Explore Prax Brain with example data. Choose a role, then try approving a change, pausing a run or opening the evidence behind an action.
Loading the interactive app preview…
This preview uses example data. Downloads, sharing and invitations are simulated.
The engine never changes what it does. A translation layer sits between it and you, and turns every status, stage, finding and failure into words for the person reading. 62 concepts, three translations each. Pick an event and read it as each person does.
| What you can see about a verification | Guided | Balanced | Detailed |
|---|---|---|---|
| The outcome: verified, not verified, pending | Plain language | Plain language and state label | State label and machine value |
| Why a check did not pass | Plain language, no internals | Plus the failing check’s name | Check name, rule ID, evaluated inputs |
| Which checks ran | Hidden | Count and names | Full list with versions |
| Evidence chain entries | Hidden | Linked, summarised | Full chain with hashes |
| Timing and retries | Hidden | A “checking…” state | Attempt counts, backoff windows |
| Signing details | Hidden | Hidden | Key identifiers only. Never a key or signature value, at any depth. |
| Raw receipt export | Hidden | On request | Direct |
A build passes through up to twelve engine stages. Nobody running a bakery needs twelve names. The app folds them into the steps a person actually follows, and names each step after your thing: “Building the order entry screen”, never “Building”.
Everyone sees the same progress panel: module names, checkmarks, a step count. Engineers can open the technical view on top of it. Click a step to see which stages it hides. What each stage does is on the Engine page; what each mode is for is on Modes.
Some things a sentence cannot carry: a scope, a price, a choice, a live preview, a failure. For those, Prax uses one of seven cards. Every card has the same three parts: a header with an icon and a label, a body, and at most three actions. Switch the depth and the same card changes its words, never its meaning.
Before Prax starts anything that costs money, it shows the whole scope and the total, and waits. While it builds you see what is done and what it has spent. Afterwards you get a receipt against the estimate. Move through the three moments.
Spending is fenced at seven levels, from a single AI call up to the month. An owner sees two of them, a daily limit and a monthly limit, already set to safe values. Every other level is one tap away for anyone, with a plain explanation. Nothing is hidden by who you are.
Drag the month forward. Watch what the app does at 50, 80 and 100 percent, and what it never does: stop work that is already running, or restart work because a limit was raised.
Software usually fails ugly: a code, a stack trace, lost work. Prax checks each failure message against five rules at the moment it is shown. Miss one and the message is replaced by a safe fallback. Pick a raw error and watch it go through the gate.
A failed task is retried up to three times, with a diagnosis and a fix between each attempt. If the third try fails, everything Prax knows goes to the support team automatically. You are not charged for any of it. Step through what an owner sees.
The first screen asks one thing: “What brings you to Prax?” Six answers. Nothing is pre-selected and you are never asked how technical you are. The labels describe the experience you want. Your choice sets the starting depth, and you can change it in Settings at any time.
The second screen takes your idea three ways at once: type it, drop in files, or tap “Not sure yet”. On a phone the file button opens the camera, so an owner can photograph a paper order book.
Sarah runs a bakery and has never written code. Her customers phone in orders and she writes them on paper. This is her first hour, in the order it happens. She writes in her own words, and Prax answers in plain language.
Prax keeps working while you are gone, so the first thing it shows on your return is what happened. How much it shows depends on how long you were away. Drag the slider.
Four plans: Starter, Pro, Teams, Enterprise. They differ in how many projects, which modes, how the team works together and how long history stays online. The governance does not differ. Click a plan to light its column. The switch above the app window uses the same four.
Ask in the conversation, type prax submit in a terminal, or write /prax build in a channel. The request enters the same pipeline, hits the same checks and leaves the same record. Play one task and watch all four surfaces move together. Approve it from any of them.
Every wait in the app follows one ladder of five rungs. The longer something takes, the more the app tells you. No part of the product invents its own loading pattern. Choose how long the wait is.