Proxy tab: Proxy groups, nodes, and latency tests
Use the Proxy tab to choose an outbound route. In the Clash Verge Rev 2.x desktop app, open Proxy from the sidebar, find a proxy group, then check which option is selected. Groups are usually defined in the configuration and may be named “Manual,” “Auto,” or something custom. A group can list nodes, reference another group, or include the built-in DIRECT option. Names are just labels; check the group type to understand where traffic will go.
Selected options in manual and automatic groups
selectis a manual selection group. Choose a member to route new connections using this group through that option. If you select a nested group, the nested group determines the final route.url-testselects a member based on test results. Iftoleranceis configured, a slightly lower latency may not trigger a switch.fallbackselects the first available member in the configured order, keeping the earliest option that passes its health check. It does not sort members by lowest latency.
Latency readings, timeouts, and mode switches
The test button measures latency to a specified test URL through a candidate route; it does not measure download speed. For example, 86 ms means this test took about 86 milliseconds. Timeout means the test did not succeed within the time limit. If the test URL is unreachable, several nodes may time out at once. Compare nodes using the same test URL, then verify real-world performance in the app you intend to use.
After checking the nodes, confirm the mode. Rule mode matches rules from top to bottom; the first match determines the route. Selecting a node in a manual group only affects requests routed through that group. Global mode uses the current selection in the global group, so selections in other manual groups may not apply. Direct mode sends traffic handled by the core straight to its destination. None of these modes routes traffic into the core; changing modes does not enable the system proxy or TUN.
Profiles tab: Import, update, and switch are separate actions
A profile card in the list only means the client has a record for that profile. In the Clash Verge Rev 2.x desktop app, for example, open Subscriptions in the sidebar to import a remote subscription. Other desktop clients may call this page “Profiles” or “Config.” Before importing, make sure the source provides a configuration compatible with your core, not a subscription format that needs separate conversion.
| Action | What it changes | Where to check |
|---|---|---|
| Import | Adds a remote subscription URL or local YAML file to the profile list. | Check that the new entry appears and look for any format errors. |
| Update | Fetches the remote subscription again. Edit or reimport local files to apply changes. | Check the update result. A successful fetch does not necessarily mean the profile is active. |
| Switch | Sets the selected profile as active and loads it into the core. | Check that the active profile is highlighted or checked, then confirm its proxy groups appear on the Proxy tab. |
Subscription fetched, but expected proxy groups are missing
First check which profile is active, especially if the list contains subscriptions with the same name. Then distinguish between two types of errors: HTTP 401 or 403 points to subscription retrieval, so check the URL, authorization, and any notes from the provider. YAML parsing errors or references to nonexistent proxy groups occur while reading or loading the configuration; check the field named in the error. An HTTP 200 response only means a response was received—it could be a login page rather than usable YAML.
Settings and Connections: Check that app traffic reaches the core
A successful test in the Proxy tab does not mean your browser is using Clash. The core runs the test request, while regular apps must route traffic to the core through the system proxy, their own proxy settings, or TUN. For an initial check, test one way of routing traffic at a time; avoid changing the mode, node, and traffic entry method all at once.
- System proxy: Find the System Proxy toggle under Settings, then test with a browser that follows the system proxy. Some apps ignore system proxy settings. If you enter proxy settings in an app instead, make sure the address and port match the client's active listener.
- TUN: In Settings, look for Virtual Network Adapter or TUN Mode. TUN routes traffic through a virtual network adapter and routing rules, and may require system permissions or a client service. Even when enabled, check whether the target app's connection appears.
Here is an example of top-level fields for mihomo's local mixed port. 7890 is an example, not a port shared by every client; use the value shown in the client settings and active configuration. With this example, you can enter 127.0.0.1:7890 as an HTTP or SOCKS proxy in an app, selecting the protocol the app requires.
mixed-port: 7890
allow-lan: false
To verify the traffic entry method, make a new request, then find the target domain or address in the client's Connections view. If the client shows both the matched rule and the proxy chain, check the rule first, then the final route. The highlighted selection in the Proxy tab shows the current option for a group; the proxy chain in Connections shows the route for that specific request. If the test app reuses an existing connection, close it or restart the app and try again.
Logs tab: Filter errors by time and level
In the Logs tab, clear any keyword filter in the sidebar and note the time before making another request. Looking only at the last line can make an old error look like the current issue. Find entries that match the time and the target domain, port, or profile switch, then use the Connections view to see how far the request got.
infois a good choice for everyday use. It includes routine runtime information as well as warnings and errors.- Use
warningorerrorto narrow down an issue. Seeing no errors does not mean app traffic is reaching the core. - Use
debugfor brief troubleshooting sessions. Switch back toinfowhen you're done to avoid accumulating detailed domain and connection records.
The Logs tab's display filter is separate from the core's log output level. A filter only hides or shows existing entries; if the core is running at info, selecting debug in the UI will not generate debug entries. For more detail, change the core log level in the client settings or specify the top-level field in a supported configuration override, then apply the configuration and test again.
log-level: debug
Where to look for three common error types
address already in use: Wait for the error to identify the port, then find which process is using it. If you change the local proxy port, update it in the system proxy or any apps where you entered it manually.lookuperrors: First determine whether the failed lookup is for the node server's hostname or the domain the app is trying to reach. They may use different DNS resolution paths; a “timeout” alone does not tell you which DNS setting to change.i/o timeout: Check the address and port in the same log entry, then inspect the related connection and proxy chain. A timeout for the test URL, the node connection, and the destination request are different issues and should be investigated separately.
Retest the first connection across all four tabs
Retest with the same app and destination, changing only one thing at a time. This helps distinguish an unapplied configuration from traffic that never reached the core or an unavailable route—without guessing by repeatedly switching nodes.
- Activate a profile on the Subscriptions tab and wait for it to load. Open Proxy and confirm that its groups appear, not those from the previous profile.
- Set the proxy mode to Rule. If you know the destination's rule points to a manual group, select a member in that group that passes the test. If the rule points to
DIRECT, changing the group's node will not affect the result. - Make sure your chosen entry method—system proxy, app proxy, or TUN—is active. Open the target app, find the new request in Connections, and check the matched rule and actual route.
- If no new connection appears, check whether the app is using the selected entry method. If a connection appears but takes the wrong route, check the mode and rule order. If the route is correct but the request fails, use the timestamp to find the relevant stage in the logs.
A successful subscription update, a latency reading for a node, and a successful visit to a destination are three separate checks. They answer whether the configuration was fetched, whether the test passed, and whether the request completed as expected. None can stand in for another.