Instant replay bug reporting: save the bug that already happened
Updated
Instant replay in Midge lets you report a bug after it happened. When it is turned on for a site, Midge keeps a rolling recording of the last 30 to 120 seconds of the page, and one click saves that window as a bug report with console logs, network requests and user actions attached. It is off by default and you opt in per site or for all sites.
The usual problem with intermittent bugs is timing. A dropdown flickers, a toast shows the wrong message, a table jumps, and by the time you start a recorder it is gone. Instant replay means you do not need to reproduce it. You already have it.
How it works
Instant replay does not record video. It records the page itself: the DOM, which is the structure of elements and text that the browser draws. Midge stores a snapshot of the page and then every change to it, such as a node added, a class switched or text updated, together with scrolling and clicks. The replay player rebuilds the page from that record, step by step.
You choose how long the buffer is, from 30 to 120 seconds. Older changes fall off the end as new ones arrive, so memory use stays bounded. When you click to save a replay, Midge takes the current buffer, adds the console and network logs from the same period, and lets you create a report.
Why not continuous video?
We considered recording the screen all the time and keeping the last two minutes. We decided against it for two reasons.
- Chrome shows a sharing indicator for as long as a capture runs. A permanent "this tab is being shared" bar is distracting, and it trains people to ignore a warning that exists for their safety.
- Video costs CPU, memory and battery. Encoding frames non-stop on every tab you use would make your browser slower, which is the opposite of what a debugging tool should do.
DOM recording avoids both. It is still heavier than log capture, because it keeps a copy of the page content, so we leave it off until you decide you want it.
Turning it on
Instant replay is off by default. You can turn it on:
- for all sites, if you spend your day testing many different apps, or
- per site, for example only on
staging.yourapp.comandlocalhost.
The never-capture list still wins. If a site is on that list, Midge does not run there at all, replay included. A good starting setup is to enable replay on your staging and local development hosts, leave it off for everything else, and add your bank and email to the never-capture list.
What a replay report contains
A saved replay is a full Midge report. Every recording includes console logs, network requests, user actions and device info. A replay is no exception. A developer opening the link sees:
- the replayed page in a player,
- console messages from the same 30 to 120 seconds, including uncaught errors (see console logs),
- fetch, XHR and WebSocket traffic with bodies and timings (see network logs),
- clicks, page changes and the fields you typed in, without the text,
- browser, operating system, window size and page URL.
Privacy in replays
Because a replay contains page content, the rules here are stricter than for a tab video:
- All input values are masked. What was typed into any field does not appear in the replay.
- Text on the page is scanned with the same patterns used for logs: JWTs, Luhn-valid card numbers, API keys and private keys are removed, and emails and international phone numbers are half-masked.
- Logs are redacted too. Passwords, card numbers, tokens and cookies are redacted in your browser before anything leaves your device.
The redaction page lists every rule. Redaction is rule-based, so an unusual secret shown as plain text on screen may not be recognized. Teams can add their own field names.
Limits to know
A DOM replay is a reconstruction, not a pixel copy. Most pages replay faithfully, but content that is not described by the DOM, such as drawings on a canvas or media playing inside the page, may not look exactly as it did. If the bug is visual in that way, use a screen recording instead.
Replay also only covers the time it was running. If you enabled it for a site a minute ago, the buffer holds at most that minute.
When to use instant replay
- Flaky UI bugs you cannot reproduce on demand.
- Exploratory testing, where you only know it was a bug after it happened.
- Demos and walkthroughs on staging, when something breaks in front of a colleague.
For bugs you can reproduce at will, a regular recording or an annotated screenshot is often enough. For general advice on what to include either way, read how to report a bug. And if you only need the network side of an incident from a teammate who was not running Midge, ask them for a HAR file and clean it with the free HAR sanitizer before it goes into a ticket.
Midge is free: unlimited recordings, videos up to 15 minutes, and no watermark.
Frequently asked questions
- Is instant replay a video?
- No. It records changes to the page structure (the DOM) and rebuilds them in a replay player. That is lighter than video and avoids Chrome's constant screen sharing indicator.
- Is instant replay on by default?
- No. It is off until you turn it on, either for all sites or only for the sites you choose. Console and network capture work without it.
- How far back can it go?
- You choose a window between 30 and 120 seconds. When you save a replay, Midge keeps that many seconds before the moment you clicked.
- Does instant replay record what I type?
- No. All input values are masked in the replay, and text on the page goes through the same redaction rules as logs, so tokens, card numbers and similar values are removed.