You pay per active program, not per contact
Most lifecycle email tools price by list size. That rewards bloat. Send to a million contacts you'll never convert, pay more. Mara prices by what she's actually running for you. Add a churn-save program, the bill moves. Stop running it, the bill moves back.
Everything below is one row of that table, explained. The two capacity rows first, since they're the only real difference between the plans, then the ten things she does, which are the same whichever plan you're on.
Active programs
A program is one lifecycle sequence Mara runs end to end: the trigger, the emails, the waits between them, the branches. There are seven kinds she can build today. Welcome, activation, feature adoption, re-engagement, churn-save, expansion, win-back. Each running program is one unit against your plan. Starter runs three at once, Growth runs ten.
The count is programs that are actually live. One you're still drafting doesn't count. One you paused doesn't count. An archived one doesn't count. Pause a seasonal win-back in July and the slot is free until you resume it. The check happens at the moment you activate, so nothing gets switched off behind your back when you cross a line.
Dunning, the failed-payment rescue, has a page on this site but isn't one of the seven yet. Mara can't draft it, so it can't take a slot.
Need more than ten programs, a dedicated IP, or a volume commitment? Reach out and we'll size a plan to you.
Sends a month
Twenty-five thousand included on Starter. A hundred and fifty thousand on Growth. A send is one email to one person.
The included number is an allotment, not a wall. Go past it and the extra sends are billed at a dollar fifty per thousand. They aren't blocked and your programs don't stop mid-sequence, because a lifecycle sequence that halts halfway through is worse than a small invoice. Mara tells you when you cross eighty percent, and again at a hundred, so the bill is never the first you hear of it.
Contacts
Unlimited on both plans. Import your whole list, connect a product with a hundred thousand users, none of it changes the price.
Same argument as the value metric itself: you're billed for the work Mara does, and the work is the programs and the sends. A row sitting in a table isn't work.
Everything she does, on both plans
The brain doesn't change with the price. Every specialist, every gate, every loop below runs the same on Starter as on Growth. Growth buys capacity, not a better Mara.
Every email drafted in your voice
She builds a voice profile from two places. First your public material: your marketing site, your changelog or release notes, your public roadmap if you have one, and your GitHub repo if you connect it. Second, and this is the one that matters, the emails you paste during onboarding. Give her one to three you actually wrote and she distills them into a fingerprint. How formal you are, how long you run, your cadence, how you sign off, whether you write as I or as we, plus verbatim excerpts she keeps as reference. That captured voice overrides anything she guessed from your website. Skip the step and the site voice is the fallback.
Before a draft reaches you she grades it against eight named things a good founder email does, then rewrites what fell short. Humble, first person, reader first, ending on a question rather than a shouty button. Welcome and activation emails default to that conversational shape. Button-style calls to action are kept for genuinely transactional mail.
Programs proposed from your product
Mara doesn't ask you to invent a strategy on a blank page. She watches real events from your product, through the Mara API, your billing webhooks, Clerk, or Supabase, and lays them against the stages your customers actually move through: visitor, signed up, activated, paid, retained.
She proposes the mapping and you confirm it. You never type an event name. She shows you what she saw and you say that's right, or move it, or that's not a step. Where the map has a gap she proposes the program that fills it, and as results accumulate she keeps proposing new ones. Those arrive as ideas in your queue, never as live sends.
Nothing sends without your yes
Every email Mara writes lands in a queue as pending approval. New accounts start in review mode and there's no way to skip your first approval, because the first email a founder sees from Mara should be one they chose to send.
If you later switch to autopilot, three things still hold. Your first program has to have been approved by hand at least once. Win-back, churn-save, and expansion always come back for review, because a wrong send there is expensive. And any reply class you haven't explicitly graduated still waits for you. You can pause a running program at any time and it stops.
Variants written and tested on live sends
On a step with testing enabled she doesn't write one email. She writes several versions of it, splits the real sends between them, and reads what came back: delivered, opened, clicked, replied.
It's on by default for welcome, activation, and feature adoption. High volume, low stakes, and copy differences show up quickly there. It's off by default for the high-stakes programs, where she personalizes to the individual instead, which is the better trade. You can flip it either way per step.
Winners get more send share, losers get rewritten
Two separate things happen with those results, and this loop is the actual product.
The first is allocation. Send share shifts toward whatever is winning, inside the set of variants you already approved. Nothing new goes out, the mix just changes.
The second is rewriting. When a variant is clearly beaten, and only on a clear signal rather than noise, she commissions one rewrite of it and retires the loser. The rewrite is new copy, so it comes back to you for approval like anything else. That's the difference between a tool that reports a result and someone who acts on it.
Personal context on win-back and churn-save
On the four high-stakes programs, win-back, churn-save, re-engagement, and expansion, she reads that specific person's history before writing to them. What they did in your product, how far they got, what you already sent them, what they replied.
That becomes real context in the email, not a merge tag. And she can't invent one. The fields she's allowed to reference are checked against what your data actually contains, so an email can't go out promising a link that doesn't exist.
Replies read, sorted, and answers drafted
Replies come back to Mara instead of into nothing. She classifies each one, interested, feedback, support request, referral, complaint, unsubscribe, out of office, an auto-bounce, and drafts an answer in your voice.
By default every one of those drafts waits for you. You can graduate a class to send itself, and the guardrails are in the code, not just on the settings screen. Only five classes are ever eligible. A class only unlocks after you've approved twenty replies in it, so she's demonstrably learned your voice first. Complaints and legal threats can never be graduated, by anyone. And a single conversation runs at most two automatic turns before it comes back to you.
A weekly report in plain words
Once a week, a short note from Mara. Two to four paragraphs on what she sent, what happened, and what she wants to change, ranked so the thing that matters is first. Not a dashboard link, not forty tiles. The numbers are on your Results page when you want them.
Your own sending domain
You point DNS at a subdomain you own, usually mail.yourdomain.com, and Mara sets up your own mailbox behind it. Nothing sends until that domain verifies, which is deliberate. Reply handling is switched on at the same moment, without you having to ask for it.
The reputation you build is attached to your domain. It's yours while you're a customer and it's still yours if you leave.
Her playbook of what worked, with evidence
What she learns gets written down where you can read it. Each learning is a sentence, a category, and the evidence behind it, which you can open. Each carries a confidence level: hypothesis, supported, or validated.
The level decides what it does. Supported and validated learnings feed into her drafting. Hypotheses don't, they go back out as things to test. Every piece of cited evidence is checked against your own account before the learning is stored, so she can't cite a result that doesn't exist. You can write your own learnings, retire hers, or turn the whole mechanism off in settings.
Try before you hire
Point Mara at your URL and watch her draft a starter program in your voice. That part is free, no card.
When you're ready to send, every plan starts with a seven day trial. Card up front, cancel any time before it ends and you pay nothing.
What's not in any plan
Paid ads. SEO content. Social posts. Broadcast newsletters. Transactional sends. CRM. Sales outreach.
This list isn't a "coming soon." It's a "not the product." If you need any of those, Mara isn't the hire.
A few questions you might have
Why per active program instead of per contact? Per active program maps to what Mara does. Per contact rewards list bloat. Per program rewards adding the win-back you've been putting off. Add a churn-save program, you pay for the value you got. Your list can be any size.
Does Mara write a different email for every single person? No, and that's on purpose. The emails improve through iteration. Mara drafts a few variants per segment, sends them, watches opens, clicks, and replies, then shifts to what works and rewrites what didn't. That learning loop is the product. On high-stakes programs like win-back and churn-save she does add real per-person context, because there the personal touch earns its cost.
Will Mara sound like us, or like a chatbot? She reads your public material, and during onboarding you paste one to three emails you actually wrote, which become her reference for how you sound. Drafts come back as plain text. Read them, edit them, or kill them. What you reject and rewrite is signal she keeps.
What happens if she sends something wrong? There are two modes, and new accounts start in review, where every email waits for your yes. The other is autopilot, and even there your first program needs a manual approval, the high-stakes programs still come back to you, and you can pause anything instantly. Every send is logged with the reasoning behind it.
Do I have to leave my current email tool? No. Mara runs alongside it and sends from your own domain. She's the brain and brings her own pipe, so your newsletter tool keeps its job.
Can I leave with my data? Yes. Your programs and your sender reputation stay with you, and your contacts already live in your own product, not locked in ours. Cancel and none of it moves.
Hire Mara
Start your trial → Try the wedge first → See what she runs →