How to chat with any brand's Meta ads inside Claude in 90 seconds

Meta's Ad Library is free. In practice, that's a lie.

Try it: open facebook.com/ads/library, search for any DTC brand, and tell me which of their ads has been running the longest. You can't. The library doesn't sort by run time. It doesn't paginate cleanly. It doesn't let you compare formats, group by hook, or pull spend ranges into a table. It loads in a single-page app that breaks every scraping tool ever written. The data is technically public; the analysis is technically impossible.

That's the gap AdWhispr fills. We've ingested the ads for thousands of brands, snapshotted them daily, and exposed it all as an MCP server you connect to Claude in a couple of clicks.

After this post you'll be able to ask Claude things like:

• "Which of CalAI's current Meta ads has been running the longest?"
• "Compare Roman and Hims on Meta — who's iterating faster?"
• "Show me UGC ads from supplement brands that have run for 60+ days, with their hooks."
• "Generate a brief on Liquid Death's ad strategy from the last 90 days."

…and Claude will actually answer, with data, in seconds. No browser tabs. No CSV exports. No manual scrolling. Whether you're a software founder studying how SaaS competitors run paid acquisition or a DTC operator dissecting category leaders, the queries work the same way.

Install in 90 seconds

This works in Claude.ai (web) today. Three steps, no manual tokens.

1. Sign up at adwhispr.com. Free account is fine — 5 messages a month, no credit card. That's enough to walk through this tutorial.

2. Open Claude.ai → Settings → Connectors → Add custom integration.

Paste this URL into the Server URL field:

https://adwhispr.com/api/mcp

Click Connect.

3. Authorize the connection.

Claude.ai will pop open an AdWhispr authorization page. Sign in (or stay signed in if you already are), review the access AdWhispr is requesting (read and write scopes for your brand data), and click Authorize. You'll bounce back to Claude.ai with the connection live.

That's it. No tokens to copy, no config files to edit, no headers to format. The whole exchange is OAuth — Claude.ai handles the credential dance for you.

To confirm the connection, start a new chat in Claude.ai and click the tools icon at the bottom of the message box — you should see "AdWhispr" listed with eight tools (search_brands, get_brand_ads, get_brand_stats, search_ads, add_brand, compare_brands, clone_ad, generate_brief). Click any tool to see its description.

Prefer Claude Desktop? There's a one-command CLI that writes the Claude Desktop config for you:

npx adwhispr-mcp-server config

Restart Claude Desktop and the same eight tools show up. Either path — Claude.ai web or the desktop CLI — connects to the same server.

Your first six queries

These are the queries we'd run on day one if we were you. Copy them into Claude verbatim.

1. Confirm a brand is in our index

What does AdWhispr have on CalAI?

Claude will call the search_brands tool, find the brand, and report back: how many ads we've ingested, when we first saw them, what's currently active. If a brand isn't indexed yet, ask Claude to add it — add_brand handles ingestion in the background and the brand will be queryable in a few minutes.

This is the right first query whether you're researching a SaaS competitor like CalAI, an ecom brand like Liquid Death, or a creator-tool like a no-code builder. Same tool, any brand.

2. Find the longest-running ad

Which of CalAI's currently active Meta ads has been running the longest, and what's the hook?

Days-running is the only honest performance proxy in Meta's data. Brands don't keep losing money on the platform — if an ad has been running 100+ days, it's working. Claude will pull the top performers, show you the actual creative, and surface the hook structure.

For software brands this is gold: SaaS founders rarely share which ads work. Days-running tells you exactly which CalAI hooks have survived A/B testing for months. That's a roadmap.

3. Estimate spend without fabricating it

Estimate Hims's monthly Meta ad spend. Show your work.

Meta gives spend ranges, not exact numbers — and most ad-spy tools will hand you a confident "this brand spends $387K/mo" that they invented. We don't. AdWhispr triangulates the Meta range × engagement signals (scraped via Apify) × days-running × creative count and gives you a tighter estimate with the inputs cited. You can argue with the math; you can't with vibes.

For DTC brands like Hims this matters because the Meta range alone might be "$100K–$1M" — too wide to plan against. Triangulated, the estimate often narrows to a $50K window. That's the difference between guessing and building.

4. Compare two brands side-by-side

Compare Hims and Roman's current Meta ad strategies. Who's testing more new creative? Whose offer changed recently? What's the format mix?

This is the analysis that takes an hour by hand. Claude will pull both brands' active sets, diff them, and tell you which one is iterating faster, which one just pivoted offers, and what each one's format mix looks like. Cross-brand comparison is the buying trigger for agencies — it's why Agency-tier customers exist.

5. Find adoptable patterns in a category

Show me the longest-running UGC testimonial ads from DTC supplement brands in the last 90 days. I want hooks I can adapt.

This is competitor research on hard mode — pulling cross-brand patterns within a category. Claude will surface the top 5–10 winners, show you the hook structure, and (if you're on Pro or Agency) you can immediately ask it to clone one for your brand.

6. Generate a brief

Generate a competitive brief on Liquid Death's last 90 days of Meta ads. Include hook taxonomy, format mix, days-running distribution, and a strategic verdict.

Briefs are the deliverable a media buyer actually pastes into a campaign plan. Claude calls generate_brief, formats the analysis with our template, and (Pro+) you can export it as a downloadable PDF or Markdown. The brief leads with the derived-intelligence panel — days-running distribution, engagement-verified reach, creative iteration rate — before any qualitative analysis. Evidence first. Opinion second.

Why this isn't ChatGPT-with-Meta-API

It's a fair question — Claude and ChatGPT are great research tools. With MCPs available, why pay AdWhispr instead of just handing Claude a Meta API token?

Because the Meta API token doesn't do what people think it does:

• Meta has two APIs. The Marketing API (your own ad account) and the Ad Library API (competitor data). Your token unlocks one of them. Your competitors are in the other.
• The Ad Library API is rate-limited to the point of unusable for production work. It returns paginated SPA pages whose creatives have to be extracted with Puppeteer. It returns spend and impressions only as ranges. It only shows what's running right now — no history.
• Asking Claude to scrape on demand costs $5–$50 per query in tokens (you're shipping creative through the context window for analysis). AdWhispr ingests once, serves at near-zero per query.

So Claude alone can't actually answer "which of Hims's Meta ads has run the longest" — Meta's API doesn't return historical data. AdWhispr can, because we snapshot every brand we track every day. The history is the product.

What AdWhispr is, honestly: the Meta competitor-ad data layer for AI assistants. The cognition is Claude's. The data, the methodology, and the productized outputs are ours.

What you can build from here

Once the MCP is connected, three workflows unlock:

Cloning a competitor's winner. Pro and Agency tiers include the clone_ad tool. Ask Claude to clone a specific competitor ad for your brand — you'll get image variants generated in your brand identity, a video storyboard with shot list and UGC creator brief, full ad copy adapted to your voice, a targeting hypothesis, and a landing page outline. Every output cites the source ad and the methodology used. You take the spec into your own Ads Manager and ship.

Tracking competitor changes automatically. Agency tier includes change alerts — we detect when a tracked brand pivots offer, launches a new format, or kills a long-running ad, and notify you. Stop checking. We watch.

Cross-brand intelligence at the category level. Agency tier unlocks the cross-brand dashboard — pull 10+ brands at once and ask Claude category-level questions ("what is every DTC mattress brand doing right now?"). This is the work that costs $5K/mo at a research agency. With AdWhispr it's a single chat message.

Pricing, plain

A "call" is one chat message in our app or one MCP tool call from Claude. Brand ingestion is always free — you only pay for analysis.

Connect once. Then ask anything.

The whole point of MCP is that the tool you already use becomes more useful. You don't switch apps. You don't open a dashboard. Claude is your strategist; AdWhispr is the eyes for Meta competitor research that Claude doesn't have on its own.

Connect Claude.ai to https://adwhispr.com/api/mcp, or run the desktop CLI:

npx adwhispr-mcp-server config

Then ask the first question.

If you hit anything that breaks or any data that looks wrong, ping us — every signal in our pipeline is auditable, and we'd rather fix a bug than ship a hallucinated metric. That's the whole bet.

AdWhispr is the Meta competitor-ad data layer for AI assistants. We don't fabricate metrics. Every signal we surface cites its inputs. Get started free →