If you want to connect your ad account to Claude or ChatGPT and have your AI assistant launch real campaigns — not draft copy, not "suggest a strategy," but put live ads into a live ad account — this is the tutorial. Fifteen minutes, one OAuth flow, and your assistant goes from advisor to operator.
This became possible because AdWhispr's MCP server, which started as a competitor-research tool, now ships execution tools: connect an ad account once, and the same chat that researches your competitors can launch and manage your campaigns.
Here's the full setup, then a first launch.
Step 1: Add the AdWhispr MCP server to your assistant
The server works in Claude (via claude.ai connectors), ChatGPT (via connectors / App integrations), Claude Code, and Cursor. The endpoint and per-client instructions are documented at adwhispr.com/integrations; CLI users can run the npm package directly:
npx adwhispr-mcp-server config
If you've already set it up for research (we've covered Claude Code and Cursor in detail), you don't need to redo anything — the execution tools show up in the same server.
Step 2: Connect your ad account with one prompt
In your assistant, say:
"Connect my Google Ads account"
The assistant calls connect_ad_account, which hands you a standard OAuth link. You approve access in your browser — the same consent screen you'd see connecting any third-party tool — and you're done. No API tokens to paste, no developer accounts to create.
Then verify what's connected:
"List my connected ad accounts"
That's list_ad_accounts. It returns every account you've linked, so multi-account operators (and agencies juggling clients) can confirm the assistant is pointed at the right one before anything launches.
Step 3: Know what's generally available vs beta
Be precise about this, because the platforms are not all at the same stage:
| Platform | Status | Launch tools |
|---|---|---|
| Google Search | Generally available | launch_search_campaign |
| Google Performance Max | Generally available | launch_pmax_campaign |
| TikTok | Generally available | launch_tiktok_campaign (full video-ad builder), connect via TikTok Business Center |
| Meta | Beta (gated) | — |
Google Search, Performance Max, and TikTok are the ones you can rely on today. Meta execution exists but is beta and gated — if you need it, ask; don't build your workflow around it yet. (The research side is a different story: Meta Ad Library ingestion, TikTok ad research via research_tiktok_ads, and Google keyword intel via research_keywords and research_competitor_keywords are all fully live.)
Step 4: Your first launch
Let's do a Google Search campaign, because it's the cleanest first run — no creative assets required, just keywords and copy.
Start with research so you're not guessing:
"Research the keywords my competitors are bidding on for [your product category]"
That's research_competitor_keywords. Review the clusters, pick your angle, then:
"Launch a Google Search campaign around the top 10 of those keywords.
Budget $40/day, send traffic to mysite.com/landing-page."
The assistant calls launch_search_campaign with your keywords, budget, and destination. You'll confirm the details in chat before anything goes live — read the summary carefully, exactly like you'd double-check a campaign before hitting publish in the native editor.
Prefer TikTok? The full loop looks like this:
"Find the longest-running TikTok ad from @competitor, clone it for my
brand, and launch it as a TikTok campaign with a $50/day budget"
That chains research_tiktok_ads → clone_tiktok_ad → launch_tiktok_campaign in a single prompt. We walk that clone-and-launch pattern step by step in this tutorial, and the bigger picture of why research-to-launch-in-one-prompt matters is the subject of our flagship post.
For Performance Max:
"Launch a PMax campaign for my store using the creatives in my
library, $60/day"
launch_pmax_campaign pulls from your saved assets (list_my_creatives shows what's there).
Step 5: Manage it from the same chat
Once campaigns are live, the management tools keep you out of the native dashboards for day-to-day operations:
list_campaigns— "What's running right now?"get_account_performance— "How did the account do this week?"update_budget— "Bump the TikTok campaign to $75/day."pause_campaign/resume_campaign— "Pause the search campaign over the weekend, resume Monday."
Performance data here is your own account data via OAuth — real numbers, because it's your account. On the competitor-research side we hold the opposite line: Meta's Ad Library doesn't expose CTR or ROAS, so we never print them, and we think no tool should.
Why connect research and execution in one server?
You could connect an execution-only MCP server instead — Adspirer is the best-known one, with broad multi-platform coverage of accounts you already run. That's a legitimate tool for account operations. But it has no competitor ad library and no creative engine, which means every launch still starts with you supplying the strategy and the creative from somewhere else.
Connecting your ad account to AdWhispr means the "somewhere else" is the same chat: 380+ tracked brands, ~194,000 competitor ads, days-running as the proof signal, cloning built in. Your assistant doesn't just execute your ideas — it can source them, from ads your competitors have already paid to validate.
Pricing and getting started
Free gets you 5 MCP tool calls a month and 1 brand — enough to connect an account and feel the workflow. Pro is $39/mo (3-day free trial) with unlimited research tool calls, 3 brands, and 10 ad clones a month; Agency is $149/mo for 10+ brands and 50 clones. Flat pricing — no per-tool-call metering, so you never hesitate before asking a question.
Head to adwhispr.com/integrations to wire it up, or start free at adwhispr.com. Your first prompt after setup should probably be "connect my ad account" — everything else in this post follows from there.