Chrome network logs, captured with every bug report
Updated
Midge captures Chrome network logs automatically with every bug report. Fetch and XHR requests come with headers, request and response bodies, status codes and timings; WebSocket messages are recorded as they are sent and received; scripts, styles, images and fonts are listed with status, size and timing. A developer can open any request, copy it as cURL, or download the whole log as a redacted HAR file.
You do not need DevTools open, and the person reporting the bug does not need to know what a HAR file is. Every recording includes console logs, network requests, user actions and device info.
Why the DevTools network panel is not enough
Chrome's Network panel is excellent when you are the one debugging. For bug reports, it has three problems:
- It must be open before the request happens. Requests made before you opened DevTools are not there, and most people open it after the bug.
- It clears on navigation unless someone remembered to tick "Preserve log".
- A HAR export contains secrets. Cookies,
Authorizationheaders and tokens in query strings all end up in the file. Our guide on how to remove sensitive data from a HAR file explains why that matters.
Midge captures network activity in the background on each page, keeps recent requests in memory, and attaches them when you create a report. Secrets are stripped before upload.
What a request looks like in a report
Each request is placed on the same timeline as the video and the console. Open one and you see the full exchange:
POST /api/v2/invoices HTTP/1.1
Content-Type: application/json
Authorization: [REDACTED]
{"customerId": "cus_8812", "lines": [{"sku": "PRO-M", "qty": 3}]}
HTTP/1.1 409 Conflict
Content-Type: application/json
{"error": "invoice_already_open", "invoiceId": "inv_3021"}
Along with the bodies you get the status code, the duration, and when the request started. Because requests are sorted by time next to the video, the 409 above sits right after the click that caused it instead of being buried among dozens of image loads.
WebSocket messages
Real-time apps often fail over a socket, not over HTTP. Midge records WebSocket messages in both directions with timestamps, so you can see that the server sent {"type":"price_update"} twice and the client rendered only the first one.
Other resources
Scripts, stylesheets, images, fonts and documents are listed with status code, size and timing, without their bodies. That is usually what you need: a 404 on app.4f2c1.js after a deploy, or a 3 MB hero image that delays the first paint.
GraphQL, done right
GraphQL makes network logs hard to read. Every request goes to the same /graphql endpoint, and errors often come back with HTTP 200:
{
"data": { "updateProfile": null },
"errors": [{ "message": "Not authorized to update this profile", "path": ["updateProfile"] }]
}
Midge shows the operation name, such as UpdateProfile, next to each GraphQL request and flags 200 responses that contain an errors array. A failure that a status code filter would hide is marked as a failure.
Copy as cURL
Any captured request can be copied as a cURL command, so a backend developer can replay it from a terminal and see the same response. Because redaction happens first, the Authorization header and cookies arrive as [REDACTED]. Swap in a token for your own test account and run it. That is deliberate: a bug report should never hand out a working session.
Export a redacted HAR
Some teams and some vendors want a HAR file. Midge lets you download the network log of a report as HAR, and the file contains the redacted data. You can open it in the free HAR viewer, or attach it to a support ticket with an API provider without leaking your keys. If you have a HAR exported straight from DevTools, clean it with the HAR sanitizer, which uses the same rules as the extension.
What gets redacted
Passwords, card numbers, tokens and cookies are redacted in your browser before anything leaves your device. In network data that means:
Authorization,Cookie,Set-Cookie, API key and CSRF headers, and any header whose name contains words like token, secret, session or auth.- Sensitive URL parameters such as
access_token,code,api_keyandsignature. - Body fields like
password,otp,cvv,ibanorcard_number, including nested ones. - JWTs, Luhn-valid card numbers and known API key formats anywhere in the payload.
Emails and international phone numbers are half-masked. The complete list is on the redaction page.
Limits
Request and response bodies are captured for fetch, XHR and WebSocket; other resource types show metadata only. Traffic from apps outside Chrome is not part of the report, and Midge does not run on sites you added to its never-capture list. If you need a complete browser-level capture for a one-off investigation, our guide on capturing network traffic in Chrome covers the DevTools route.
Midge is free: unlimited recordings, videos up to 15 minutes, and no watermark.
Frequently asked questions
- Which requests does Midge capture?
- Fetch and XHR requests with headers, request and response bodies, status codes and timings, plus WebSocket messages. Other resources such as scripts, stylesheets, images, fonts and documents are listed with status, size and timing.
- Can I export the network log as a HAR file?
- Yes. The network log of a report can be downloaded as a HAR file, and the download contains the redacted data, not the raw values.
- How does Midge handle GraphQL?
- It shows the operation name for each GraphQL request and flags responses that return HTTP 200 but contain an errors array, which a status code filter would miss.
- Will copy as cURL include my auth token?
- No. Authorization headers and cookies are redacted before upload, so the copied command contains [REDACTED] in their place. Add your own token when you run it.