Every claim licensed against your own material
Cold email that can prove what it says.
MarketingOS reads your site, your docs and your decks into a set of facts it is allowed to claim. Then it plans the campaign, writes each email against that set, and blocks anything it cannot source. The number in your subject line came from your material or it does not ship.
Invite-only while we are in testing. Sign in with the account you were given.
- Evidence ledger
- Numbers, prices, quotations and URLs are checked back against your sources before a draft can ship.
- Cold reads
- Every draft is read by a model that never saw the brief. If it cannot say what you sell, the draft loses.
- Priced before you buy
- A run tells you what it will cost in model calls before it spends anything.
A run is a fixed sequence, not an improvisation.
No model decides what happens next. The order is written in code, which is why two runs of the same campaign are comparable and why a failure has a place to be rather than a mood.
-
It learns your business once
Your site, PDFs, decks and screenshots are compiled into six artifacts — your evidence, your voice, your audience, what you sell, and what you cannot yet prove. Keyed to a fingerprint of the material, so an unchanged corpus is never paid for twice.
-
It maps the market you are actually in
Who this buyer is really deciding between, which of your claims are yours alone, and which arguments every competitor is already making — so the campaign does not open with a sentence three other emails also opened with.
-
It plans the sequence
One strategy pass turns your request into a brief per email: what each one is for, which evidence it spends, and what the reader is meant to do next.
-
It writes, checks, and rewrites
Each draft goes through deterministic gates, then a cold reader who has never seen the brief, then a critique, then a rewrite of what did not land. Taste sends a draft back; only code stops it shipping.
What actually blocks a draft
The difference between a tool that writes marketing copy and a tool you can put your name on is what it refuses to send.
Unsourced facts
Every figure, price, quotation and link is matched back to your own material. Comparisons ignore typography on purpose — a CMS makes apostrophes curly and a model quotes back in ASCII, and an exact match would drop correct proof.
Emails that never say what you sell
A draft has to name the offering somewhere the reader is actually looking. The sign-off does not count: it answers who sent this, not what this is.
Proof that was never spent
The same check runs in reverse. An email built on a piece of evidence that never made it into the copy is an email that argued from nothing.
Arguments everybody makes
A short, closed list of framings your competitors are already using. If the campaign leans on one of them, it is sent back.
Drafts nobody could decode
The cold reader's one veto. A predicted click-through on an email the panel could not parse is an answer to a different question.
Runs that would waste your money
A campaign with nothing a stranger could check is stopped before the first model call, with the gap named, rather than after it has been paid for.
What it is not
Worth reading before you sign in, so the first run matches what you expected.
- Not a sending platform. It writes and checks the campaign; you send it from wherever you already send.
- Not a list builder. It qualifies companies against a segment you defined; it does not scrape addresses.
- Not a one-click button. It needs your material. Given nothing to work from, it refuses the run rather than inventing something.
- Not free to run. Every run spends real model quota, which is why the cost is shown before you spend it and why accounts are invite-only for now.
Bring your material. Get a campaign that can defend itself.
Sign in with the account you were given, add a business, and point it at your site. The first campaign starts from everything it learned.