Skip to content
Voicetrail

● RECFor web app reviews

Review a web app out loud.

Use the app and say what you see. Voicetrail records the pages, your voice, your clicks and the errors underneath on one timeline, then writes up the issues. Your team gets a link. Your coding agent gets the whole thing over MCP.

Chrome, Edge, Brave and other Chromium browsers. Card required; cancel before day 15 and you pay nothing.

The Voicetrail toolbar recording a shop page. The reviewer has circled a product card and is pointing at its price, which reads $NaN; the picker shows the element's CSS selector.

Recording our test shop: the reviewer circles a card and points at a broken price.

One review, one timeline

Real data from the recording above

Timeline of a 62-second recorded review: 17 spoken sentences, 13 clicks and page changes, 5 annotations and 6 errors and failed requests, all on one time axis.

00:12What it records

Everything a bug report leaves out.

A good review is what someone says while using the product. Voicetrail keeps that, and lines it up with what the page was doing at the same second.

Page

A replay of the page itself, not a video

The page structure and every change to it, so text stays sharp at any zoom and a 10-minute review is a few megabytes.

Voice

What you said, with live captions

Your microphone, transcribed while you talk and again after you stop, with a timestamp on every sentence.

Clicks

Clicks, typing, scrolling, page changes

Each click names the element and its CSS selector. Passwords, card numbers and codes are hidden before upload.

Marks

What you pointed at

Pick an element, circle it, draw an arrow, drag a box or type a comment. Each mark records the elements under it.

Errors

Console errors and failed requests

Exceptions, rejected promises, console errors and every request that came back 400+ or never arrived.

Issues

The issue list

An AI pass groups what you said into issues, each with your quote, the element, the page and the errors around it.

00:31How it works

Four steps, and only one of them is work.

  1. 01

    Start on the tab

    Open the app you want to review and press Start, or Alt+Shift+R. It works on production, staging and localhost.

  2. 02

    Use it and talk

    Click around and say what you notice. When you say "this button", point at it with the picker.

  3. 03

    Stop

    The recording uploads to your workspace. A minute later it has a transcript and an issue list.

  4. 04

    Hand it over

    Send the link to your team, or ask your coding agent to fetch the review and fix it.

The recording in the viewer: the page replay on the left with the circled card, a scrubber with colored event markers, and the transcript on the right where each sentence lists the clicks and errors that happened while it was spoken.
Replay, scrubber and transcript. Click any sentence to jump there.
The Issues tab: issues grouped from the narration, each with severity, the reviewer's quote, the element and related errors.
The issue list, written from what the reviewer said.

00:58For coding agents

Your agent reads the review, not a video.

Every recording comes with timeline.md: what you said, merged with what happened at that second, with CSS selectors, errors and screenshot references. A 10-minute review is a few thousand tokens.

Connect Claude Code, Cursor, ChatGPT or any MCP client once. Then ask for the review by name.

you ›

Get my latest Voicetrail review and fix the issues.

Set up your agent
timeline.mdexcerpt, unedited
# Feedback session: Test Shop
- Recorded: 2026-11-01 18:06 UTC, length 01:02
- App: http://localhost:8411

## Feedback items (AI summary)
- [00:03] (bug, high) Add to cart fails silently with a 500 error — button "Add to cart" in div.card "Blue mug $12 Add to cart" ([data-testid="product-1"] > button.btn.add); cart counter link "Cart 0" (a[href="/cart"]) — "Let me add the blue mug to the cart. Hmm, nothing happened. The cart counter up here still says zero."
- [00:11] (bug, high) Product price renders as $NaN — div.card "Notebook $NaN Add to cart" ([data-testid="product-3"]) — "And look at this notebook card. The price shows NAN dollars. That's clearly broken."
- [00:20] (bug, medium) Filter button does nothing on click — button "Filter" (#filter) — "This filter button here doesn't do anything when I click it."

## Timeline
00:00  NAV       page /  "Test Shop"
00:01  SAY       "Okay, I'm on the Test Shop products page."
00:03  SAY       "Let me add the blue mug to the cart."
00:04  CLICK     button "Add to cart" in div.card "Blue mug $12 Add to cart"  ([data-testid="product-1"] > button.btn.add)  [shots/001.jpg]
00:04  NET.FAIL  POST /api/cart → 500
00:04  CONSOLE.ERROR Add to cart failed: HTTP 500
00:06  SAY       "Hmm, nothing happened. The cart counter up here still says zero."
00:09  POINT     at: link "Cart 0" in nav "Products Cart 0 Account About (full page)"  (a[href="/cart"])  [shots/002.jpg]
00:11  SAY       "And look at this notebook card."
00:14  SAY       "The price shows NAN dollars."
00:14  CIRCLE    around: div.card "Notebook $NaN Add to cart"  ([data-testid="product-3"])  [shots/003.jpg]
00:17  SAY       "That's clearly broken."
00:20  SAY       "This filter button here doesn't do anything when I click it."
00:21  ARROW     pointing at: button "Filter"  (#filter)  [shots/004.jpg]
00:21  CLICK     button "Filter"  (#filter)  ×2
00:24  SAY       "Let me try apply discount."
00:26  CLICK     button "Apply discount"  (#break)
00:26  EXCEPTION Uncaught TypeError: Cannot read properties of undefined (reading 'percent')  at /app.js:30:86
…

01:20Who it's for

Anyone who has to explain what's wrong with a screen.

QA testers

Test the way you think out loud and skip the write-up. Every bug arrives with steps, the element, the console error and the failed request.

Product managers and designers

Walk through a build and say what bothers you. Engineers get your words next to the exact screen and element.

People building with AI coding agents

Review what your agent built, then let the same agent pull the review over MCP and fix it. No copying screenshots into chat.

Agencies reviewing client sites

Record a pass over a staging site and share the link with the client or the developers, with every issue already listed.

01:47Compared with video

Why not just send a screen recording?

Video is fine for showing someone a flow. For fixing it, the developer (or the agent) needs to know which element you meant and what broke underneath.

QuestionA screen recordingA Voicetrail recording
Which element is "this"?You guess from the pixelsThe element you pointed at, with its CSS selector
Errors behind the screenNot in the videoConsole errors and failed requests at the second they happened
Size of 10 minutesTens or hundreds of MBA few MB, most of it audio
What an agent can readFrames, through a vision modelA text timeline of a few thousand tokens
Finding a momentScrub and watchSearch the transcript, click a line, jump there
Text on screenCompressed, blurry when zoomedReal text, sharp at any zoom

02:05Privacy

Hidden in the browser, before it leaves.

Masking runs in the extension, so masked text is never uploaded. Recording happens only in the tab you started, between Start and Stop.

How we handle recordings
  • Standard masking, on by default

    Passwords, card numbers, one-time codes, token-like fields and secrets in URLs are replaced with dots.

  • Strict masking for real customer data

    All typed text and all page text become asterisks, screenshots turn off. The layout still replays.

  • Allowed domains

    Admins can limit recording to your own sites, like *.staging.yourco.com.

  • Retention you choose

    One year by default, or 30 or 90 days. Deleted recordings are gone for good after 7 days.

02:30Pricing

$27 a seat a month. 10 hours each, pooled.

Or $22 a month billed yearly. People who only watch are free. Start with a 14-day trial and 3 recording hours. Prices in USD, + tax where applicable.

02:51Questions

FAQ

Is it a screen recorder?

No. Voicetrail records the page itself (its structure and every change to it) instead of pixels. The replay stays sharp at any zoom, a 10-minute review is a few megabytes, and every click is tied to the exact element that was clicked, with a CSS selector.

Which browsers does it work in?

Chrome, Edge, Brave, Arc, Opera and Vivaldi on desktop. Firefox and Safari aren't supported.

Do I need to add code to my app?

No. It's a browser extension, so it works on any site you can open: production, staging or localhost. Site owners can optionally mark elements with data-fbr-mask or data-fbr-block to hide them from every recording.

What does the AI do with my recording?

ElevenLabs transcribes your voice. Then a language model reads the text timeline (what you said, merged with what happened on the page) and groups it into issues. It only sees text, never audio or screenshots, and it doesn't train on your data.

Which coding agents can use it?

Anything that speaks MCP: Claude Code, Claude.ai and Claude Desktop, ChatGPT, Cursor, VS Code with Copilot, Windsurf, Codex CLI and Gemini CLI. Your agent runs on your own subscription; Voicetrail only serves the data.

What languages does it understand?

Transcription is done by ElevenLabs Scribe, which handles more than 90 languages, including Spanish, French, German, Portuguese, Japanese and Vietnamese. We test the issue list mostly in English today.

What happens when the trial ends?

The trial lasts 14 days and includes 3 recording hours. A card is required to start. If you don't cancel, it becomes a Team subscription on day 15; we email you 7 days before. Cancel during the trial and you pay nothing.

What if we run out of hours?

Live captions pause for the rest of the month, and everything else keeps working. You can buy 10 more hours for $12. Recordings past 150% of your hours still upload, and are processed next month or when you add hours.

Record your next review instead of writing it.

14 days, 3 hours of recording, then $27 a seat a month.

Start the trial