← All posts

June 30, 2026 · 5 min read

How to Run Ads from Claude in 2026: The Full Research-to-Launch Loop

How to run ads from Claude in 2026: connect one MCP server, research competitors' winners, clone the best ad, launch it live, and manage budgets by chat.

You can run ads from Claude now. Not "draft ad copy in Claude and paste it into Ads Manager" — actually run them: research what's working in your category, clone the best competitor ad for your brand, launch it as a live campaign on Google or TikTok, and then pause it, resume it, or move budget, all from the same chat window. If you've been searching for a "Claude ads manager," this is what that looks like in practice, and this post is the full walkthrough.

The setup is one MCP connector. The workflow is three verbs: research, clone, launch.

Connect AdWhispr to Claude

Claude.ai supports remote MCP servers through connectors. Add AdWhispr and you get 27 tools spanning research, creative, execution, and management:

  1. In Claude.ai, open your connector settings and add a custom connector.
  2. Point it at the AdWhispr endpoint — the current URL and walkthrough live at adwhispr.com/integrations.
  3. Authenticate via OAuth with your AdWhispr account. The free tier (5 MCP tool calls/month, 1 brand) is enough to verify everything works.
  4. To launch anything, connect an ad account: just tell Claude "connect my Google Ads account" and the connect_ad_account tool walks you through OAuth. list_ad_accounts confirms what's linked.

Prefer the terminal? The same server runs in Claude Code — claude mcp add adwhispr -- npx adwhispr-mcp-server — and the Claude Code setup post covers it step by step. ChatGPT user? There's a ChatGPT connector guide too. The loop below is identical in every client.

Step 1 — Research: find the proven winner

Don't start with a blank creative brief. Start with what's already converting in your category:

Who has the best ads in my space? I sell cold brew concentrate.

Claude saves your brand context (save_my_brand), names real competitors, and pulls each one's ads sorted by longevity (search_brandsget_brand_ads). Longevity is the tell: at the time of writing AdWhispr tracks 380+ brands and ~194,000 ads, and 74,000+ of those ads have run 100+ days. Nobody pays to run a loser for 100 days. The longest-running ad in the library has been live for 2,587 days — here's why days-running beats fabricated metrics.

Drill in:

Show me [competitor]'s longest-running active ads and break down
the hook, format, tone, and offer on the top three.

Every ad is AI-classified on those four axes, so Claude answers with structure, not vibes. For TikTok-native research there's research_tiktok_ads; for search intent there's research_keywords and research_competitor_keywords — covered in depth in competitor keyword research with an AI assistant.

One thing Claude will not do here is quote a competitor's CTR or ROAS. Meta's Ad Library doesn't expose those numbers, so AdWhispr doesn't print them. Any tool that does is making them up.

Step 2 — Clone: make the winner yours

Found the ad that's been running 200+ days? Clone it:

Clone that ad for my brand.

clone_ad handles image and video ads; clone_tiktok_ad handles TikTok creative. The output is original creative grounded in the proven winner — your brand, your voice, the structure that's already been market-tested. Clones land in your creative library, retrievable any time with list_my_creatives. If you want the ethics and mechanics of this, read how to clone a competitor's Meta ad.

Pro includes 10 clones/month; Agency includes 50; one-time packs (10/$19, 30/$39, 100/$99) cover heavy testing sprints.

Step 3 — Launch: put it live from the chat

This is the part that didn't exist a year ago. With an ad account connected:

Launch a Google Search campaign for my cold brew concentrate,
$40/day, targeting the keywords we found from [competitor].

launch_search_campaign builds and launches it. Performance Max works the same way via launch_pmax_campaign — both Google formats are generally available. TikTok is generally available too, with a full video-ad builder behind launch_tiktok_campaign:

Take the clone we just made and launch it on TikTok at $50/day.

That's launch_cloned_ad — the shortest path from someone else's proven winner to your live campaign. Meta execution exists in beta. (And yes, the whole loop compresses into a single message — we wrote up the one-prompt research-to-launch workflow separately.)

Step 4 — Manage: campaign ops as conversation

Once campaigns are live, Claude becomes the ads manager you talk to instead of the one you click through:

You say Tool that fires
"How are my campaigns doing this week?" get_account_performance
"List everything currently running." list_campaigns
"Pause the TikTok campaign." pause_campaign
"Resume it Monday morning — actually, resume it now." resume_campaign
"Move the Search campaign to $60/day." update_budget

The compounding win is context. The assistant that pauses your underperformer is the same assistant that knows why you launched it, which competitor ad it was cloned from, and what the category looked like when you made the call. When performance dips, "what should I run instead?" is answerable in the same thread — go back to step 1 without switching tools.

What this replaces (and what it doesn't)

Running ads from Claude replaces the tab carousel: ad spy tool → screenshot → brief doc → creative tool → Ads Manager → spreadsheet. One chat, one context, one loop.

It doesn't replace judgment. Claude won't stop you from launching a bad idea with a big budget — it executes what you ask. The research layer exists precisely so that what you ask for is grounded in ads that have already survived the market. Execution-only chat tools skip that layer entirely; that's the structural difference we unpacked in AdWhispr vs Adspirer. Adspirer manages the campaigns you already have. AdWhispr tells you what to run — then runs it.

Pricing to get started

Free gets you 5 MCP tool calls a month and 1 tracked brand — enough to connect Claude and run one honest research pass. Pro is $39/mo (3-day free trial) with unlimited research tool calls, 3 brands, and 10 clones/month. Agency is $149/mo for 10+ brands, 50 clones, and cross-brand comparison. Flat pricing — research calls aren't metered, so you never hesitate before asking a question.

Connect the server at adwhispr.com/integrations, or start free at adwhispr.com. Then open Claude and type the only prompt that matters: "Who has the best ads in my space?"

The ads manager of 2026 isn't a dashboard. It's a conversation that ends with a live campaign.