Growthmind watches real people use your product and tells you where they get stuck, with the recording and the numbers attached. That part costs nothing and always will. What you pay for is everything after: the experiment gets specced, tried first against a simulated version of your customer, dispatched to the coding agent you already use, verified live, measured, and closed with a straight answer.
51 of the 340 people who got this far clicked Save on the deck editor more than twice in ten seconds. The click registered. Nothing saved, and nothing told them it had failed. Most of them left the page without finishing setup.
The numbers that change between plans are how many experiments Growthmind can have running at the same time, and how much of your product it is responsible for. It caps experiments anyway, because two of them fighting over the same screen give you an answer you cannot trust.
Two kinds, one word. Read is a real person: Growthmind skims every session cheaply and only reads one properly when the skim says something went wrong. Simulated is one run of a simulation you wrote, executed on our servers instead of your machine, on demand or on a schedule. Those are the only two things counted, because they are the only two that cost us anything. Your replay tool charges you for every session it records. We only charge for the ones we read. Everything else is unlimited: seats, questions, findings, pull requests, verdicts.
| Plan | Sessions read a month | Sessions simulated a month | Experiments at a time |
|---|---|---|---|
| Free | 500 | 5 | 1 |
| $20 | Drawn from the included $20 | Drawn from the included $20 | 3 |
| Custom | As agreed | As agreed | As agreed |
Point Growthmind at your own model keys and your own infrastructure and none of the cloud numbers apply, because none of the cost is ours. No limit on sessions read, sessions simulated, or experiments. No meter to read. The same product, not a hobbled community edition.
Growthmind publishes a set of skills for whatever coding agent you already use. It reads them and writes your simulations as code, in your repo, reviewed like anything else you ship. None of that touches our servers, so none of it is charged.
Running them is the other half. A simulated customer working through your product for ten minutes would take over the machine you are trying to work on, so we run them on ours, on demand or on a schedule you set. That is the part we count.
You can install the skills, never pay us, and still end up with better simulations than you have today. That is fine. It is how you find out whether the rest is worth $20.
Not free for fourteen days. If you never pay a penny you still get to see where your product loses people, with the recording attached.
Growthmind hands the work to the coding agent you already pay for, so that runs on your token budget rather than a markup of ours. It shows you the estimated cost before it dispatches anything.
The whole company can read the findings and ask questions. Charging for the second person to care about growth would be a strange way to sell this.
Each run stops at 40 steps or ten minutes, whichever comes first, so one run always costs about the same. If a simulation needs longer than that, it is really an integration test, and you should run it in your own pipeline.
They live in your repo as code you can read. If you leave, you keep them, and they still run on your own machine or on a self-hosted install.
Cancel in one action, any month. Annual exists if you want it and is never required. The price you sign at is the price you renew at.
Some weeks Growthmind has nothing worth telling you. On those weeks it says so, and it still costs the same. Volume was never the product.