What it is

advntg.ai connects as an MCP server in Claude. Performance analysis and creative operations directly from chat, without manually building reports or switching between dashboards.

Why use it

For tasks that are either not available in the dashboard or require manual work:

• Creative pairing — see which creatives have already been tested together, what worked, and which pairs are still untested

• Cross-app gaps — find winners from one app that have not yet been tested in another

• Creative fatigue and breakouts — see which creatives are gaining or losing spend momentum, not just which ones have the highest spend

• Scaling winners — copy a winning live ad into other ad sets while preserving social proof

• Drill down without rebuilding reports — refine the analysis with follow-up questions instead of setting up filters again

Safety

Publishing follows a draft → review → confirmation flow.

Nothing is sent to the ad network without your explicit confirmation. This is enforced at the server level, not just as an instruction to the agent.

The agent also will not invent information that should come from you, such as the landing page URL, copy, budget, geo, account, or creative selection. If something is missing, it will ask.

How to connect

Server URL:

https://advntg.cloud/mcp

Personal plan (Free, Pro, Max)

1. Settings → Connectors → "+" → Add custom connector
2. Paste the server URL, click Add
3. Click Connect and sign in with your advntg account

Team or Enterprise plan

Your workspace Owner does this once: Organization settings → Connectors → Add → Custom → Web, paste the URL.

Then each person: Settings → Connectors → find advntg → Connect → sign in with your own advntg account.

In a chat

Click "+" at the bottom left → Connectors → toggle advntg on.

Your first message

Once connected, ask:

What can I do with advntg MCP for my apps?

The agent will show the workflows available for your apps and connected networks.

From there, you can get more specific, for example:

What should we test next in [app name]?

or:

Which winning creatives haven't been tested across my other apps?

You can keep refining the same conversation with follow-up questions instead of restating the full context each time.

For the full reference — prompts, limits, edge cases, and troubleshooting — see: advntg MCP — Reference

Feedback

You can send feedback directly from chat:

Send feedback: [your message]

The most useful feedback is the exact wording of requests that did not work as expected.