For Meta media buyers spending $10K+/month on Meta ads.
Become irreplaceable by turning Claude into your personal junior Meta media buyer that launches ads for you for $17.
Plug Claude into your Meta ad account. Brief it like a junior. Ship 3x the tests in a third of the time without touching Ads Manager UI for the mechanical work. Become the buyer your boss (or your client) can't replace.
Works with Claude (recommended) or ChatGPT. Same workflow runs in Codex, Manus, or Anti-Gravity if your team uses one.
30-day "ship a paused batch or refund" guarantee. No subscription. Already running in 5 live client accounts.
On it. Loaded your saved settings. Running pre-launch safety check…
- ✓ pulled saved account, page, pixel, geo
- ✓ loaded 5 creatives from Drive
- ✓ rate-limit check passed
- ✓ built campaign, ad set, 5 ads
- ✓ all PAUSED · nothing has spent
Total time: 5 minutes. Open Ads Manager to approve.
Two types of media buyer in 2026
There's the buyer who still opens Ads Manager, clicks through 17 fields per ad, and spends 3 to 5 hours building a refresh batch by hand. Then there's the buyer who briefs Claude in plain English, drinks a coffee, and reviews paused campaigns 5 minutes later. Same headcount. Same brain. 2 to 3x the output. One of those buyers is getting promoted this year. The other is the line item the agency cuts when the AI math finally clicks for the founder.
This kit is what makes you the first kind. You download a $17 USD bundle. You paste one prompt into Claude. You run a 5-minute wizard that captures your ad account, page, pixel, geo, age and budget defaults per client. From that moment on, Claude is wired into every account you run and knows exactly how to launch ads safely.
You brief it the way a senior buyer briefs a junior. "Pull the 30 videos from this Drive folder. Launch them across the two coaching offers, paused, $100 a day each, using my saved settings." Claude loads your defaults, runs a 2-gate pre-launch safety check, builds the campaigns, uploads every creative as its own ad, and hands you the campaign IDs. You unpause the winners in Ads Manager.
Under the hood it uses two methods. Meta's official MCP (the free Connector they shipped April 29, 2026) and the Meta Marketing API directly. The kit's launch gate decides which one to use per action, because each one is missing something the other does well. You don't need to know which is which. You stay the strategist. The kit does the clicks.
The product is the wiring. Claude does the work. The kit makes sure the work is safe, paced, and shippable into a real Meta ad account. Setup once, brief forever, stay ahead.
"Brief Claude in English. Paused campaigns in your account in under 15 minutes. You spent the saved hours on the work that gets you promoted."
What a real Tuesday looks like with the kit installed.
"2 launch gates plus 12 hard safety rules block the launches that get juniors fired and accounts restricted. Your name stays on the offer, not the incident report."
What ships in the box.
The seat you're sitting in is being repriced
Be honest about what's already happening. Meta shipped their official MCP on April 29, 2026. Every founder who runs $10k+/month is now one Claude window away from launching their own ads. Three agencies I know cut their junior bench in Q1. The senior buyers who kept their seats all have one thing in common: they're already briefing an AI for the mechanical work and using the saved hours on strategy, creative, and accounts.
The 17-clicks-per-ad workload is the part of the job that gets repriced first. Roughly 60 to 70% of a buyer's day is mechanical setup that has nothing to do with strategy. That's the part the kit eats. The strategy, the creative judgment, the offer thinking, the account whisperer instincts — those are what the founder still pays for. The buyer who keeps doing both wins twice. The buyer who only does the clicks loses.
Here's what the shift sounds like in practice. You send a voice note:
Voice note
"Pull the 30 videos from this Drive folder. Launch them across the two coaching offers, paused, $100 a day each. Use my saved settings for both accounts. Write five primary text variants and five headlines per ad."
You walk to the espresso machine. Answer two Slack messages. Come back. Two paused campaigns are sitting in Ads Manager. 30 ads, each with five primary text variants and five headlines built in. Nothing has spent a dollar. You review and unpause the winners.
That is the picture. One brief. Coffee. Done.
What you do with the 15 to 20 hours a week you just freed up is what decides whether the seat keeps you. Ship 2 to 3x the tests. Take on a second client. Build the angle library the rest of the team uses. Run the audit on the dental account nobody else has time for. The buyers who do that become the ones nobody can cut.
Secure checkout
AI Junior Media Buyer
The drop-in kit that turns Claude into your AI Meta media buyer. Brief it in English. It builds the campaign, configures every setting, and ships every ad paused for your approval.
One-time payment · No subscription · 30-day refund guarantee
- ✓ Built to keep your ad account alive. 2 launch gates, 1.5-second action pacing, 25% budget-ramp guard and 12 hard safety rules block the burst-launch and budget-spike behaviours most often blamed for 2026 account bans
- ✓ Every ad ships PAUSED by default. Nothing spends a dollar until you unpause it in Ads Manager
- ✓ Works inside Claude, ChatGPT, Codex, Manus, or Anti-Gravity. No new app to learn
- ✓ Multi-account ready. One install runs every client account via preferences.json
- ✓ Lifetime access. Yours forever, every future update included
- ✓ One-time $17. No subscription, no per-account fees
- 🎁 Bonus: 12 ready-to-run campaign recipes (cold prospecting, retargeting, conversion, lead form, app install) so your first launch is copy-paste, not a build
Your details
We'll send your kit to this email after payment.
Why this exists
Two things happened in the last 90 days that every working media buyer has heard about, even if they haven't acted on either.
Meta released their official MCP server. April 29, 2026. Free, elegant, and on launch week it had an OAuth bug that broke it inside Claude Code (GitHub issue #55002 is still public). Two weeks in, the verdict from every industry write-up is the same line: "right for reporting, wrong for launching ads." Chris Pollard at Ads Uploader said it cleanest. "MCP is the wrong shape of tool" for launching at scale.
What the official MCP cannot do, as of this week:
- Upload creatives from a local folder or your Google Drive. Every creative must be pre-hosted on a public CDN. That alone kills the "drop the MP4 in chat" workflow you actually want.
- Roll up multiple Business Manager accounts. Each one is scoped separately. If you manage 5+ clients, the MCP alone makes you log in and out all day.
- Save or replay a workflow. When the chat ends, the recipe disappears. No spec file. No audit trail. No undo.
- Approval gates. Anomaly detection. Spend safeguards. None of them exist. One bad prompt can rebudget an entire account and there is no rollback button.
- Disable unused tools. All 29 MCP tools load every session, eating roughly 55,000 tokens before you've even typed a question.
- Touch Advantage+ or the bidding algorithm. Meta deliberately excludes the optimisation layer. You can read, you can create, you cannot tune.
- Work without a paid Claude Pro tier and the OAuth working in your specific Claude Code build.
Fine for reporting. Dangerous for launching. Definitely not the cheat code.
The Marketing API direct is the alternative. It's also, in the words of an actual developer survey, a strong candidate for the worst API in software. The current state of the API for a buyer wiring it solo:
- Standard Access is demo mode. You need Advanced Access for anything real. Approval takes 2 to 3 days, sometimes months. One developer wrote his case as "the Advanced Access Trap that nearly killed my project."
- Tokens expire on Meta's internal fraud signals, not on a clock. Daily, weekly or hourly. Error #190 (Invalid Access Token) accounts for over 70% of authentication issues people hit.
- Rate limits are tier-gated and shared across endpoints. 60% of developers report spending more than 10 hours a week on rate-limit troubleshooting.
- API v24.0 is deprecated June 9, 2026. If your launcher isn't migrated by then, every API call dies on a single Tuesday morning.
- Burst-creating 50 ads in 60 seconds is widely reported as one of the fastest ways to get an AI-driven account flagged in 2026. The rule every agency knows: "no pacing, no pacing, banned."
The official MCP is too shallow to launch from. The Marketing API is too sharp to wire without a safety system. The kit is what sits between you and both of them. The launcher decides which method to use per action. Reads and reporting go through the MCP. State-mutating writes (campaign creation, budget changes, asset uploads, bulk launches) go through the API, paced and gated. You get the conversational convenience of the MCP plus the deterministic power of the API, with the safety wiring most off-the-shelf setups skip. That's the cheat code.
Account safety — because your reputation is the product
68% of active Meta advertisers have hit an account restriction in the last 12 months. That was 41% in 2024. The 2026 climate is enforcement-first, AI-reviewed, and unforgiving. A torched client account is the kind of thing buyers don't come back from professionally. The kit is built to make that almost impossible.
"I've lost months of sleep over it."
Brittany Hamilton, NBC 5 DFW
"I've lost a 15-year business that is essentially gone."
Teshia MacQuarrie
"About 10 seconds later, it came back and said my account was locked... and it came back with horrific news telling me I was banned for life."
Jason Keilman, Canton Township small-business owner
What it actually does to keep your accounts safer:
Pre-launch safety check
Runs before every change to your ad account. Cannot be skipped, even by you.
Pacing
Forces a 1.5-second gap between actions. Burst-creating 50 ads in 60 seconds is the most-reported AI ban trigger in 2026. The launcher paces every action on purpose.
Rate-limit budget plan
Calls plan_rate_budget() against your Meta API tier and refuses to act if the launch would exceed your rolling-hour ceiling. Auto-chunks into safe batches with 65-minute pauses, or refuses with a tier-upgrade recommendation.
Cool-down
When Meta says "slow down," the launcher actually slows down. No retry storms. No silent duplicate ad sets.
Kill switch
Any serious throttle or error arms the kill switch. The launcher refuses every further state-mutating call in that session until you reset it. Stops the 2 AM cascade before it compounds.
Budget-ramp guard
Blocks any budget raise above 25% in one move. 20% is the documented threshold above which Meta's Andromeda algorithm resets your learning phase. The launcher stays under it on purpose so you don't accidentally reset a client's learning.
Multi-account profiles
Saves a separate ad account, page, pixel, geo and budget profile per client. Switch contexts mid-session with one line. One install, every account you manage.
Default state on every API or MCP call is PAUSED. Nothing spends until you unpause. Built to be much harder to misuse than to use correctly. On purpose. Your reputation as a buyer is the product. The kit protects it.
What ships in the box
Everything you need to launch safely across every account you manage. $17 USD one-time. Yours forever.
The hybrid launcher
Hybrid method detection
Detects which method (MCP or Marketing API) is right for the action you're asking for, then runs it through the 2 launch gates before any state-mutating call fires.
PAUSED by default
Every campaign, ad set and ad ships PAUSED. The launcher cannot ship active. You are always the one who unpauses.
Multi-account round-robin
Pacing distributed across the accounts you manage so no single one gets flagged.
Kill switch
Armed on the first serious error. Prevents cascading damage in a client account.
Every Meta campaign type
Cold prospecting, retargeting, conversions, lead forms, app installs, traffic, engagement, Reels, Stories, Advantage+, CBO/ABO, lifetime or daily budgets, custom audiences, lookalikes, cost cap or lowest cost. If Meta supports it, the launcher ships it.
Meta Marketing API v22.0
Built on Meta Marketing API v22.0 with the v24.0 migration path documented.
The autoloaded operating manual
CLAUDE.md
12 launcher-specific hard rules. The model does not drift across sessions because the rules re-anchor every time.
Multi-account preferences
preferences.json holds a profile per client (account, page, pixel, geo, budget). Claude loads the right one per session.
Scope statement
Baked into the launch gate so future Claude sessions cannot scope-creep the kit beyond launching.
Nine reference skills
Full Meta API reference
Complete Marketing API reference loaded on demand.
Insights API
Meta Insights API reference.
2026 rate-limit tier system
Current rate-limit headers and tier system.
Account safety patterns
Behavioural patterns Meta is known to flag, so the launcher avoids them.
2026 Meta doctrine
Andromeda, EMQ, frequency caps, learning-limited threshold, restricted categories.
Launch method (MCP vs API)
Decides which method to use per action. Reads via MCP, state-mutating writes via the API.
12 campaign recipes
Working recipes for cold prospecting, retargeting, conversion, lead form, app install.
Google Drive fetching
Pull creatives directly from Drive folders.
Lead form recipes
Lead form campaign setup recipes.
Setup tooling, run once each
Setup wizard (multi-account)
Run once per client to add a new profile to preferences.json. Claude reads the right profile every session.
Token refresher
Swaps Meta's short-lived token for a permanent System User token in 30 seconds.
Pre-flight checker
Runs the 2 gates before any launch. Cannot be skipped.
Buyer-facing supports
README in 3 formats
Markdown, HTML, PDF. Same instructions, different reading preferences.
Troubleshooting tree
30-second fixes for the 15 most common errors. No support tickets needed.
Worked example
A real briefed campaign so you can see exactly what a session looks like before running your own.
Three layers, one job: turn Claude into the safe Meta launch arm of your media buying practice. Already wired into your accounts the moment setup ends.
The 2 launch gates
Every launch runs through these two gates before any Meta API or MCP call fires. Claude cannot bypass them. Even you cannot bypass them on a bad day. That's the whole point.
Skill-load check
Blocks Claude from launching without first reading meta-rate-limits, account-safety, 2026-meta-doctrine, and meta-api. Stops drift across sessions.
Rate-limit budget (BLOCKING)
Blocks any launch that would exceed the rolling-hour Meta API ceiling. Auto-chunks into safe batches with 65-minute pauses, or refuses with a tier-upgrade recommendation.
The 2–3x productivity math, on a real Tuesday
You drop your finished ads into a Google Drive folder. Or a local folder. Or you point Claude at a public CDN URL. You open Claude. You type a sentence or record a voice note:
You type:
"I have dropped 30 finished videos in this Drive folder. Launch all of it as PAUSED across my two coaching accounts at $100 a day each, using my saved settings."
- Loads your saved defaults from preferences.json (ad account, page, IG, pixel, geo, age, budget, optimisation goal, destination URL).
- Pulls the 30 finished creatives from your Drive folder.
- Calls plan_rate_budget() against your current Meta API tier and shows you the rate-limit plan before mutating anything.
- Runs the 2-gate launch checklist (skill-load + rate-limit budget). Refuses if either fails.
- Decides per action whether to call the MCP or the Marketing API. Reads via MCP. Bulk creative uploads and state-mutating writes via the API, paced.
- Builds the campaign, the ad set, and uploads each creative as its own ad. Everything PAUSED.
- Hands you the campaign IDs and ad IDs so you jump straight to Ads Manager and approve.
What you see in Ads Manager
✓ Campaign · Coaching | May-2026 | PAUSED
✓ Ad Set · Saved audience, $100/day, PAUSED
✓ Ad 1 · pricing-letter.mp4 · PAUSED
✓ Ad 2 · calculator-math.png · PAUSED
✓ ... (28 more)
✓ Done. 2 campaigns, 2 ad sets, 30 ads. Nothing has spent a dollar.
Brief to paused-batch in your account: roughly 10 to 15 minutes for a 30-ad refresh that would take 3 to 5 hours by hand. Do that across 3 clients in a morning. That's where the 2–3x comes from. Not magic. Just every minute of mechanical Ads Manager work, deleted.
What it does NOT do (and why that's the point)
This kit is a safe Meta ads launcher. Nothing more. Strategy, creative judgment, offer thinking, account whispering — those are still your job. That's exactly why the kit makes you more valuable instead of less. It eats the part of the job an AI was always going to eat. It leaves the part of the job that pays you.
Boundary clarity matters more than promises. Here's what the $17 USD base does not do.
It does not generate images or videos. You bring your finished creatives. Canva, Midjourney, NanoBanana, your designer, your editor. The kit launches them. The Image Gen Pro Pack add-on at checkout wires Claude to render images via NanoBanana Pro and gpt-image-2 if you want that.
It does not pick angles, hooks, or strategy from scratch. The base kit assumes you know the offer and the audience. The optional Creative Strategist Skill add-on handles angles and hooks per offer.
It does not manage ads after launch. No auto-pausing of bad performers, no daily budget rebalancing. The optional Ad Management Skill add-on handles that.
It does not send weekly reports. The optional Weekly Audit Skill add-on generates them from your Meta insights.
It does not run TikTok, Google, LinkedIn, or X ads. Meta only. Facebook, Instagram, Threads, Reels.
It does not make video creative from scratch. You bring video assets.
The kit makes no CPL, ROAS, or CTR claims. It ships a working safety system, a working hybrid launcher, and working documentation. Ad results are your strategy, your creative, and your offer. Exactly the things that keep you employed.
Who built this
I built this because I was running it across multiple client accounts at my own ad agency (Low Ticket Ads Agency) and got tired of explaining the setup to every founder who asked. My agency has spent over $2,000,000 on Meta ads using this exact wiring as the launch layer.
The buyers on my team run it daily. It's the reason a single senior buyer can hold down the workload that used to need a buyer plus a junior. It's running right now in 5 live client accounts:
Real accounts, real ads, real spend. The setup you're buying is the same one my own buyers run.
Objections, answered straight
"Won't my boss / client think I'm cheating by using AI?"
They're going to find out you used AI either way. The difference is whether they find out because you shipped 2 to 3x the work and the account performance speaks for itself, or because they tried Claude themselves and realised they could've cut a head. Bring it up first. Show them the paused output. Frame it as a force multiplier you brought to the role. The buyers getting cut in 2026 are the ones who pretended the shift wasn't happening. Be the one who walked into the room with the cheat code already installed.
"Why not just use the free MCP from Meta?"
Because the official MCP cannot upload creatives from your Google Drive, cannot roll up multiple accounts, cannot save or replay a workflow, has no approval gates, no spend safeguards, no undo, and loads all 29 tools every session whether you need them or not. Excellent for reporting. Not built to launch at scale. The kit uses the MCP for what it's good at (reads, reporting) and the Marketing API for what it's good at (paced, deterministic writes with rollback safety). You get both, gated.
"Why not just wire the Marketing API myself?"
You can. Block out 2 to 3 weeks for Advanced Access approval, another week to learn the rate-limit headers, another week to wire your own pacing logic and kill switch, and a permanent maintenance cost for every API version migration Meta forces on you (v24.0 deprecation is June 9, 2026, and they break something every 8 to 12 months). My agency did all of that. The kit is the result. $17 USD buys you the destination, not the journey.
"I'm not technical. Will I be able to set this up?"
Setup is one wizard you click through once. Same difficulty as connecting Stripe to a new tool. After that you only ever talk to Claude in plain English. "Launch these ads, paused." You will never look at code. The README walks you through the 5 to 10 minute setup step by step. If you can paste a token from one website into one text box, you can run this.
"What if Meta deprecates the API or changes the MCP?"
The launcher detects the method per action, so if the MCP shifts the API path keeps you running. For version deprecations (v24.0 is June 9, 2026), you get a video training that walks you through updating the kit yourself line by line. You own the code; you don't wait on me to ship a patch.
"I don't trust AI with my client's ad account."
Good. Don't. That's why everything ships PAUSED, the launcher refuses state-mutating calls during a throttle event, the budget-ramp guard blocks any +25% move, pacing is forced at 1.5s between actions, and a kill switch arms on the first serious error. The kit is built around the assumption that the buyer trusts the AI exactly as far as they can verify in Ads Manager. You stay the human in the loop. You unpause the winners. The AI never spends a dollar without you.
"Will Meta ban my client's account?"
Meta does what Meta wants. Nobody can promise zero restrictions. What the launcher does is remove the behaviours that get flagged most. The 50-ads-in-60-seconds burst is impossible inside the launcher because it forces 1.5-second pacing. The 25% budget-ramp guard keeps you under Meta's Andromeda learning-phase-reset threshold. Everything ships paused. The launcher refuses state-mutating calls during a throttle event. You're a lot less likely to get hit, but no tool (mine or anyone else's) can guarantee Meta's enforcement won't touch you.
"I have never run Meta ads before."
Then this is the wrong starting point and I'll tell you straight. The kit assumes you already know how to read a campaign, judge an audience, and call a winner. If you don't, you need to learn the craft first. The kit makes a working buyer 2 to 3x more productive. It does not make a non-buyer into one.
"Is there a refund?"
Yes. 30 days. Ship a paused batch from a folder of creatives. If it does not ship, full refund, no forms, no calls.
Guarantee
Guarantee
30 days. Ship a paused batch from a folder of creatives. Or your $17 USD back.
If you cannot get the kit set up and ship a paused batch from a folder of finished ads into a Meta account within 30 days of purchase, email me. I refund the $17 USD. No forms. No calls. No upsell.
The guarantee is built around a specific deliverable, on purpose. Not "if you do not love it." Not "if you do not double your salary." The job is to ship a paused batch. That either happens or it does not.
FAQ
Is this the same as buying a course?
No. There are no videos. There is a short README, an operating manual Claude reads at the start of every session (not you), reference skills it reads on demand (not you), three setup scripts you run once each, a worked example, and an empty folder for creatives. You are buying a working tool, not lessons.
Do I need a Claude or ChatGPT subscription?
Yes. Claude Pro at $20 a month works for small batches. Claude Max at $100 a month is recommended if you are running parallel research and 5+ ad batches per session. ChatGPT Plus also works. Neither is bundled. The same workflow runs in Codex, Manus, or Anti-Gravity if your team uses one of those instead.
Do I need a Meta developer app?
Yes. The README walks you through creating one in about 5 minutes. You add the Marketing API product, copy your App ID and App Secret, generate an access token. The refresh_token.py script swaps the short-lived token for a permanent System User token automatically. If you are using the official MCP path, the OAuth handles itself once you are eligible.
Can I run this across multiple client accounts?
Yes. preferences.json stores a profile per client (ad account, page, pixel, geo, budget). Run the setup wizard once per client. Switch contexts mid-session with one line, set_active_account('client_b'), and Claude loads the right profile. Pacing is distributed across accounts so no single one gets flagged. The official MCP cannot do this on its own.
Can I pull creatives from Google Drive, or only local folders?
Both. The google-drive.md skill teaches Claude to fetch directly from a Drive folder. Local folders, Dropbox links, and S3 buckets also work. This is one of the limitations of the official MCP, fixed at the kit level.
Will this work for my niche?
If Meta ads work for your niche, this works. The reference skills are niche-agnostic. Niches it has already shipped in: B2B SaaS, low-ticket fitness, dental, hospitality, AI tools, coaching.
Who built this?
Francis Sprenger. I run IntelliBots and Low Ticket Ads Agency, which has spent over $2,000,000 on Meta ads. I built this as the production launch layer for my own client work, then productised it after enough buyers asked how to set up the same wiring on their own.
The seat is being repriced either way
For the cost of one bad lunch you turn Claude into the launch arm of every Meta account you manage. The 3 to 5 hour ad launch becomes a 10 to 15 minute brief. You ship 2 to 3x the tests, with safety rails Meta's official MCP doesn't have and the Marketing API doesn't give you on its own.
The buyers who install this in Q2 2026 are the ones who keep getting paid in Q4. The buyers who don't are the line item nobody fights for. If it does not ship a paused batch in 30 days, you get the $17 USD back.
30-day "ship a paused batch" guarantee. No subscription. Already running in 5 live client accounts.
2 launch gates. 12 hard safety rules. Hybrid MCP + API. PAUSED by default. Yours forever for $17 USD.