internal product tool · fintech

Honest MOCK

The production app, on your computer. Explore the installed production release with test accounts you control: follow flows, use the app and inspect its activity in one workspace.

role
product owner · design · weekly maintainer
for
product, design, operations and compliance teams
status
in use at work · 2026
Setup, the Payments flow and the Inspector open together. An internal product tool, built at Honest; shown with synthetic test data.

01why it exists

Seeing your own product should not need an engineer.

To look at one screen for one kind of customer, a product manager used to need a QA account in exactly that state, a test phone, and usually someone from engineering to set both up. Questions like “what does a customer two days past due actually see?” took a day to answer, so they often went unasked.

Honest MOCK runs the production Android app on a Mac or Windows PC against a simulator that stands in for the company’s backend. Choose a customer type and the account is rebuilt in seconds. Follow a journey screen by screen, use the app, watch what it sends, and move the calendar forward to see what happens next.

It is rebuilt each week from the release customers have, so it never drifts from the product.

412
screens and states, rendered from the app’s own code
8
journeys, from sign-in to repayment
2
languages: English and Bahasa Indonesia
6
account presets: new and existing × three card types
weekly
rebuilt for each production release

02a guided look

What you can do.

Eight features. Select any screenshot for a closer look.

01 / Workspace

Choose the panes you need.

Setup, Flows, Emulator and Inspector work together or on their own.

  • Choose a page from Setup or a journey from Flows. A pane’s name in the top menu shows or hides it.
  • Close a pane and the others fill the space. Any pane opens in its own browser window; Emulator and Inspector can also float.
  • Open panes, their sizes, the language and the appearance are remembered on reload.

1 of 8

One pane, any combination, or all four.

03getting started

Claude handles the set-up.

  1. 01

    Install the plugin

    One plugin from the company’s marketplace, on a Mac or a Windows PC.

  2. 02

    Ask Claude to set it up

    It installs the emulator and the tools it needs, and downloads this week’s ready-made app.

  3. 03

    Follow the sign-in prompts

    Each step is explained in plain language. You type the passwords and press the sign-in buttons yourself.

  4. 04

    Open your workspace

    The first visit opens the how-to guide. From then on it updates itself with each release.

The guide is the starting point on a first visit, and one key away afterwards.

04how it works

Runs locally. Uses test data.

The app, the simulator and the console run on your computer. There is no connection to QA or production backends.

01 / your computer

App + simulator + console

The app uses production code. A local simulator supplies responses and synthetic account data, constructed to mimic production.

02 / the landing page

A page into your local console

The internal site serves a static page. Your browser shows the local console inside it; console data never reaches the site.

03 / releases

Built weekly. Verified locally.

A maintainer builds from the production release. Each computer checks a release’s signature before installing it.

  1. Monday evening Find the release The version customers now have.
  2. build Build and check Production’s release plus the mock overlay. A drift report lists new calls. Every screen is checked; only changed screens render again.
  3. publish Signed release The app, the screen renders and the simulator, signed.
  4. each computer Updates by itself Downloads what changed, checks each file, installs.

05decisions

Why it is the way it is.

accuracyOnly the backend is simulated.
Production’s own code runs: screens, navigation, strings, validation and on-device rules. Only the network layer points at a local simulator, which answers with production’s fields and error codes. What you see is the product, not a picture of it, and a wrong answer fails visibly instead of looking plausible.
audienceIt is for people who are not engineers.
The QA app answers “does this build work?”. MOCK answers “what does production look like and do for this customer?”, for product, design, operations and compliance. So set-up is done by Claude in plain language, and nobody has to learn the Android toolchain to look at their own product.
installMost people never build anything.
Three roles. A user downloads the ready-made app (about 7 GB with the emulator). A prototyper adds the code and build tools. One maintainer builds each week’s release and publishes it. The common case is the small one.
consoleThe console reads like the account, not like a settings form.
Every value is edited where it is shown, one at a time; choosing a preset loads it with no Load button; there is no console sign-in, because it follows whichever account the app is signed in to. Destructive actions confirm in the page. Fewest buttons possible, because testers use it beside the live app.
test dataA test account looks like a live customer today.
Existing-customer presets carry three months of history dated up to now, with the latest statement still to pay, refreshed weekly so dates never go stale. Values follow production totals and averages; none comes from a real customer’s record, and a value production does not make clear stays unknown instead of being invented.
languagePlain words instead of version-control terms.
People make a prototype, save it, share it and bring it up to date. Every failure returns a message a non-technical person can read, plus a fix for Claude to try next. New copy is asked for in English and Bahasa Indonesia and never machine-translated.
prototypesA prototype replaces production on the emulator; it is not a second app.
It is chosen in the same place as a card type, holds only its changes against this week’s release, and ships with production’s “before” render for every changed screen. One owner per prototype; others make their own copy. Old ones are archived after eight weeks, not deleted.
upkeepEvery sync problem is written down with its fix.
Each release can break the overlay. Each problem goes into a playbook (symptom, then fix) that the set-up skill reads before it resolves the next one, and the fix is automated where it can be. The second time costs nothing.
honestyIt says what it does not reproduce.
Real money movement, server decisions such as credit scoring, vendor web pages, and real latency or outages are not simulated, and the guide says so on its first screen. A weekly report lists the calls the simulator cannot answer yet.
safetySynthetic data, no calls out, signed releases.
Test accounts are made by the simulator; every backend address is local; vendor keys are blank. Each computer installs a release only when its signature and checksum match, because the shared folder the releases sit in is readable by many teams.

06questions, answered

Accuracy, access and upkeep.

Why not use the QA app?

The QA app tests whether a build works. Honest MOCK shows what production looks like and does for a given customer, for people who are not engineers. QA keeps its time for releases.

How can the app behave like production with no backend?

Production’s own code runs. Only the network layer points to the simulator, which answers with production’s fields, types, enum values and error codes, taken from the app’s schema and each service’s code. Vendor SDKs (card security, the face check, pushes, feature flags) get stand-ins.

How do we know the answers are accurate?

The starting point is the app team’s own mock responses. Values such as balances, limits and fees follow production totals and averages. A value production does not make clear stays unknown. The weekly drift report lists the calls with no answer yet.

What can Claude do on my computer?

Only what you can do. It never stores a key or token in a repository, it keeps its files in two folders, and it asks before accepting licences, scheduling the weekly sync, pushing or publishing.

What if the maintainer is away?

That week’s release is late. Everyone keeps the previous release; the tool is not business critical. A new maintainer makes a release key and adds it by a reviewed change.

What does it cost to run?

Storage for the current release and the last four on a shared drive. The landing page is static and scales to zero. Builds run on the maintainer’s Mac. AI tokens are used only for set-up, the weekly sync report and prototype changes; the console, the emulator and Flows use none.

[built at Honest, where it is an internal tool · synthetic test data throughout · not available outside the company]

the whole ledger patricktanna.com