Gemini integration
Meta Ads and Google Ads in Gemini
Gemini can use Adrails through a custom app connected to the remote MCP server. Sign in with OAuth, choose one workspace and the permissions to grant, then use Gemini to read ad performance, investigate a result and prepare a change that is applied from Adrails.
01
What Gemini can do on your ad accounts
Gemini can ask Adrails for the figures behind a paid-media decision: spend, results, revenue, ROAS, CPA, CTR, CPM, reach and frequency for the chosen accounts and period. It can compare periods, inspect campaign structure and read connected business context such as Shopify or Stripe revenue.
When the next action is clear, Gemini can prepare a campaign launch, a pause, a test, an edit or a budget recommendation through Adrails. The prepared proposal has its own review page and does not change a live advertising object until an authorized person applies it in Adrails.
- Read the figures and business context behind a question
- Prepare a specific campaign action with its evidence
- Review and apply platform changes in Adrails, never in chat
02
How to connect Gemini
Gemini connects Adrails as a custom app. Enter the public MCP server URL, then complete the OAuth sign-in in Adrails. The authorization screen asks for the workspace and permissions only, and does not expose provider credentials to Gemini.
After the connection is enabled, select Adrails from Gemini with @ before asking it to work with your accounts. Gemini's tool confirmation covers only the tool call: an advertising change is applied from its page in Adrails.
- 1. In Gemini, open Settings, then Connected Apps
- 2. Under Custom apps, choose Add a custom app and enter the Adrails MCP server URL
- 3. Complete Adrails sign-in, select the workspace and choose the permissions
- 4. In a Gemini conversation, type @ and select Adrails before asking a question
What runs where
Gemini, Adrails and the ad platform, step by step
| Step | In Gemini | In Adrails |
|---|---|---|
| A question about results | Gemini picks the tool and reasons over the figures | Reads the workspace's accounts and returns the table; nothing changes on the platform |
| A new launch | launch_brief asks what is missing, then campaigns_launch or google_ads_build prepares it | The proposal waits on the Review advertising change page; applied, it is built paused |
| A pause or a budget move | One exact change, prepared in the conversation | An owner or admin applies it after Adrails reads the platform again; undo while it still matches |
| The permission prompt | Gemini's own setting for tool calls | The consent screen: one workspace and the permissions you grant |
| The record | The conversation | Every tool call logged, every change kept with its evidence |
| The cost | Your Gemini plan does the reasoning | 0 credits per tool call; paid plans from $49 a month to apply changes |
From the Adrails MCP documentation and Gemini's own documentation on 4 October 2026.
Read more
- Connect an AI clientEvery client, step by step, then how to check the connection.
- MCP tools referenceEvery tool, its inputs and who can call it.
- MCP server cardEvery tool with the schema the server registers, as a directory reads it.
- MCP securityWhat a connection can reach, and what Adrails records.
- Best MCP servers for adsThe official connectors, Adrails, Pipeboard, GoMarble, Adspirer, Ryze AI, Windsor.ai and Polar, by job.
Common questions
Does Gemini need my Meta or Google Ads password?
No. Meta Ads and Google Ads are connected once in Adrails. Gemini signs in to Adrails through OAuth and receives only the data and tools your workspace permission allows.
Can Gemini launch an ad campaign by itself?
No. Gemini can prepare a launch proposal through Adrails, but an owner or admin must apply it from the Review advertising change page. The resulting campaign is created paused.
Which accounts can Gemini read?
The Meta Ads and Google Ads accounts in the workspace you selected during the OAuth consent flow.
How do I remove Gemini access to Adrails?
Disable the custom app in Gemini, then revoke its grant from Your connections in Adrails if you no longer want it to access the workspace. Revoking the grant takes effect for future tool calls.