Check the active configuration and proxy group
Seeing a subscription in the list only means the client has saved that configuration. In the desktop client Clash Verge Rev, first check that the intended profile is selected on the “Subscriptions” page, then open “Proxies” to see whether its groups and nodes are available. Clash Verge Rev uses the mihomo core; other Clash clients may call the same section “Profiles” or “Configuration.” Menu names and locations vary by client.
- Keep Rule mode enabled. Rule mode routes each connection according to your rules. Direct mode does not use proxy nodes for regular connections, while Global mode uses the global selection instead of the groups chosen under Rule mode.
- Know the difference between nodes and groups. A node is a specific proxy outbound. A manual selection group lets you choose an entry, but that entry may itself be another group. For example, if “Proxy” is set to “Auto,” the actual node is determined by the latter. Group names are set by the configuration author and may not include “Select Node.”
- Choose a node in the group used by the target rule. If the test domain routes through the “Proxy” group, check the selection in that group or a subgroup it references. Changing a different group—even one containing the same node—won’t change the outbound used by this rule.
Run a latency test: check for successful results before comparing times
On the “Proxies” page, use the speed-test control next to a node to test it individually. To compare nodes, run a group-wide test. The core sends a request to the specified test URL; this is not an ICMP ping to the server or a bandwidth download test. If your configuration uses a url-test group for automatic selection, its health-check URL and interval may differ from the manual test in the interface.
- Keep the test conditions consistent. Run two or three tests on the same network with the same test URL, then try a real request in the app you want to use. Results from different test URLs aren’t directly comparable because the route from the test destination to the outbound changes.
- Rule out nodes that repeatedly time out first. For example, one node might return 180 ms, 195 ms, and 186 ms across three tests, while another returns 120 ms once and times out twice. For troubleshooting your first connection, the first node is a better starting point. These figures illustrate how to assess results; they are not measurements from this site.
- If every test times out, it doesn’t mean every node is down. First check that your local network is working, then review the test URL, active configuration, and core errors. If the test destination is unreachable, an entire group may show the same result. If you can change the test URL, change only that setting per run so you can identify what caused the results to change.
The core runs latency tests directly, so they can return normal results even when the system proxy is off. A successful test only shows that the core made a request to the test address through the tested node; it doesn’t prove your browser has handed traffic to the client.
Enable the system proxy: set up a clear test path
On desktop, verify your browser with the system proxy first; don’t enable TUN at the same time. In Clash Verge Rev, for example, find the “System Proxy” toggle under “Settings.” This lets apps that respect system proxy settings use the local endpoint provided by the client; it doesn’t force every program to use it. On a managed device, make sure you’re allowed to change proxy settings and don’t overwrite existing configuration set by your organization.
For comparison, on Windows 11 go to “Settings” → “Network & internet” → “Proxy” and check whether the manual proxy points to the local client. The address is commonly 127.0.0.1; the port must match the active listener. For example, if the client shows a mixed port of 7897 and the system proxy uses that mixed endpoint, the system setting should be 127.0.0.1:7897. If the client uses a separate HTTP proxy port, check that port instead. Don’t copy the example blindly or use the controller port as the proxy port.
- Test one app first. Use a browser that follows system proxy settings. If a browser extension sets its own proxy, rule out its effect on this test first.
- Verify with a new connection. Existing connections may keep using the previous outbound after you switch nodes. If the result doesn’t change, quit and reopen the browser, then revisit the test page to avoid reusing an old connection.
- Desktop and mobile setups differ. The system proxy toggle in this section is for desktop. If an Android client routes traffic through the system VPN, check VPN permission, session status, and app coverage instead of following the Windows manual proxy steps.
Verify proxy routing with connection records and your exit IP
First, check that the request reached the client and see which rule it matched
Open the client’s “Connections” page, then visit your chosen test site in a browser. Find the matching connection by domain and start time so you don’t mistake another domain loaded by the page for your test request. Check the matched rule and outbound chain: the rule explains why that policy was selected, and the outbound chain shows which node or DIRECT the group ultimately used. Short-lived requests may disappear from the active connections list. If so, revisit the site or check the “Logs” for the same time; don’t conclude that no request was made just because it’s no longer listed.
Then compare the public IP shown by the same test site
On the same network, turn off the system proxy and keep TUN off, then visit the same IP-check site and note the direct-connection result. Turn the system proxy back on, visit the site using a new connection, and check the outbound chain for the test request. Compare the same address family, such as IPv4 with IPv4. Don’t compare a direct IPv6 address with a proxied IPv4 address, and don’t rely only on the country or city reported by the site.
If the check request matches DIRECT, an unchanged IP is consistent with the current rules and doesn’t mean the proxy toggle failed. Test another destination known to use a proxy group, or, if supported, temporarily switch to Global mode and explicitly select the node you’re testing—but only for this request. Switch back to Rule mode afterward and check your usual routing. A changed IP only shows that this test request used a different outbound; it doesn’t mean every app or domain is routed through the proxy.
Latency tests pass but websites won’t load: troubleshoot by connection symptoms
| Symptom | Check first |
|---|---|
| Latency test returns a result, but no matching record appears for a new request | Check whether the browser uses the system proxy, whether the destination is on the proxy bypass list, and whether the local listening port matches the system settings. |
| The request reaches the client but goes direct | Check the matched rule, proxy mode, and group selection. In Rule mode, an earlier matching rule may already determine the outbound. |
| Outbound chain shows a proxy node, but the website times out | Check the logs from the same time to distinguish a DNS resolution error from a connection timeout. Keep the test destination the same and retry with one different node that passed the latency test. |
| The browser works, but another app has no connection record | First check whether the app supports system proxy settings or has its own proxy settings. If you need to route its traffic through the client, assess TUN permissions, routing, and app compatibility separately. |
For an initial check, you only need one complete chain of evidence: the target app made a new request, the client recorded it, and the outbound chain points to the expected node. If the destination is an IP-check page, use the exit IP as a cross-check. If you need to troubleshoot further, note the destination domain, request time, matched rule, and outbound chain. Before sharing logs, remove subscription URLs, credentials, and unrelated traffic records.