Google Ads MCP setup guide
Google Ads MCP Setup for Claude, ChatGPT & Codex (2026)
Set up a Google Ads MCP server with Claude, ChatGPT, Codex, Cursor, or Gemini, compare official self-hosting with AdsCove, and verify access safely.
Published and last verified: July 29, 2026 · AdsCove Product & Engineering
Direct answer
You have two practical routes. Google's official server is open source and can run locally or on your cloud, but its documented setup requires Python, a Google Ads developer token, an enabled API, credentials, and MCP client configuration. AdsCove is a hosted alternative: connect your Google Ads account in the browser, connect the AI client to https://adscove.com/api/mcp, then verify with list_connections.
Setup steps
- 1
Choose hosted or self-hosted
Use Google's official project when you want to own the runtime and credentials. Use AdsCove when speed, remote OAuth, cross-platform context, and approval-gated operations matter more.
- 2
Connect the Google Ads account
In AdsCove, complete Google OAuth and choose the customer account. Manager-account access is supported when the connected user can reach the customer.
- 3
Connect the AI client
Point Claude, Codex, Cursor, Gemini, or another remote-MCP client to https://adscove.com/api/mcp and authorize with OAuth or an AdsCove API key.
- 4
Verify the customer
Call list_connections and google_list_accessible_customers. Confirm the customer ID before requesting reports or preparing a campaign.
- 5
Test reporting first
Read campaigns, performance, conversion actions, or search terms for a narrow period before preparing any mutation.
- 6
Approve the exact change
Review campaign type, customer ID, daily budget, bidding, keywords, assets, URLs, and initial status before approving a pending execution.
Official Google Ads MCP prerequisites
Google's official integration guide documents a Python MCP server with a Google Ads developer token, Google Ads API enablement, credentials, and MCP client configuration. It can also be deployed as a remote service with an OAuth proxy.
This is a strong choice for technical teams that want to inspect and operate their own server. The tradeoff is setup and maintenance: your team owns credentials, deployment, updates, and access controls.
- Python and pipx for the documented local setup
- A Google Ads developer token with the required access level
- A Google Cloud project with Google Ads API enabled
- Application Default Credentials, OAuth proxy credentials, or a compatible google-ads.yaml setup
- Customer and manager-account IDs where applicable
Hosted AdsCove connection details
AdsCove operates the remote MCP service and platform integration. Your AI client connects to one endpoint, while Google Ads and optional Meta Ads or Shopify accounts are authorized in the browser.
Server name: adscove
Endpoint: https://adscove.com/api/mcp
Transport: Streamable HTTP
OAuth: supported by compatible remote-MCP clients
API-key fallback: Authorization: Bearer sk_live_your_key
Verification tools: list_connections, google_list_accessible_customersSafe first Google Ads prompts
Make the customer ID and date range explicit. Separate diagnosis from execution so the AI cannot turn a broad optimization request into an unreviewed account change.
- For customer 123-456-7890, show campaign performance for the last seven days. Do not change anything.
- Analyze search terms with spend and no conversions. Return negative-keyword candidates only.
- Research this seed keyword, show monthly volume and bid range, and do not select keywords automatically.
- Prepare one paused Search campaign draft and return the pending execution for review.
Common Google Ads MCP errors
Most setup failures come from credentials, developer-token access, or choosing the wrong customer context. Confirm those three facts before changing the MCP client configuration.
- Customer is accessible only through an MCC: include or select the correct login customer.
- Developer token is limited to test accounts: request the appropriate Google Ads API access level.
- OAuth user lacks customer access: authorize a Google user that can open the account in Google Ads.
- Mutation rejected: check policy, required assets, account eligibility, and the exact API error before retrying.
Frequently asked questions
Is there an official Google Ads MCP server?
Yes. Google publishes an official open-source Google Ads MCP server and an integration guide. The documented server can run locally or as a separately hosted remote service.
Does Google Ads MCP work with Claude?
Yes. Google's configuration format can be added to compatible MCP clients, including Claude Code. AdsCove also provides a hosted remote endpoint that Claude can authorize with OAuth.
Do I need a Google Ads developer token?
You need one when operating the official Google Ads MCP setup described by Google. With AdsCove, AdsCove operates the platform integration, so the merchant authorizes the Google Ads account through the browser instead.
Can Google Ads MCP create campaigns?
Capabilities depend on the server. Google's current official documentation emphasizes account search, customer access, and resource metadata. AdsCove separately exposes paused Search, Performance Max, and Display draft workflows plus approval-gated changes.