Clash Setup Guide

This guide starts with an installed client and an existing subscription URL or local config. Follow four steps to connect, then check the route used by a specific app. Haven’t installed a client yet? Find downloads by platform, then return to step one.

Step 1: Import a Subscription and Enable the Config

Open the client and go to “Config,” “Subscription,” or “Profiles.” If your provider gave you a link, choose the option to import from a URL when adding a config. Paste the full address into the field, then select Download, Import, or Save. The query string—the question mark and everything after it—is part of the URL, so don’t remove it. If your provider gave you a local YAML file, choose the file import option on the same page and locate it in the system file picker. A local file and a remote subscription are two different sources; you don’t need to upload a local file to the internet to use it.

After importing, a new entry should appear in the config list. Seeing it in the list doesn’t mean the client is using it. Select the new entry or choose Enable from its menu, then confirm the active config is marked with a check, highlight, or changed name. Next, open the “Proxies” or “Proxy Groups” page. If the config includes policy groups and nodes, they should appear there; expand a group to see its options. It’s normal for older configs to remain in the list. Just confirm which one is active—there’s no need to delete anything to clean up the list.

If the new entry fails to import, first check where the error appears. A download timeout usually means the client couldn’t retrieve the content; check your device’s network and make sure the URL is complete. A parse error may mean the client received a response, but it isn’t a config it can read—for example, the URL returned a login page, or the provider supplied a format for another client. Check what the link is intended for on the provider’s site, and look for one explicitly labeled as Clash-compatible. Don’t paste subscription URLs into screenshots, public logs, or online parsers; they may contain access credentials.

If an error appears only after importing a local file, note the field or line number reported by the client and troubleshoot a copy of the original so you don’t overwrite your only working config. If switching configs fails, switch back to the one that worked. This helps determine whether the new file or the client is at fault. Don’t change proxy modes yet: until the config has loaded, switching modes won’t make missing policy groups appear. Once the new config is enabled, continue to the next step.

Step 2: Choose a Proxy Mode

Find the mode selector at the top of the “Proxies” page or in “Settings.” You’ll usually see Rule, Global, and Direct. For your first connection, choose “Rule.” In Rule mode, the core checks each connection that reaches the client against the rules in the current config, in order, and applies the policy specified by the matching rule. That means one site can use a proxy while another connects directly on the same device. Stay on the page after switching and confirm the mode indicator shows Rule, then expand the policy group you want to use.

Policy groups and nodes are two different layers. A policy group is a selection point referenced by rules. Its entries may be individual nodes, another group, or a direct-routing policy. Group names are set by whoever wrote the config, so they aren’t always called “Proxy” or “Node Selection.” If the client highlights the current selection, check what it is. For nested groups, keep expanding until you know where the request will go. Some configs use an automatic selection group to choose the route, so changing a different group manually may not affect the policy used by the current rule.

Global mode is mainly useful for short-term comparisons. After switching, check which policy the global mode actually selected, not just the mode label at the top. If the global policy is Direct, requests may still bypass the proxy. And if an app isn’t sending requests through the client, switching to Global won’t route them through it. Direct mode helps you see what requests look like without a proxy route, but it doesn’t turn off the system proxy or disconnect your phone’s VPN. There’s no need to switch modes back and forth when getting started. Stay in Rule mode and use the connection logs below to see how the config routes traffic.

Before testing, choose a target. If you can find its domain in the rules, note which policy group it should use. If it’s supposed to connect directly, use it as a comparison—not as your only test of the proxy route. When the rule list includes a fallback, requests that don’t match earlier rules use that policy, which may not be a proxy. For help changing rule order or understanding the difference between domain and IP rules, see the full guide. For now, keep the existing config and continue connecting.

Step 3: Choose a Node and Enable the Proxy

Select a node in the manual policy group you checked earlier, then wait for the interface to show the selection. If the group offers latency testing, run a test to see which options respond from your current network. The result only reflects how the client reaches the test address and depends on both the target and your network. A responsive node doesn’t mean your app is connected, or that other destinations will perform the same way. If you’re using an automatic selection group, check the current route first; there’s no need to change the test address or selection criteria right away.

On desktop, enable “System Proxy” from Settings, the home screen, or the system tray menu. With the switch on, send your first test request from a browser that follows the system proxy settings. If the browser has its own proxy extension or profile, or is managed by workplace policies, first check whether it uses the system settings. The system proxy switch confirms that the proxy entry point is configured; it doesn’t mean every app is using it. For your first test, use one entry point at a time. Don’t run multiple proxy tools or change ports, DNS, and TUN settings just to test a browser request.

On mobile, the connection button is usually on the home or Proxies page. Tap it and, if the system asks to create a VPN configuration, check the app name in the prompt before approving. Return to the client, wait for the connection status to update, and check in system settings that the VPN for this client is active. If another VPN is already running, check whether it can be paused. Don’t remove a managed connection from a work device without authorization. Some clients let you choose which apps to route; if you use this feature, include the app you want to test before sending a request.

If the switch turns itself off immediately, check the client logs from the same time. If the service failed to start, troubleshoot that first. If a local listening port is already in use, check whether another instance is running. Don’t mistake an entry point error for a node failure, and don’t replace the entire subscription to fix a local conflict. If a browser works through the system proxy but another desktop app doesn’t, check whether that app supports its own proxy settings, or see the TUN section in the full guide for requirements involving virtual adapters, routing, and permissions. Once the entry point is stable, move on to step four.

Step 4: Check the Route Used by the Request

Open the “Connections” or “Logs” page in the client, then send a new request from the app you want to test. For a browser test, open a new page and note the target domain and time so you can find it in the connection list. An existing page may reuse an old connection or load from cache. If you can’t find a matching record, close the app’s existing sessions and try again. A page loading successfully only tells you the request got a response; it doesn’t tell you whether it went direct or through a proxy. Don’t mistake the client’s own latency test for an app request.

When you find the record, check these three things in order: the destination, the matching rule, and the policy used. If the interface only shows a group name, open the connection details or check the group’s current selection on the Proxies page. If a request expected to use a proxy matched the right rule and the final route matches the selected node, it was routed as expected. If the record shows DIRECT, check the config first to see whether that destination is supposed to connect directly. Then test a destination expected to go direct to distinguish “the proxy route works” from “traffic is being routed as expected.”

If there are no records for the app at all, check the entry point first: whether the desktop app uses the system proxy, whether the mobile app is covered by the VPN, and whether you sent the request after enabling the entry point. If there is a record but it used the wrong policy, return to step two and check the mode and rules. Remember that a site’s main domain, static asset domains, and login domains may follow different routes. If the policy is right but the request fails, keep the same destination and entry point and try another node in the same group. Change one thing at a time to identify whether the issue comes from the entry point, rules, or proxy route.

You can also visit an IP lookup service in the same browser and compare the results before and after enabling the proxy. First confirm that the lookup domain uses the proxy under the current rules. A changed IP only shows that the lookup request used a different route; it says nothing about the whole device or other apps. You’re done when the config is active, the app’s request appears in the connection logs, and the final policy matches what you expected. When you’re finished, turn off the client’s system proxy on desktop or disconnect through the client on mobile, then confirm the system settings have returned to their previous state.