Set up an app end-to-end without leaving your terminal. Run minitest init and Mini handles onboarding where you work, then hands off to the web app to build and run.
Onboard from your terminal
Set up an app end-to-end without leaving your terminal. Run minitest init and Mini handles onboarding where you work, then hands off to the web app to build and run.
- Guided setup, in the CLI: Mini reads your repo, asks what it needs, and drafts your scenarios and personas conversationally. Also available over MCP for your AI editor.
- Optional GitHub step: connecting GitHub is now optional, so you reach a first run faster.
- Multi-select answers: hand Mini more context in fewer round-trips.
- Live all-set screen: watch scenarios populate live, then jump into your first run.
GitHub or Bitbucket, your choice
Mini now speaks Bitbucket, not just GitHub. Pick your provider during onboarding and everything downstream adapts.
- Choose at onboarding: connect GitHub or Bitbucket when you create an app.
- Provider-aware everywhere: commit context, wording, icons, links, and the connected-repo card all reflect your provider.
- Same power on both: repo-linked builds, commit runs, and PR triggers work identically on either.
Richer web testing
Web testing is now wired to your repo and your real viewports.
- Run a specific commit: pick a commit and Mini builds and serves a live preview of your web app from your repo, then tests that exact version.
- Desktop & tablet, first-class: the scenario modal and demo render desktop and tablet formats, with one target chip per browser/viewport on the report.
- Preview, verified: Mini confirms the preview URL is reachable before starting a run.
A reworked Runs experience
The run page is now board-first, with progress always in reach.
- Board-first run detail: see every scenario's status at a glance mid-run.
- Floating runs bar: an always-visible bar tracks in-progress runs as you move around the app, with clear progress and queued/building states.
- Pass-rate donut: an accurate donut summarizes each run, and platform tabs show issue counts in red.
The New user persona
Every app now ships with a default "New user" persona so a first-time-user journey is always covered.
- One persona, guaranteed: apps enforce at least one persona, with a designed New user persona out of the box.
- Recognizable everywhere: its avatar renders consistently across scenario and run views.
- Dedicated inbox: the New user persona comes with its own email inbox, so Mini can validate OTP flows end-to-end.
Scenarios: dependency board & filters
The Scenarios page is now an edit-focused dependency board built for large suites.
- Dependencies view first: see the whole flow with dependencies by default.
- Category filter: filter by category to focus on part of the suite.
- Cleaner detail: open a scenario as a floating card to edit it without losing your place.
Test maintenance from the CLI
Keep your suite in sync with your code from the terminal.
- Maintenance in the CLI: the new minitest maintenance command group, with interactive apply-now vs. review-link UX and incremental/audit modes.
See exactly what Mini did
The run recording now shows the agent's every move. A live overlay draws each on-screen action as it happens, so you can tell "the app broke" from "the agent mis-tapped" at a glance.
- Every gesture, drawn: taps, long-presses, swipes, drags, and pinches render on the recording, synced to playback.
- Human-like and readable: natural tap entrances and comet-like swipe traces keep the overlay clear without hiding your app.
- On by default, toggleable: flip the overlay off per run whenever you want the raw frame.
Note: this is only rolled out for IOS runs for now. We’re releasing the Android support in the incoming weeks.
Mini now flags UI/UX issues
Beyond pass or fail, Mini spots design and UX improvement opportunities as it tests. These findings never change a scenario's verdict, they just surface things worth a look.
- Findings on the run: a UI/UX findings section sits under the acceptance criteria, showing what Mini spotted in that run and whether it's new, still present, or resolved.
- A dedicated UI/UX Suggestions tab: every finding in one place, sorted by last appearance, with the screen it was seen on, and one-click triage.
- The right evidence: point-in-time findings show a screenshot at the exact moment; temporal ones show the clip, labelled with its interval.
Smaller improvements
- Advanced Build Settings: build toggles now live under a collapsible Advanced Build Settings section, including a new "Disable looping animations" toggle (on by default) that cuts flaky animation noise.
- Firefox live recordings: live device recordings now work on Firefox too.
- Smarter build reuse: Mini rebuilds when your env vars or build options change instead of serving a stale build.
- Cleaner run emails: real app name and icon, build label, and a direct commit link.
- Per-app Android signing key: each app gets its own Android debug keypair, shown in App Settings.
- Build-time env vars for web apps: edit them directly from settings.
- RevenueCat sandbox patches: an opt-in toggle on Build Settings.
- Revamped App Settings: a cleaner single-scroll layout with a sticky sub-nav.
- Sharper git context: provider-aware labels and links, with commit context preserved on rerun.
- Self-healing CLI auth: the CLI recovers from a stale or revoked refresh token instead of locking you out.
