Know why users leave.
Not just where.
Your coding assistant can build anything in an afternoon. It cannot tell you why nobody comes back. Growthmind can: it reads your sessions, names why users leave, and hands you the fix spec.
- Before launch: 200 simulated users from your ICP, and 31 give up at the invite step.
- Week one: 31 of 47 real sessions ended at the invite panel, median hesitation nine seconds.
- Verdict: the onboarding checklist moved nothing in fourteen days, so it was killed and reverted.
- Verdict: skippable invites put 18 more people a week in front of first value, so it was kept.
Every other tool waits
behind the launch.
Dashboards can only see users you already have, and only after they've stalled. A funnel chart needs hundreds of users to show a leak. You built this in a weekend; you don't have hundreds of users yet. That is exactly when the leak costs the most.
THE REACTIVE STACK
Dashboards need traffic to say anything. You just launched; you don't have it. Ship first, instrument later, wait for volume that isn't coming, decode the chart yourself. For a new product, the insight doesn't arrive late. It never arrives.
GROWTHMIND
A living model of your product and your audience, in the loop from the brief onwards. It finds the stall in simulation, fixes it pre-launch, then proves it with real behaviour once you're live.
Watch the cliff form.
Then read why.
Sixty seconds, start to verdict: one read-only command, a live launch, a user gone at the invite wall — and the memo that names the wall, prices the fix and files the change order before a dashboard would have enough data to draw the chart.
- Day zero. One command, read-only.
- Growthmind reads the repo and maps your onboarding. Four steps. No skips.
- Launch day. A real user lands, and the watch begins.
- Two steps clear in nineteen seconds. The simulation holds.
- Step three demands teammates. The cursor stalls.
- Nine seconds of hover. Tab closed. That is the cliff.
- A dashboard would hand you a number. Growthmind starts typing.
- It names the wall — with the evidence pinned to it.
- Then prices the fix, spec'd for your coding assistant.
- Signed, stamped, on your desk before you knew to ask.
Six steps. One owner.
Every loop smarter.
You keep your workflow — brief your coding assistant exactly as you do today. Growthmind stands where a growth product owner would: questioning the plan, guarding the quality, owning the number.
Brief
Tell your coding assistant what to build — "an onboarding flow that does x, y, z."
Interrogates the brief like a growth product owner: is this the right flow for your users? It challenges what your assistant never will.
Build
Work with your assistant exactly as you do today. Nothing changes.
Arms it before it starts: proven patterns for your product type, your audience's context, your past learnings. Quality baked in, not reviewed in.
Simulate
Get a pre-launch verdict instead of a guess.
Runs a simulated audience of your ICP through the build, finds where they stall or bail, and fixes it before anyone real ever sees it.
This is where every other tool begins. Growthmind has been in the room since the brief.
Ship
Go live with confidence — the flow already survived an audience once.
Ships instrumentation and guardrails with it, so every real interaction is measurable from minute one.
Observe
Read one honest memo a week — not a dashboard to decode.
Watches real behaviour against what the simulation predicted, finds the leaks, and says plainly what's working and what isn't.
Improve
Initial small, clean change orders. Your coding assistant does the typing.
Proposes tweaks, measures honestly, keeps or kills — then writes every lesson back into its model, ready for your next brief.
Every experiment ends
keep or kill.
No zombie flags, no dead code, no maybe-later. Winners get their flags removed and their code kept; losers get deleted — with the reasoning attached and the lesson banked. We publish our kills, because honest verdicts are rarer than wins.
WHY IT COMPOUNDS
Every loop, Growthmind compares what it predicted in simulation with what real users did. That calibration data feeds the model of your product and audience — so your next brief starts smarter than your last, and every week it works for you is a week no competitor can copy. Nobody else is collecting this.
WHAT YOU FEEL
Like having a growth product owner from day zero: someone who questions the plan, guards the quality, and owns the number — without the six-month search for a hire you can't afford.
Paranoid by design.
READ-ONLY START
The first memo needs zero write access. Repo write comes later, if you grant it.
CHANGE ORDERS ONLY
No autonomous deploys. Every change is a spec'd order your coding assistant implements and a human approves before it ships.
FLAGS & ROLLBACK
Changes ship flag-gated with staged rollouts and automatic rollback on guardrail breach.
NO TRAINING ON YOUR CODE
Your code and data orchestrate the work; they never enter a training set.
FULL AUDIT LOG
Every action the agent takes is logged, append-only. The security review reads itself.
Free to find.
From $20 to close.
Finding it is free. Closing it is the job. The only metered thing is sessions — real ones we read, simulated ones we run — because we only charge for the ones we read. Everything else is unlimited, on every plan.
- Session recordings, with numbers — what stalled, how often
- Problems identified and priced, with estimated value
- Pull request descriptions, ready to hand your coding assistant
- Runs one experiment at a time, closed with a real verdict — three at once is the job
- $20 of usage included every month — drawn down by sessions read and simulated runs at the per-unit rates above
- Activation and retention, both — retention cohort experiments, churn signals in real time
- Overnight guardrails with a kill switch Growthmind can pull
- Kill criterion and readout date declared before anything runs; every experiment closes persist, kill, or inconclusive
- Overage is off by default. Past the included $20, Growthmind tells you, keeps working on a smaller sample, and asks before any overage is enabled — only a human can switch it on, and off the same way. Optional monthly spend cap.
- Cross-product learning — what one product learns, all inherit
- One unified experiment ledger across the portfolio
- Dead-idea prevention: what's been tried never gets re-tried
- Guest and team access controls, SSO/SCIM, audit exports, named support
- Metered usage exists here — but only a human can switch it on, and off the same way
The full product, not a community edition. Free for teams under 50 — no limits of any kind, session data stays on your infrastructure; bring model keys, Postgres and a container runtime. Over 50 people: written license, single sign-on, audit logging, a named support contact, air-gapped deployments supported. Your history transfers to cloud whenever you're ready.
Asked plainly.
Answered plainly.
The rest of this page speaks in memos. These are the same facts in the words people actually type — fit for a search result, an assistant, or an email to your co-founder.
What is Growthmind?
Growthmind is an AI growth product owner for software teams. Product analytics records what users did, then leaves you to dig through funnels and session replays to work out why; Growthmind does that digging itself, inside the workflow you already have. It connects to your repository, your coding assistant and your analytics, challenges the plan and simulates your audience before the code ships, reads real sessions after launch and explains every drop-off with the evidence attached, then specifies each fix as a pull request, runs it as a measured experiment, and closes it with a keep-or-kill verdict — so you build a product people use, then use again.
Is Growthmind a PostHog or Amplitude alternative?
No — it sits on top of them. Growthmind connects to PostHog, Amplitude or Mixpanel and reads what they record; the difference is who does the reading. Analytics hands you funnels, charts and hours of session replays and leaves the why to whoever has an afternoon to trawl them; Growthmind reads the sessions itself and sends you the why, evidence attached, wherever you work. It is also on the job before launch, simulating your audience while a dashboard still has nothing to chart. And if you have no analytics stack yet, it brings its own AI analytics engine, so you can start without one.
Does Growthmind work with Claude Code, Cursor or Copilot?
Yes — it stands beside your coding assistant rather than replacing it. You brief Claude Code, Cursor, Copilot, Codex or Windsurf exactly as you do today; Growthmind interrogates the brief, arms the assistant with your audience's context before it builds, and hands changes back as spec'd change orders your assistant implements and a human approves.
I built my app with Cursor or Claude Code, why is nobody using it?
Because building and activating are different problems. A coding assistant can produce a working product in an afternoon, but it has no model of your audience: it doesn't know which step of your onboarding will make a real user close the tab. Growthmind supplies that missing judgment. It simulates your ICP through the flow before launch, names the step where they'll stall, and once you're live it reads every real session to catch what the prediction missed — then hands your assistant a spec'd change order to fix it. The finding is free on every plan.
Do I need analytics if my product just launched and has no users?
Traditional analytics needs traffic before it can tell you anything, which is exactly what a new product doesn't have. Growthmind works the other way round: it reads your repository, simulates your audience through the product before launch, and predicts the drop-off with zero real traffic. Once real users arrive, it reads each session individually rather than waiting for statistical volume, so the first honest memo can arrive from your first handful of users.
How do I improve activation and retention without hiring a growth person?
A growth hire is a six-month search most early teams can't afford, and the job is mostly judgment: deciding what to test, guarding quality, and killing what doesn't work. Growthmind does that job as software. It questions the brief, simulates the audience, specifies each fix for the coding assistant you already use, measures honestly, and closes every experiment keep or kill. You approve; it owns the number.
Why doesn't my analytics tell me why users drop off?
Because analytics answers where, not why. The funnel names the step; the reason is scattered across session replays, events and screens that someone has to watch and cross-reference by hand — and nobody has the hours. Growthmind does that manual layer as software: it reads each session, works out what actually went wrong, and sends the answer in plain English with the replay and the numbers attached — a message wherever you work, not another dashboard to decode.
How do I find where users drop off in onboarding?
Ask Growthmind — it reads your real sessions and names the step, the reason and the cost, instead of leaving you to cross-reference a funnel chart with hours of replays. Before launch, when there's no traffic to read, it runs a simulated audience of your ICP through the flow to catch the stall early. The free plan does the finding: what stalled, how often, and what it's costing you.
Can Growthmind run A/B tests?
Yes — on paid plans it owns the whole experiment. The kill criterion and readout date are declared before anything runs, the change is tested in simulation first and ships flag-gated as a pull request, and every experiment closes with an explicit verdict: persist, kill, or inconclusive. Losers are deleted, not left behind a dormant flag.
Is Growthmind open source?
Yes. The product is open source under MIT at github.com/growthmind-ai/growthmind — growth advice you can't inspect is astrology, so ours ships with the reasoning attached. Self-hosting the full product is free for teams under fifty, with session data staying on your infrastructure.
How much does Growthmind cost?
Finding problems is free forever — not a trial. Closing them is $20 a month, with $20 of usage included every month, drawn down by sessions read and simulations run; usage beyond that is off by default until a human switches it on. Portfolios and enterprises are priced by agreement. See pricing for the plans in full.
Has Growthmind launched yet?
Not yet — the stamps say coming soon for a reason, and the figures in the specimen memo and ledger above are illustrative, not customer results. The MVP is being built in the open right now: follow the changelog, watch the repo, or leave your email to be first in line.
Build a product people use,
then use again.
The audit is free, the access is read-only, and the first memo lands within the day. Worst case: you learn where your setup leaks and what it's costing you.
Filed. One memo when Growthmind takes connections — that is the whole sequence.
That didn't send. Try again in a minute, or email founders@growthmind.ai and we'll add you by hand.
One email at launch — no drip, no digest. Something to discuss now? Talk to the founders.