Clash / mihomo
Clash FAQ
First, identify whether the issue is with configuration loading, traffic interception, or rule matching. Choose a section based on the symptoms, then use the update log or connections list to find where things went wrong.
Basics
Understand the differences between the client, core, configuration, and traffic entry points. Choosing a mode in the interface doesn’t guarantee that traffic has reached the core.
What’s the difference between a Clash client and the mihomo core?
The client imports configurations, switches policies, and displays connections. The mihomo core handles incoming traffic, matches rules, and processes DNS. Their version numbers are independent. If a menu is missing, check the client documentation first. If a configuration field isn’t recognized, review the startup log and verify which core version is actually loaded. Some clients update the core separately, so record both versions when troubleshooting. Before switching clients, make sure they support your subscription format and protocols, then migrate your configuration.
What’s the difference between a subscription link, config.yaml, and a proxy node?
A subscription link is an address used to fetch a configuration or a list of nodes. A complete config.yaml can also include ports, DNS settings, rules, and proxy groups. A node is a single optional outbound connection; it cannot replace the full configuration. If an import shows nodes but none of the expected rules, first check whether the link returns a complete Clash configuration or only a node list. Then decide whether to use the provider’s configuration or add the nodes to an existing one. Subscription URLs often contain account identifiers, so don’t share them in screenshots or public logs.
Which mode should you use: Rule, Global, or Direct?
For everyday use, start with Rule mode, where the core selects an outbound connection according to the order of rules in your configuration. Global mode sends traffic entering the core to the GLOBAL policy group and is useful for brief comparison tests; after switching, check which outbound connection that group has selected. Direct mode bypasses proxy exits for traffic entering the core, but doesn’t stop traffic interception. To compare behavior with the client on and off, disable the relevant system proxy or TUN, then start a new connection.
How do system proxy and TUN differ? Should you enable both?
A system proxy uses the operating system’s proxy settings to route requests through the client, covering only apps that follow those settings. TUN takes over traffic through a virtual network interface and routes, making it suitable for apps that ignore the system proxy, but it depends on system permissions and routing settings. When setting things up, enable just one method at first and check the connections list to confirm the target request reaches the core. If problems arise with both enabled, turn one off to compare results. Don’t change DNS, the outbound connection, and routing rules all at once.
Setup
If an import fails, check the subscription update log first. If startup fails, check the core log. These errors point to different parts of the process.
Importing a subscriptionSubscription updates return 401, 403, or 404. How can you tell if a subscription has expired?
Check the request result in the update log. A 401 usually means authentication is required or the credentials weren’t accepted. A 403 means the request was denied, possibly because the provider restricts its source. A 404 may mean the URL has changed. Check your subscription status and supported format in the provider’s account page, then copy a new Clash or mihomo subscription URL. Keep the old configuration until the new one has been verified. If the log shows only a timeout and no HTTP status code, troubleshoot network connectivity before resetting the subscription.
Why is the node list still empty after a successful configuration import?
First, make sure the imported configuration is set as the active one. Then check the startup log for YAML parsing errors, unsupported fields, or references to missing proxy groups. If you use proxy-providers, a successful import only means the main configuration was read; remote nodes still need to be fetched separately. Check the provider’s update results in the client and make sure the response isn’t a login or error page. If nodes were fetched but the proxy group is still empty, check its use references and filter conditions.
TUN reports insufficient permissions. What should you check?
On Windows, check the client documentation to see whether service mode must be installed and enabled, then make sure the service is running. If elevation is required, grant it only for the specific installation or startup step. On Android, confirm that the client is authorized in the system VPN settings and check whether another VPN is already using the connection. If TUN still fails, note any device or routing errors in the log and check the relevant settings supported by the client. Don’t disable system security features as a first step.
Startup says mixed-port is already in use. Can you just change the port?
You can, but first check whether another instance is using the port. Common causes include an old client still running in the system tray or both a GUI client and a standalone core running at once. If you find a duplicate, exit it normally and restart the client. If both instances need to run, assign each a different local port and update any apps where you manually configured a proxy address. Finally, make sure the system proxy points to the current instance so requests aren’t still being sent to the old port even though the interface shows the client as running.
Tips
Note the current outbound connection and matched rule. Change one setting at a time, then start a new connection to compare results.
Local overrides and mergesDoes a Clash node that times out in a speed test have to be unusable?
Not necessarily. A speed test records the result of using a node to reach a specific test address; it doesn’t represent every destination. Compare it with other nodes in the same group, then test the site you actually need and confirm the selected outbound connection in the connections list. If only one test address fails, check whether that address is reachable. If one node consistently fails for several destinations, troubleshoot the node or check the provider’s announcements. Don’t delete a configuration based on a single test; retesting on the same network makes it easier to isolate the issue.
Why does Rule mode send a site to DIRECT when you expect it to use a proxy?
Rules are evaluated in order, and the first match determines the policy. Open the request in the connections list and note the matched rule and final outbound connection. If a broad direct rule matched, move a more specific rule above it. If the expected rule matched but the request still went direct, check whether its proxy group has selected DIRECT. If the log shows only an IP address and no domain, first check whether the core can obtain domain information rather than widening the match. Close the existing connection after making changes, then send a new request.
Why did my local rule changes disappear after updating the subscription?
When a client refreshes a subscription, it may replace the local copy of the remote configuration with new content, wiping out direct edits. Save custom rules in a separate override or merged configuration, then bind it to the subscription as described in the client documentation. Rules that need to take precedence should go near the top of the rules list, not after MATCH. After saving, inspect the active configuration to confirm the rules are present and correctly ordered and their proxy group references are valid. Then refresh the subscription manually to verify the changes persist.
TUN blocks access to a printer or NAS. How can you restore access?
Turn off TUN first and check whether local network access returns. Note the device’s actual IP address and subnet. After turning TUN back on, try accessing the device by IP. If the IP works but the device name doesn’t, check local DNS or device discovery first. If the IP also fails, check for overlapping routes and exclude only the local network subnet you actually use, using a method supported by the client. Test local devices and external sites separately after making changes. Don’t exclude every private IP range at once.
Troubleshooting
If the connections list is empty, first check how the app connects. If it has entries, check the rules, outbound connection, and DNS.
DNS configuration guideSystem proxy is on, but an app isn’t using Clash. How do you troubleshoot it?
Open a new browser tab and check the client’s connections list. If the browser appears there but the target app doesn’t, the app probably ignores the system proxy or uses its own settings. Check the app’s proxy options or compare results with TUN enabled. If neither appears, verify the system proxy address, listening port, and core status, and check whether the system is still using an old auto-proxy configuration. If the connection is listed but goes direct, check the matched rule and proxy group.
The browser works, but a Windows UWP app can’t use the local proxy. What should you do?
If the app uses a system proxy pointing to 127.0.0.1, first confirm that a regular desktop app can use the same port. Consider AppContainer’s local loopback restrictions only if specific UWP apps fail. If the client provides a UWP loopback setting, follow its instructions to enable the exemption only for affected apps, then fully exit and restart them. If you’re unsure whether to change this setting, test TUN first if your system permits it. On managed devices, follow your organization’s policies.
How do you troubleshoot a Clash DNS leak or unexpected DNS resolution path?
First, identify which upstream resolver you expect to use and which domains should resolve through local DNS. Temporarily use just one traffic interception method, note the affected domains and test time, then compare the client log with your system DNS settings. Secure DNS in a browser may use a different path, so compare results with it on and off on the same network. A test page showing the resolver’s location doesn’t prove a leak on its own; base your assessment on the actual request path and your configuration’s intended behavior.
How can you restore internet access if it still doesn’t work after quitting Clash?
Reopen the client, turn off system proxy and TUN in its interface, then exit normally. If the client won’t start, check the system network settings to see whether the proxy still points to a local port that’s no longer listening. Revert only the settings enabled for this session; keep any proxy configured by your workplace or school. If you changed DNS manually, restore its previous value, then reconnect and test your network. Don’t reset all network settings, as you could lose existing static IP addresses or internal network configuration.