For Bolt.new apps

Shipped it with Bolt.
Ship it again tomorrow.

Bolt.new lets you prompt full applications into existence. Lastest makes sure the next prompt doesn't break the last one - with deterministic tests that replay forever for free.

Start Free →View source
Survives rapid regensZero tokens per run$0 self-hosted
Lastest test recorder capturing a user flow on a bolt.new StackBlitz app
Click. Done.

Why Not Just Ask the AI?

You can ask Bolt to test what it built. Here's why that breaks.

Bolt.new is built for speed - prompt, regen, regen, regen. Every iteration is another chance for something subtle to drift. The fastest way to ship broken things is to let the same agent that builds them also test them.

01

Ask the AI to test your Bolt app.

The LLM runs a vague flow in a chat, declares it's fine, and moves on. A different path every run. No screenshot, no baseline, no history. After a dozen regens nobody knows what used to work - and every re-run is another token bill.

  • Non-deterministic - never the same flow twice
  • Tokens compound - every regen is a new round of testing tokens
  • No visual layer - layout drift sneaks through
  • The model grades its own homework
02

Use Lastest. Generate once. Replay forever.

Point Lastest at your running Bolt app. Record the flows you care about once. Lastest stores deterministic Playwright code with a visual baseline and replays the exact same thing after every regen - no AI calls, no token cost, no drift. Prompt-driven workflows finally get a safety net.

  • Deterministic replays - same flow, same screenshots, every run
  • Zero-token replays - regen 100 times for $0 in test cost
  • 3 visual diff engines - pixelmatch, SSIM, Butteraugli
  • Self-healing selectors - 7-layer fallback survives Bolt re-renders
  • Works with any URL - StackBlitz preview, deployed build, anywhere

Built for Bolt Workflows

Prompt fast. Regress nothing.

Two features that matter most when every prompt is a potential regression.

01 · Record

Click through your Bolt app. That's the test.

Open the recorder, paste your bolt.new preview URL, walk through the flow, hit stop. Lastest captures every click, type, and navigation and generates deterministic Playwright code - no selectors to maintain, no docs to read, no source code access required.

  • Works with StackBlitz preview URLs or deployed builds
  • Captures multi-step screenshots per test run
  • Tests the running app - survives any Bolt regen under the hood
Recorder timeline capturing a bolt.new app user flow

02 · Visual diff

Did the last Bolt prompt break the UI?

Bolt's rapid-iteration loop is exactly where visual regressions hide. A prompt rewords a button, the font-weight shifts, the card grid collapses - and DOM tests are still green. Lastest compares screenshots with three diff engines (pixelmatch, SSIM, Butteraugli) so regen churn can't silently ship broken pixels.

  • Text-region-aware diffs reduce false positives from dynamic content
  • AI failure triage: real regression vs. flaky vs. env issue
  • Approve or reject each change with a one-click slider
Visual diff slider comparing baseline vs new bolt.new screenshot

Common questions for Bolt apps

Why does Bolt.new break things on every prompt?
Bolt regenerates whole files on each prompt: class names shift, components rename, layouts move. The page still renders; the regression hides until a user finds it. Lastest captures a visual baseline; the next prompt either matches it or shows a screenshot diff for one-click review.
Can the LLM just test what it generated?
It can click a happy path. Different path next run, no artifact, no baseline. Tokens compound. Lastest records the flow once and replays the exact flow forever, deterministically, against a baseline you reviewed. Replays are plain Playwright, so they cost zero tokens.
Will it work with Bolt’s deploy preview URLs?
Yes. Point Lastest at any URL Bolt serves. The reusable GitHub Action runs against preview deploys on PR open and fails the PR on a non-trivial diff.
What about full-stack Bolt apps (Supabase backend)?
Lastest tests the running app end-to-end: clicks the button, hits the API, reads the DB, renders the result. Backend regressions surface as visible diffs.
How does it differ from Chromatic or Percy?
Chromatic requires Storybook (Bolt doesn’t generate one). Percy and Applitools bill per screenshot or per seat. Lastest tests any URL with three diff engines. Self-hosted is free forever; Cloud is a flat $299/month with no per-screenshot or per-seat fees.
What does it cost?
Self-hosted is free forever: open source (FSL-1.1-ALv2), run it via Docker with unlimited screenshots and replays. Lastest Cloud is a flat $299/month if you'd rather skip the ops.

Add real tests
to your Bolt app.
In five minutes.

No cloud account. No API key. No procurement. Point it at your bolt.new preview URL and go. Open source, FSL-1.1-ALv2 licensed, $0 forever.

★ Star & Install from GitHub →
install

# 1. clone

$ git clone https://github.com/las-team/lastest

 

# 2. start

$ cd lastest && docker-compose up -d

 

# 3. point at your bolt.new preview URL

$ open http://localhost:3000

 

✓ View the source