The official Google Ads MCP Server is no longer limited to reading advertising performance data. Google has added tools that let AI agents create ad assets, upload videos to YouTube, and manage labels in advertising accounts. The changes were merged into the main GitHub branch on November 2, 2026, and the project’s configuration already lists version 0.0.6.
This marks a shift from an AI assistant for analytics to a tool capable of carrying out specific operations in Google Ads. However, the new tools do not provide access to bidding, budgets, or campaign launches. Each change requires separate user confirmation.
Key Takeaways
- On November 2, 2026, Google merged code for creating ad assets and managing labels into the MCP server’s main branch.
- The new tools support text, image and video assets, sitelinks, callouts, and structured snippets.
- An AI agent can upload local video files to YouTube through the Google Ads API.
- Labels can be created, edited, deleted, and associated with supported advertising resources.
- Write operations are protected by a two-step Propose & Confirm mechanism.
- Write tools are enabled in the default configuration, but they can be disabled.
What Actions Are Now Available Through Google Ads MCP?
Google Ads MCP Server uses the Model Context Protocol to connect AI agents to the Google Ads API. Previously, its primary use case was reporting: an agent could retrieve campaign data through Google Ads Query Language (GAQL), inspect resource metadata, and identify accessible advertising accounts.
The main branch documentation now describes four additional tools that can modify data.
Creating Ad Assets
The create_asset tool lets users create Google Ads assets: text, images, YouTube video references, sitelinks, callouts, and structured snippets. Additional parameters are available for other asset types.
An important limitation: creating an asset does not mean it will automatically be used in an ad. The new tools are not presented as a general-purpose way to attach assets to campaigns or ad groups.
Uploading Videos to YouTube
The upload_video_asset tool allows a local video file to be uploaded to YouTube through the Google Ads API for subsequent use as a video asset. The upload runs as a background task.
According to Google’s documentation, this capability is supported only in local stdio mode. With remote deployment over HTTP, uploading a local video file this way is not available.
Managing and Assigning Labels
The manage_labels tool handles the creation, updating, and deletion of account labels. A separate tool, manage_label_associations, can add and remove label associations with CAMPAIGN, CUSTOMER, AD_GROUP, and AD_GROUP_CRITERION resources.
For advertising teams, this could make large accounts easier to organize, for example by tagging campaigns according to market, test type, or the responsible media buyer. These operations do not, however, change bids, budgets, or targeting.
How Propose & Confirm Protection Works
Google has implemented a two-step process for all four write tools. First, the agent prepares a proposed operation, and the server checks it without applying any changes. Where supported by the Google Ads API, the validate_only mode is used.
The server then returns a description of the proposed changes and a checksum tied to the operation’s parameters and the advertising account ID.
After explicit user approval, the agent can resubmit the operation with the same parameters and the received checksum. If the parameters do not match or the checksum is missing, the server rejects the request to modify data.
This mechanism reduces the risk of accidental write operations but does not replace human review. Users need to evaluate the proposed operation before confirming it, especially when an agent works with several client accounts.
What Google Ads MCP Still Cannot Do
The introduction of write operations does not mean that an AI agent has gained full control over the advertising account. The current tool list does not include dedicated functions for changing budgets, adjusting bids, adding keywords, pausing campaigns, or launching new campaigns.
As a result, a request such as “find campaigns with high CPA and reduce their budgets” is not yet supported by this MCP server’s standard toolset. The agent can analyze performance through GAQL but has no documented tool for making that budget change.
This distinguishes Google Ads MCP from the advertising platform’s built-in AI features. For example, UAGEEK previously covered AI-generated text previews in Google Ads AI Max. In that case, Google generates ad content inside the platform’s interface, whereas MCP gives an external AI client access to specific operations through an API.
How to Disable AI Access to Changes
All available tool categories are enabled in the default Google Ads MCP configuration file. Administrators should therefore review the settings before connecting an agent to a production advertising account.
For read-only operation, Google recommends disabling the labels and assets categories in the tools_config.yaml file. The relevant configuration snippet looks like this:
namespaces:
labels: false
assets: false
The configuration path can be supplied through the GOOGLE_ADS_MCP_TOOLS_CONFIG environment variable. The server can also use a file in the current working directory or the package’s default configuration.
Google also warns that connecting an AI client to MCP gives it access to advertising account data within the permissions granted. Agencies and teams therefore need to control OAuth access, the account list, and the environment in which the agent operates.
Version 0.0.6: What to Know Before Updating
The pyproject.toml file in the repository’s main branch lists version 0.0.6. However, at the time of checking, the GitHub Releases page still showed v0.0.5 as the latest published release.
This means that a feature’s presence in the main branch does not guarantee it is available in an already installed package. Before using it, check the specific version, installation source, and actual list of tools exposed by the MCP server.
The project also remains experimental. It makes sense to test new capabilities on a test or noncritical account first, rather than immediately connecting them to a large number of client campaigns.
What This Changes for Agencies and Media Buyers
The most immediate practical benefit of Google Ads MCP is combining reporting and preparatory tasks in a single AI workflow. For example, an agent could analyze an account’s structure, propose a labeling system, create those labels after approval, and help prepare new ad assets.
For teams that regularly produce large numbers of creatives, this could reduce manual work. However, automation still does not cover the full campaign-launch cycle: attaching assets, changing budgets, and managing ad delivery require other tools.
Google continues to expand AI capabilities across its advertising ecosystem. UAGEEK has also covered AI tools for creating and localizing ad creatives. The new MCP functions complement that direction, but they do not amount to the arrival of a fully autonomous media buyer.
The key takeaway for advertisers is that AI can now do more than explain Google Ads metrics: it can also perform limited account operations. Access permissions, review of proposed changes, and version control are therefore becoming part of everyday advertising automation.
Sources for Verification
Sources for Verification
Google Ads MCP — official repository. A list of available tools, limitations, the Propose & Confirm mechanism, access settings, and deployment modes.
Google Ads MCP Server Documentation
Google Ads MCP — Pull Request #139. Confirms the addition of asset creation, video uploading, and label management features, and the merge into the main branch on November 2, 2026.
Code Changes and Merge History
Google Ads MCP — pyproject.toml. Confirms version 0.0.6 as specified in the main branch and the Python requirements.
Project Version Configuration
Google Ads MCP — GitHub Releases. Shows which versions have been published as separate releases.
Google Ads MCP Release History
