checked

Share this article