Fix ERR_PROXY_CONNECTION_FAILED in Chrome

Here is the page, word for word as Chromium's source defines it (checked in the main branch on 2026-10-30):

• Heading: No internet
• Summary: There is something wrong with the proxy server, or the address is incorrect.
• A "Try:" list on desktop that includes Checking the proxy address and Contacting the system admin
• The code line at the bottom: ERR_PROXY_CONNECTION_FAILED

The "No internet" heading is misleading. Your connection is usually fine. Chrome shows the same heading it uses for a dead Wi-Fi link, which is why so many people restart their router first. Don't. If other apps on the machine still load (Spotify, Slack, a game launcher that ignores the proxy), the network is up and only the proxy path is broken.

One more detail on that page is worth reading. Expand the details and Chrome adds "If you use a proxy server... Check your proxy settings or contact your network administrator to make sure the proxy server is working. If you don't believe you should be using a proxy server:" followed by platform instructions. That second sentence is the most common real situation: a proxy you forgot about, or one something else installed.

Chromium's network error list gives this error the number -130 and describes it precisely: "Could not create a connection to the proxy server. An error occurred either in resolving its name, or in connecting a socket to it." It then adds a line that saves a lot of wasted effort: the error "does NOT include failures during the actual 'CONNECT' method of an HTTP proxy."

In plain terms, the browser failed at one of two very early steps:

1. It could not look up the proxy's hostname. A typo in the host, a hostname that no longer exists, or DNS that is broken for that one name.
2. It could not open a TCP connection to the proxy's port. Wrong port, nothing listening there, a local tool that has quit, or a firewall in the way.

Anything that happens after the connection opens, such as a rejected password, produces a different error. We confirmed this on Chrome 156.0.8078.74 on Windows Server 2022 on 2026-10-30, driving the installed Chrome through Playwright with each broken proxy setting:

These are single runs from one machine, so treat the timings as rough. The pattern is the useful part: if the code at the bottom of the page reads exactly ERR_PROXY_CONNECTION_FAILED, stop suspecting your password. The host or port is wrong, or the thing at that address is not there. The auth case in the last row is automation-specific: in a normal Chrome window a missing password shows a Sign in prompt reading "The proxy [proxy address] requires a username and password." instead of an error page.

Every other proxy-related code (407, 502, timeouts) is covered in proxy error codes explained. This page sticks to the one Chrome shows when it can't reach the proxy at all.

Work down this list and stop at the first step that changes something. Each one costs less time than the next.

Before touching any settings, find out if the proxy answers at all. Copy the host and port exactly as your dashboard shows them. For SparkProxy that is gateway.sparkproxy.io, with port 11000 for HTTP/HTTPS and 13000 for SOCKS5.

On Windows, open PowerShell (Start menu, type "PowerShell") and run one line:

Test-NetConnection gateway.sparkproxy.io -Port 11000

Read the last line of output. TcpTestSucceeded : True means the host resolved and the port accepted a connection. That was our result on 2026-10-30. False (with a WARNING line above it saying the TCP connect failed) means Chrome will fail the same way, and nothing inside Chrome will change that.

On macOS, open Terminal and run nc -vz gateway.sparkproxy.io 11000. A line ending in "succeeded!" is the pass condition.

How to read the result:

• Test passes, Chrome still fails. Chrome is not using the proxy you think it is. Go to Step 2. This is the most common outcome.
• Test fails with a name-resolution message. The hostname is wrong or retired. Compare it letter by letter with the dashboard.
• Test fails on a correct host and port. Something between you and the proxy is blocking the port: a corporate firewall, a guest Wi-Fi network, or local security software. Try the same test from a phone hotspot. If it passes there, the network you're on is the cause.

If you want a fuller health check that includes the exit IP and authentication, the walkthrough in how to test if your proxy is working covers it. For this error, a TCP pass or fail is all you need.

Chrome on desktop has no proxy form of its own. It reads one of four sources, and the higher ones override the lower ones:

The fastest way to see which one is active is to open chrome://settings/system and read the line under the system section. Chromium's settings strings show exactly what you'll see:

• Open your computer's proxy settings: Chrome is using the operating system's proxy. Go to Step 3.
• Chrome is using proxy settings from an extension: an extension owns the setting. Go to Step 4.
• Chrome is using proxy settings from your administrator: a policy owns it. Go to Step 4.
• Browser proxy settings are controlled by multiple sources: both a policy and an extension are applying settings.

For the exact address Chrome is trying, open chrome://net-internals/#proxy. The page lists Effective proxy settings (what Chrome is actually using) and Original proxy settings. If the effective line names a host or port you don't recognise, you've found the culprit before changing anything. Take a screenshot of that panel now; it's your "before" record for backing out later.

One quick isolation test while you're here: fully quit Chrome (also from the tray icon on Windows), then start it once from the Run dialog (Win+R) with chrome.exe --no-proxy-server. Chromium documents that flag as overriding "any other proxy settings provided." If pages load in that window, the internet is fine and the proxy configuration is the whole problem. If a policy is in force, the flag is ignored, which is itself a clue.

When Chrome says it is using your computer's proxy settings, the fix is in the OS. Chrome, Edge and many desktop apps share this one setting, so a value left behind by a VPN client, a debugging tool, or an old proxy trial breaks all of them at once.

Windows 11 and Windows 10. Open Settings and go to Network & internet > Proxy. There are two blocks to check:

• Under Automatic proxy setup, look at Automatically detect settings and Use setup script. A setup script pointing at a file or server that no longer exists can send Chrome to a dead proxy. If you didn't configure a script on purpose, turn Use setup script off.
• Under Manual proxy setup, click Set up next to Use a proxy server. The Edit proxy server dialog shows Proxy IP address and Port. A value like 127.0.0.1 with some port usually means a local tool registered itself as the proxy and then quit without cleaning up. If you don't use a proxy, switch Use a proxy server off and click Save. If you do, retype the host and port from your dashboard rather than editing the old entry.

Windows uses one proxy value for HTTP and HTTPS in that dialog, so put the HTTP port (11000 for SparkProxy) there, never the SOCKS5 port.

macOS. Choose Apple menu > System Settings > Network, click your active network service, then Details > Proxies. You'll see toggles for Auto proxy discovery, Automatic proxy configuration, Web proxy (HTTP), Secure web proxy (HTTPS) and SOCKS proxy. Two traps catch people here:

1. Each network service has its own proxy settings. Fixing Wi-Fi does nothing if you're on Ethernet or a USB tether. Check the service that's actually connected.
2. An HTTP endpoint goes in both Web proxy (HTTP) and Secure web proxy (HTTPS). A SOCKS5 endpoint goes only in SOCKS proxy, with the SOCKS port.

Click OK after any change; the panel doesn't apply edits until you do.

Full first-time setup on each OS is already written up in setting up a datacenter proxy on Windows and configuring a datacenter proxy on a Mac. Here you only need to find and correct the wrong value.

After saving, go back to chrome://net-internals/#proxy and click Re-apply settings, then Clear bad proxies. Chrome keeps a list called "Recently failed proxies, marked as bad" with a retry time for each, and clearing it stops Chrome from skipping a proxy you've just fixed.

Extensions. Proxy switchers, free VPN extensions and some privacy add-ons use Chrome's extension proxy API. When one is active, chrome://settings/system names it, and the settings page shows "[extension name] is controlling this setting" with a link to disable it. A switcher that still points at a trial endpoint you stopped paying for is a classic cause of this error.

To test: open chrome://extensions, switch off every extension with "proxy" or "VPN" in its name or description, and reload a page. If the error goes away, turn them back on one at a time to find which one. If you rely on a switcher such as FoxyProxy, open its options and correct the saved host and port there. The OS settings won't help, because the extension is overriding them. How to set one up correctly is covered in configuring a proxy in Chrome and Firefox.

Policy. Open chrome://policy and search the page for "Proxy". If you see ProxySettings, ProxyMode, ProxyServer or ProxyPacUrl with a status of OK, the device or browser is managed. According to the Chromium policy definitions, when the proxy policy is set, Chrome "ignore[s] all proxy-related options specified from the command line," and you can't change it from settings either. On a work laptop that's your IT team's proxy, and the fix is a ticket to them. On a personal machine, an unexpected proxy policy is worth investigating as possible adware, since legitimate home software rarely sets one.

Strictly, these don't produce ERR_PROXY_CONNECTION_FAILED. They matter here for two reasons: people misread the code at the bottom of the error page, and fixing the host and port often uncovers one of these next.

Protocol mismatch. The scheme must match the port. In our test, sending HTTP to SparkProxy's SOCKS5 port produced ERR_EMPTY_RESPONSE, and sending SOCKS5 to the HTTP port produced ERR_SOCKS_CONNECTION_FAILED after about 15 seconds. If you see either, check which field you typed the port into. HTTP and HTTPS fields take 11000; the SOCKS field takes 13000. The difference between the protocols is explained in HTTP, HTTPS and SOCKS5 proxy protocols.

Auth method. SparkProxy supports two methods, IP whitelist and username/password. They fail differently:

• Username and password. Chrome's built-in path has no field for credentials, so Chrome shows the Sign in prompt. Wrong details bring the prompt back. You never get ERR_PROXY_CONNECTION_FAILED from this.
• IP whitelist. The proxy checks your public IP. If your home IP changed overnight, the proxy refuses you with a 407, which again shows as a prompt or a 407-type failure, not the connection error.

From the command line the two are easy to tell apart. With curl 8.21.0 on Windows, a missing or wrong password returned curl: (7) CONNECT tunnel failed, response 407, while a dead proxy returned Could not connect to server. If you want to stop the prompt for good, whitelist your current IP in the dashboard; the trade-offs are covered in how proxy authentication works.

Be clear about who owns which part. A support ticket for the wrong cause wastes a day.

Notice that most rows land on your side. If Step 1 passed from your machine, the SparkProxy gateway was reachable and the cause is local configuration. If Step 1 failed on every network you tried, send support the host, port and the exact test output.

If you only needed the proxy for a test, or you want the machine back to how it was, undo each layer you touched. A half-removed proxy is how this error returns a week later.

1. Windows: Settings > Network & internet > Proxy. Turn off Use a proxy server and Use setup script. Leave Automatically detect settings as it was before you started; your Step 2 screenshot is the reference.
2. macOS: System Settings > Network > your service > Details > Proxies. Turn off every proxy toggle you enabled, then click OK. Repeat for each network service you edited.
3. Extensions: In the switcher, choose its direct or disabled mode, or remove the extension from chrome://extensions.
4. Launch flags: If you made a desktop shortcut with --proxy-server, delete it or remove the flag from its Target field.
5. Chrome's cache of failures: In chrome://net-internals/#proxy, click Re-apply settings and Clear bad proxies. The effective settings line should now read direct, or match what you set.
6. Dashboard: If you whitelisted an IP for a one-off test, remove it so a later occupant of that IP can't use your plan.

Compare the result with the screenshot you took in Step 2. When Effective proxy settings matches what you intended, you're done.

Want to try a clean setup with a working proxy afterwards? SparkProxy has a 24-hour free trial with no credit card, plans starting from $75/month with unlimited bandwidth, and both IP whitelist and user:pass auth over HTTP/HTTPS or SOCKS5. Pick Random Rotation for general browsing, or Sticky Sessions when a site needs the same IP for 5 minutes.