When Proxifier does not work, separate the failure into one of four layers: the proxy endpoint, the Proxifier rule, the application process, or name resolution. Changing all four at once makes the log harder to interpret and can introduce a connection loop.
This workflow follows the official Proxifier v4 for Windows documentation. Check the installed edition before applying platform-specific menu instructions. If Proxifier already shows an error message or a code such as 10061, start with the error message table.
Fastest triage
Proxy Checker first; application rule second; route verification last.
If Proxy Checker fails, fix the endpoint before editing application rules. If Proxy Checker passes but the app remains direct, inspect rule order and the actual process in Proxifier's Connections and Output panes.
Use the symptom to choose the first check
| Symptom | Most likely layer | First action |
|---|---|---|
| Proxy Checker fails | Address, port, protocol, authentication, network, or proxy health | Test the endpoint outside application rules |
| Proxy Checker passes but app shows the direct IP | Rule order or process mismatch | Inspect Connections and Output |
| No connection appears in Connections | Observation settings, process identity, or no new request | Check the Direct connection visibility limits before concluding that no connection occurred |
| App cannot connect through one protocol | Protocol capability | Compare SOCKS5, HTTPS CONNECT, and plain HTTP support |
| Hostnames fail but direct IPs work | Name resolution | Review Profile → Name Resolution |
| Everything slows or stops after adding a local proxy | Connection loop | Keep loop detection enabled and exclude the local proxy process |
| Only one helper or updater stays direct | Child process not covered by the rule | Add the exact process only when required |
Find your Proxifier error message
Match what Proxifier shows in its log, proxy list or dialogs, then jump to the fix. Quoted messages are Proxifier’s own wording from its documentation; numbered errors are the standard Windows network (Winsock) codes that Proxifier’s error reference lists.
| What you see | What it means | Fix |
|---|---|---|
10060 connection timed out, or a proxy marked [failed] | The proxy did not answer in time: wrong host or port, a firewall, an overloaded proxy, or a timeout set too short. | Re-test in Proxy Checker (Step 1). Check the address and port, and raise Connection to proxy timeout if the proxy is slow; use Reset Fail Status for All Proxies after fixing it. |
10061 connection refused | The host answered but nothing accepted the connection on that port, or it rejected your address. | Confirm the port and proxy type with the provider. Many providers only accept whitelisted source IPs, so add your current public IP from IP Lookup or switch to username and password. |
10054 connection reset, or 10053 connection aborted | The connection opened and was then closed: often a protocol mismatch (SOCKS5 set as HTTPS or the reverse), rejected credentials or a provider limit. | Check the protocol in Step 2 and the credentials below. Compare SOCKS5 and HTTPS if only one type works. |
| A proxy login prompt keeps appearing, and connections wait | The proxy rejected the saved login or password. While the prompt is open, Proxifier holds all connections; choosing Ignore makes authentication fail silently each time. | Enter the correct credentials and use Save to Profile. For HTTP 407 Proxy Authentication Required, follow the 407 checklist. |
| An HTTP proxy works in the browser but HTTPS sites fail through Proxifier | Proxifier’s documentation gives the cause as “SSL support is most likely unavailable”: many HTTP proxies only allow HTTPS tunnels to port 443. | Add the proxy as type HTTPS only if it supports SSL tunnelling to any port, or use SOCKS5. In Chrome the same failure shows as ERR_TUNNEL_CONNECTION_FAILED. |
| “Infinite loop connection detected. All new connections will be blocked automatically till the user action.” | Proxifier is capturing a local proxy, tunnel or antivirus that sends the traffic straight back out. | Add a Direct rule for that program and turn off Handle Direct Connections if it is on (Step 9). Keep loop detection enabled. |
| “Name resolution over proxy has been automatically disabled to prevent infinite loop.” | Resolving hostnames through the proxy looped back through Proxifier. | Review Profile → Name Resolution (Step 6) and check for leaks with the Proxifier DNS leak fix. |
| An IP-based rule stops matching after enabling Resolve hostnames through proxy | Names resolved through the proxy get placeholder addresses in 127.8.*.*, so rules written for real IPs no longer match. | Match by hostname or application instead (Step 5, and rule examples). |
| No error, but the app still shows your real IP | The connection matched the Default or a Direct rule, or a different process made the request. | Read the matched rule in Connections (Step 3) and identify the real process (Step 4). |
| “Proxifier drive is not running” | Proxifier’s network driver did not start, so nothing can be proxified. | Reinstall Proxifier with administrator rights; only an administrator can install or start its driver. |
Step 1: test the proxy independently
Open View → Proxy Checker or use the Check button in Proxy Server Settings. The official checker can test SOCKS4(A), SOCKS5, HTTPS, and HTTP endpoints and measure response time.
For the first test, leave Work through Proxifier disabled. That is the documented default. Enabling it causes the checker itself to follow Proxifier and is intended only for a specific scenario, such as a proxy that is not directly reachable.
A pass covers the enabled tests against the checker's configured target and port. It does not prove that another application's process, destination, or rule uses that proxy. Keep the checker result separate from the application observation below.
Verify:
- Hostname or IP address.
- Port.
- Protocol selected in Proxifier.
- Username and password or enterprise authentication.
- Source-IP allowlist.
- Target host and port used by the checker.
- Local firewall and network reachability.
Step 2: confirm the proxy protocol
A common failure occurs when an HTTP proxy is added as HTTPS. Proxifier's documentation distinguishes plain HTTP support from HTTPS/CONNECT support:
- A plain HTTP proxy can process HTTP connections.
- A proxy that supports CONNECT on arbitrary ports can be configured as HTTPS.
- Many browser-compatible HTTP proxies allow HTTPS only on port 443 and cannot tunnel arbitrary application TCP connections.
- SOCKS5 is designed for general proxy routing and supports username/password authentication.
If a proxy works in a browser but fails in Proxifier, test whether it actually supports the selected protocol. Read SOCKS5 vs HTTPS in Proxifier before changing the application rule.
Step 3: read the matched rule in the Connections pane
Open the Connections pane through View → Windows, then select Log → Screen Log → Verbose. For one new request, compare Application and its process ID, Target, Time/Status, and Rule: Proxy. The Windows v4 interface reference explains these fields: an active connection has no status label, while a failed connection is marked as failed. A closed connection alone does not establish a routing failure.
Proxifier evaluates rules from top to bottom. The first matching rule determines the action.
If the connection is missing from the list
Direct-action connections are included in the list when Profile → Advanced → Handle Direct Connections is enabled. If it is off, the absence of a Direct row does not prove that the application made no connection. First reproduce a new request and confirm the process; do not change the proxy endpoint just to make a row appear. See Proxifier's Direct Connections documentation.
Check for these mistakes:
- A broad Direct rule appears above the intended proxy rule.
- The Default rule uses the proxy when it should be Direct.
- The application path does not match the executable that opened the connection.
- The target host or port criterion is narrower than intended.
- The rule is disabled.
- The application was started with manual proxification, which ignores ordinary rules for that process instance.
Step 4: identify the real process
Applications often use more than one executable:
- Launchers start the main process.
- Browsers use multiple processes.
- Updaters connect separately.
- Meeting and chat apps use helper processes.
- Python projects can use a virtual-environment interpreter instead of the system executable.
- Command-line shells start separate network tools.
Use the Application/PID observation to identify the process, then confirm its executable path before editing a rule. A displayed name such as python.exe is not enough to distinguish two installations. If the instance was launched through the executable's Proxifier context menu, manual proxification overrides ordinary rules for that instance; restart it normally when testing an ordinary rule.
Worked example: the Python rule matches the wrong interpreter
Illustrative configuration, not a captured log or a performed Aerod test. Assume Proxy Checker passes, no earlier rule matches, targets and ports are Any, and Default is Direct:
- The intended proxy rule matches
C:\Python\python.exe. - The project actually launches
C:\Projects\Example\.venv\Scripts\python.exe.
Those paths differ, so this example's request would miss the intended rule and fall through to Default Direct. Changing credentials or DNS would not correct that mismatch.
- Confirm the interpreter path used by your own project; the paths above are examples, not values to paste.
- Change only that rule's application path to the confirmed interpreter. Leave its verified proxy and other criteria unchanged.
- Restart the target process normally and make one permitted request.
- Check the new process ID and whether the intended rule/proxy now appears. Verify the requesting process's route as described in Step 10. A different match, error, or absent row remains something to investigate, not a successful result.
The Proxifier rule examples explain exact-path matching for Python projects and other applications.
Step 5: verify the rule criteria
A connection must satisfy every populated rule criterion. If Applications, Target hosts, and Target ports all contain values, all three must match.
During diagnosis:
- Create one exact application rule.
- Leave Target hosts as Any.
- Leave Target ports as Any.
- Set the action to the verified proxy.
- Keep the rule above Default.
- Keep Default direct.
After the route works, narrow host and port criteria one at a time.
This pattern is selective routing, not fail-closed isolation: an unmatched application or destination is deliberately allowed to connect directly by Default. Do not use a successful selected-app check as proof that all other traffic is blocked.
If your intended policy is different, the Default Block guide explains a separate rule pattern, its explicit Direct exceptions, and the traffic it cannot claim to protect.
Step 6: check name resolution
When hostnames fail but numeric addresses work, review Profile → Name Resolution. Automatic detection is enabled by default. It can switch to proxy-side resolution if the system DNS service is unavailable.
Proxy-side resolution introduces placeholder addresses such as 127.8.*.*. IP-based rules no longer see the real destination address on the local computer. Use hostname-based rules and Verbose logging when that mode is active.
Read Proxifier DNS Leak Fix for a full local-versus-proxy resolution workflow.
Optional SOCKS5 comparison: which side resolves the name?
If the endpoint is SOCKS5, compare the same destination with curl resolving its hostname locally and with curl handing the hostname to the proxy. This checks the endpoint independently; it does not exercise Proxifier's rules. Use a permitted diagnostic environment where Proxifier or another transparent forwarding layer is not intercepting curl. Do not disable an enforced network policy to run it; use Proxy Checker instead if you cannot isolate the request.
Replace PROXY_IP:PORT with your provider's numeric proxy address and port. Run each line separately in Windows PowerShell:
curl.exe -q --verbose --noproxy never-bypass.invalid --connect-timeout 5 --max-time 15 --socks5 PROXY_IP:PORT https://example.com/ --output NUL
curl.exe -q --verbose --noproxy never-bypass.invalid --connect-timeout 5 --max-time 15 --socks5-hostname PROXY_IP:PORT https://example.com/ --output NUL
The curl SOCKS guide distinguishes the two modes: --socks5 resolves the destination locally; --socks5-hostname sends its name to the proxy. Using a numeric proxy address keeps resolving the proxy's own hostname out of this comparison.
-q comes first to ignore curl's default configuration file. The nonmatching --noproxy entry prevents a saved NO_PROXY environment setting from bypassing the proxy for example.com; it is not a server to contact. Time limits keep a failed request bounded. On macOS or Linux use curl and /dev/null instead of curl.exe and NUL. These examples contain no credentials. If authentication is required, use your approved secure curl configuration or Proxy Checker; do not paste passwords into commands or shared logs. See the curl option reference.
| Observed result | What to investigate next |
|---|---|
| Local mode cannot resolve the destination; hostname mode completes the HTTPS request | Investigate local destination-name resolution before changing credentials or application rules. |
| Local mode completes; hostname mode is rejected or fails | Investigate the proxy's hostname handling, resolver, or destination policy. The difference alone does not identify the cause. |
| Both fail before SOCKS negotiation | Check proxy reachability and the numeric address/port first. A timeout is not a DNS diagnosis. |
| Both complete | Return to the actual application's process, matched rule, and Proxifier DNS log; the endpoint comparison has not verified them. |
A SOCKS rejection, authentication error, TLS failure, or HTTP error is not a successful destination check. Keep certificate verification enabled. Neither mode identifies the recursive resolver or proves that all applications avoid DNS leaks.
Reproduction boundary: On October 26, 2026, Aerod checked these two curl modes against a loopback SOCKS5 receiver. Local mode sent a numeric destination; hostname mode sent localhost. The receiver deliberately refused both connections. This verifies the request-format distinction, not a working external proxy, TLS transfer, or Proxifier installation.
Step 7: inspect the Output log
The Output pane supports four levels:
| Level | Use |
|---|---|
| Error Only | Program-critical and connection errors |
| Normal | Recommended ordinary view with connection open and close events |
| Verbose | Rule processing, DNS requests, and additional diagnostics |
| Debug | Extensive diagnostic output for deeper troubleshooting |
Start with Log → Screen Log → Verbose; screen logging and file logging are separate controls. If an intermittent problem needs a saved log, choose the necessary level under Log → File Log. Proxifier can also create Debug and Traffic output, but traffic dumps can consume substantial disk space and may contain sensitive application data. Enable them only for a controlled diagnostic session, redact sensitive details before sharing, and restore the previous logging levels afterward.
Step 8: diagnose timeouts and refusals
Connection timeout
The connection was not established within the allowed time. Possible causes include an unreachable proxy, filtered port, overloaded endpoint, wrong network path, slow chain, or unavailable target.
Connection refused
A reachable host actively rejected the connection. Confirm the port and service, then check whether the endpoint is listening and whether the source is authorised.
Authentication failure
Confirm the username, password, authentication method, and whether the provider requires source-IP authorisation. HTTPS corporate proxies can use Basic, NTLM, Kerberos, or current-user credentials depending on configuration.
If the response specifically says 407 Proxy Authentication Required, follow the proxy authentication troubleshooting checklist to separate the proxy challenge from a destination website login.
Target failure after proxy connection succeeds
The proxy can be reachable while the target is blocked, unavailable, disallowed by provider policy, or unsupported by the selected protocol.
If the same proxy also fails in Chrome with ERR_TUNNEL_CONNECTION_FAILED, read its CONNECT response code as shown in how to fix ERR_TUNNEL_CONNECTION_FAILED.
Step 9: prevent infinite connection loops
A local proxy, antivirus, tunnel, or forwarding program can create a loop:
- The application connects.
- Proxifier redirects it to the local proxy.
- The local proxy opens its outbound connection.
- Proxifier captures that connection and redirects it back to the local proxy.
- The cycle repeats.
Proxifier includes Infinite Connection Loop Detection and Infinite Name Resolution Loop Detection. Keep both enabled unless you are resolving a documented false positive.
To fix a loop:
- Add a Direct rule for the local proxy or tunnel process.
- Do not proxify the local proxy's own outbound connection.
- Disable Handle Direct Connections in the local proxy workflow when the official documentation requires it.
- Disable proxy-side name resolution if a DNS loop is the trigger.
- Restart affected applications after changing the route.
Step 10: verify the corrected route
- Clear or save the current log.
- Restart the target application.
- Confirm the intended rule and proxy in Connections.
- For a browser, open IP Lookup in that same routed browser. For a script or other client, verify its own permitted request and, where supported, its public egress address; a lookup in an unrelated browser does not verify that client's route.
- Compare IPv4, IPv6, ASN, organization, and approximate location.
- For a browser, run WebRTC Leak Test and the Browser Permissions & Privacy Settings Check.
- Review DNS separately.
- Restore any temporary direct-handling and logging settings. Keep loop detection enabled.
If you are weighing a different tool instead, compare Proxifier alternatives by per-app rules, UDP and DNS behavior, then verify any replacement with the same route checks.
Escalation checklist
Collect enough information for support
- Proxifier version and edition.
- Windows version.
- Proxy address redacted as needed, port, and protocol.
- Proxy Checker result.
- Rule order.
- Actual executable path and process ID.
- Target hostname and port.
- Name-resolution mode.
- Relevant Verbose or Debug log lines.
- Whether a local proxy, VPN, antivirus, or tunnel is also active.
FAQ
Why does Proxy Checker pass but my application stays direct?
A pass covers the checker's enabled tests and configured target, not another application's route. Check the exact process, matching rule, and manual proxification. If a Direct connection is absent from the list, check the conditional visibility setting before drawing a conclusion.
Does an empty Connections pane mean the application made no connection?
No. Direct-action traffic is listed when Handle Direct Connections is enabled. Reproduce a new request and identify the process first. Only use temporary direct handling when appropriate; local forwarding and loop setups require the documented exception, not a blanket recommendation to enable it.
Why does a proxy work in a browser but fail in Proxifier?
The server may be a plain HTTP proxy that handles browser HTTP or standard HTTPS ports but does not support arbitrary CONNECT tunneling. Add it with the correct protocol and test it with Proxy Checker.
Why does Proxifier show connection timeout?
The client could not establish the tested connection within the allowed time. Check address, port, protocol, firewall, upstream reachability, proxy health, and target availability.
Why does the wrong process bypass my rule?
Launchers, helpers, updaters, browsers, and child processes can use different executables. Use the Connections and Output panes to identify the process that actually creates the connection.
What causes an infinite connection loop in Proxifier?
A local proxy, tunnel, antivirus, or forwarding program can have its outbound connection captured and redirected back into itself. Exclude that process with a Direct rule and keep loop detection enabled.
What does connection refused (10061) mean in Proxifier?
Error 10061 means the proxy host answered but refused the connection on that port. Check the port and proxy type, and whether the provider only accepts whitelisted source IPs; if so, add your current public IP or switch to username and password authentication.
Why does Proxifier keep asking for my proxy login?
The proxy rejected the saved username or password. Proxifier holds connections while the login prompt is open, and choosing Ignore makes authentication fail silently. Enter the correct credentials and save them to the profile.
What log level should I use for troubleshooting?
Use Log → Screen Log → Verbose for rule processing and DNS requests. File logging is configured separately under Log → File Log. Use Debug only when needed, protect sensitive details, and restore previous diagnostic settings afterward.
Sources checked
- Proxifier: Proxy Checker
- Proxifier: Proxy Server Settings
- Proxifier: Proxification Rules
- Proxifier: User Interface and Output log
- Proxifier: Log Files
- Proxifier: Name Resolution
- Proxifier: Connection Loop Problem
- Proxifier: HTTP Proxy Support
- Proxifier: Direct Connections
- Proxifier: Manual Proxification
- Proxifier: Proxy Redundancy and failed proxies
- Proxifier: Advanced proxy settings and authentication
- Proxifier: Installation and driver
- Proxifier: Network error codes
- everything curl: SOCKS proxy
- curl option reference