Find SaaS ideas already validated by ad spend
If an app runs hundreds of ads for months, the market has already voted. Ask Claude to find them.
- Real ads from Meta’s Ad Library
- Ranked by ad longevity
- Claude, ChatGPT & any MCP client
Real counts from our July 2026 sweep of the same public Ad Library the MCP tools search live — 1,540 software advertisers catalogued, 90 low-difficulty web apps keeping 100+ ads alive (top outlier: a creator-community platform at 3,232). Run your own sweep for today's numbers.
The method
Nobody keeps paying for ads that don't pay back. An ad library that stays full for months is a P&L you can read from the outside — here's how to read it.
Step 1 · Scan
Sweep the Ad Library for software
Claude runs dozens of keyword variations through Meta's Ad Library via RivlWatch's own Meta edge — “screen recorder”, “invoice app”, “AI headshots”, problem phrasings, translations — and dedupes every software advertiser it surfaces.
Step 2 · Rank
Sort by ads live × days running
Active-ad count says how hard they're pushing; how long those ads have stayed alivesays whether it's working. Nobody funds a losing campaign for six months — longevity is the signal that's hardest to fake.
Step 3 · Study
Read what they sell — and the pitch
Pull the exact creatives: the hook, the promise, the pricing angle, the landing page. The market already told them what converts — you get the answer sheet before writing a line of code.
One question, whole market
This is what the hunt looks like once the MCP is connected. The rows below are real counts from our July 2026 sweep — category names only, advertisers withheld.
The sweep catalogued 1,540 software advertisers; 90 low-difficulty web apps keep 100+ ads live right now. The standouts:
- A meal-planning service — 513 active ads
- An AI career-guidance app — 486 active ads
- A landing-page builder — 457 active ads
- An AI resume builder — 368 active ads
Every one of these has kept a large ad library alive long enough that the spend is almost certainly paying for itself. Want me to pull the longest-running creatives from the top one and break down the pitch — hook, promise, price angle?
Stylized replay of a real session — counts as returned on July 31, 2026. Connect the MCP and run the sweep live on today's Ad Library instead.
The same exchange, as it ran — one question, whole market.
Run a deep sweep of the Meta Ads Library for software/SaaS advertisers in: [YOUR CATEGORY, e.g. "video tools" or "bookkeeping"]. 1. Generate 20-30 keyword variations: the product nouns, the problems they solve (what a customer would type), adjacent job-to-be-done phrasings, and German, French, Italian and Spanish translations. 2. Call search_ads_library for every variation and dedupe the advertisers into one list. 3. Rank by matching-ad count and daysRunning. Treat anyone keeping 100+ ads alive for months as market-validated — that's paid proof people buy. 4. For the top 5, use get_ad_creative on their longest-running ads: what exactly do they sell, at what price angle, and what hook does the copy lead with? 5. Finish with: a validated-ideas table (category, ads live, oldest ad's days running, core promise), the hooks that repeat across winners, and the underserved angle nobody in the category is running yet.
Why builders start here
The graveyard of side projects is full of apps nobody asked for. This flips the order: find the demand first, then build.
Validate before you build
Months of someone else’s ad spend is the cheapest market research you’ll ever run. Instead of guessing whether people pay for a tool like yours, find the proof: a category where advertisers keep hundreds of ads alive is a category with revenue in it.
See the exact pitch that converts
Their longest-running ads ARE their best-performing copy — the hook, the framing, the offer that survived every test. Read it word for word, and launch with positioning the market has already approved instead of your first guess.
Watch categories heat up
Re-run the same sweep next month and compare: which advertisers doubled their ad count, which niches new players keep entering, which quietly went dark. You see a category warming up while it’s still early enough to matter.
Connect it in under 2 minutes
No integrations to wire, no data to export, and nothing to copy-paste. Add the URL, sign in with the account you already have, and ask.
Step 1
Start your free trial
3 days free. A card is required and nothing is charged until it ends — cancel in one click from your account. This is what your agent reads from.
Step 2
Add the connector
In Claude, ChatGPT or any MCP client, add rivlwatch.com/api/mcp. That's the whole setup — no key, no config file.
Step 3
Sign in and ask
Click Connect, sign in with your RivlWatch account, approve — done. Then ask for your first sweep.
Add as a custom connector, then sign in
https://rivlwatch.com/api/mcpNo API key, no config file, nothing to install. You sign in on rivlwatch.com with the account you already have, and each connection can be revoked on its own.
Want the full tool list and security model? Read the MCP docs →
Frequently asked
Build what the market already buys.
Connect once, and your next idea starts from months of someone else's ad spend — not a blank page.