Forgot to open DevTools? How to see network requests that already happened

Chrome's Network features reference says it plainly: DevTools records all network requests "so long as DevTools is open." Nothing runs in the background before that. When you open DevTools after a one-off 500, a login redirect or a race on first load, the Network tab is empty because the requests were never written down anywhere you can reach.

There is no hidden cache to dig through. Browser history stores the URLs of pages you visited, not the API calls those pages made. Every working option below records the next occurrence. Choose one based on how hard the bug is to trigger.

If DevTools was open and the Network tab still shows nothing, or not the request you expected, check these first:

• A filter is still on. Clear the filter box, click All in the type chips, and make sure Invert is unchecked. Under More filters, Hide data URLs and Hide extension URLs remove rows too.
• Recording is stopped. The round record button at the left of the action bar should be red. Grey means Chrome is not recording. Click it, or press Ctrl+E (Cmd+E on macOS) with the Network panel focused.
• You're looking at the wrong tab. A link that opens a new tab or popup gets its own DevTools window, and its requests never appear in the original tab's Network panel. Requests from iframes do show in the parent page's panel.
• A service worker is involved. A worker can answer a request from its own cache, and requests it sends on its own, such as background sync, show up in the worker's DevTools. Open that from chrome://inspect/#service-workers.
• The request happened before DevTools opened, as covered above. Reload with DevTools open.

For most bugs this is the fastest route. Open DevTools on the tab, tick Preserve log, then reload or repeat the action. Preserve log matters for login and checkout bugs, because Chrome clears the list on every document load and the request you need is usually the one right before a redirect. Preserve log in Chrome covers the details.

Tick Disable cache too if the bug only shows for a first-time visitor. Cached requests can skip the network on a reload, so the call that failed the first time may not repeat.

When it fails, export a HAR file before you close DevTools. You can open it later in the free HAR analyzer, which runs in your browser and lists the failed, slow and large requests.

If the bug happens on first load or in a tab you never think to inspect, make Chrome open DevTools for you. Quit Chrome completely, then start it with the --auto-open-devtools-for-tabs switch. Every new tab and window opens with DevTools attached, so recording starts before the page's first request.

# macOS
open -a "Google Chrome" --args --auto-open-devtools-for-tabs

# Windows
"C:\Program Files\Google\Chrome\Application\chrome.exe" --auto-open-devtools-for-tabs

# Linux
google-chrome --auto-open-devtools-for-tabs

The switch only takes effect when Chrome starts, so it does nothing if a Chrome window is already running. A separate profile avoids that problem and keeps your everyday browsing out of it. Add --user-data-dir=/tmp/chrome-debug to the command.

For popups only, there is a lighter option. In DevTools, open Settings (the gear icon), then Preferences, and check Auto-open DevTools for popups in the Global section. Chrome's preferences reference says it opens DevTools when you click links that open new tabs, which catches OAuth and payment popups that redirect before you can press F12.

chrome://net-export writes a NetLog, a browser-wide log of network events that does not need DevTools at all. It is the tool Chromium developers ask for in bug reports, and the steps are in Providing network details.

1. Open a new tab and go to chrome://net-export.
2. Click Start Logging to Disk and choose where to save the JSON file.
3. Reproduce the problem in a different tab. Keep the net-export tab open, because closing it stops the log.
4. Click Stop Logging when you are done.

Like DevTools, net-export only records from the moment you click Start. Leave it running in the background on days you're chasing an intermittent bug. A long log grows quickly, and you can cap it with the --net-log-max-size-mb startup flag.

The default capture level, Strip private information, removes cookies and credentials. Include raw bytes adds the bytes sent and received, and the Chromium docs warn that this level may include personal information. Don't share that file casually.

If the bug already happened and you can't make it happen again, the server may be the only place it was recorded. A one-off 500 usually leaves a line in your access logs, a stack trace in your error tracker or APM, and a trace if you run distributed tracing. Search around the time you saw the error, filtered by your user ID or IP. If your API returns a request ID header, log it on the client too, so you can match a user's report to a server log line next time.

Server logs show what the server received and returned. They miss requests that never reached it, such as CORS failures, blocked requests and calls that timed out on the client. For those, use one of the browser options above. See how to fix a CORS error if that turns out to be the cause.

Netraptor is a Chrome extension with a panel in DevTools. Its Keep history for this site switch records a site's API calls while DevTools is closed, so the next time a bug happens before you open DevTools, the calls are already saved.

• Turn it on from the Netraptor toolbar popup, or the History pill in the Netraptor panel. Chrome asks for access to that site, and the switch covers the domain and its subdomains. Reload the page once to start.
• Netraptor hooks fetch and XMLHttpRequest in the page when it starts loading, in every frame, so API calls made during the first load are captured too. Each call is saved with its status, headers and bodies, including the ones that fail.
• Calls are saved to IndexedDB in your browser. Saved sessions appear in the Netraptor panel's session picker next to the live one, grouped by page load (Reloaded or Navigated to) with single-page-app route changes split out.
• Filter a saved session with status-code:5xx to find the one-off 500, then export it as HAR, Postman or curl.

Know the limits before you rely on it. Keep history records API calls only (fetch and XHR). Documents, scripts, styles, images and redirects of the page itself are not saved. It records only on sites you turned it on for, and only after you turned it on, so it can't recover a bug from yesterday. The Free plan keeps the last three sessions and Pro keeps all of them.

Opening the Netraptor panel without Keep history works like the Network panel. It pulls in everything DevTools has already logged for the tab, so you don't need to reload when you switch to it, but it can't see requests from before DevTools opened either.

The practical setup is to turn on Keep history for the apps you build and test. If you want a coding agent to read those calls instead of guessing, see how to let your coding agent see your app's API calls.