Clash First Connection Guide: Pick a Node, Test Latency & Verify the Proxy

Successfully importing a subscription is just the prep work. Between "subscription imported" and "your browser is actually routing through the proxy," there are four more steps: picking a node, testing latency, enabling the system proxy, and verifying it's working. This guide walks through all four in order, explaining not just how but why.

Before You Connect: Three Quick Checks

Before clicking connect for the first time, run through these three checks in order. Skip any of them and everything that follows is a waste of time.

1. Profile is loadedOpen the "Profiles" page in your client and confirm the subscription you just imported is there and currently selected. If it's not showing up, go back and redo the import.
2. Nodes are listedOpen the "Proxies" page and check that the proxy group shows a list of node names. If the list is empty, the subscription probably failed to fetch — go back to the profile page and click "Update."
3. Mode is set to RuleSet the proxy mode to "Rule." Rule mode routes traffic by policy: sites in mainland China connect directly, while everything else matching a rule goes through your node — the easiest setup for daily use. We'll cover the difference between Global and Direct mode later; for your first connection, Rule mode is all you need.

Step 1: Pick a Node

On the "Proxies" page, find the proxy group named something like "Proxy" or "Select" (depending on your subscription, it might be called "Manual Select," "PROXY," or similar). Click into it, and you'll see every node your subscription includes. Whichever one you click becomes your active exit node.

Two rules of thumb for picking a node:

• Favor nodes that are geographically close. For users connecting from mainland China, nodes in Hong Kong, Japan, or Singapore are physically closer, which usually means lower latency and snappier page loads than nodes in Europe or North America.
• Favor lower latency, but don't over-trust the number. Low latency means faster response times — it doesn't guarantee download bandwidth. Whether video streams smoothly also depends on the node's current load. We'll cover how to test latency next.

If there's a group called "Auto" or "url-test," selecting it lets the client automatically re-test latency at regular intervals and switch to whichever node is currently fastest. Use it if you don't want to pick manually; if you want to stick to one specific route, select a node by hand instead.

Step 2: Run a Latency Test

Near the node list, you'll usually find a lightning-bolt icon or a "Latency Test" button. Click it, and the client tests every node in the group; a few seconds later, a millisecond value shows up next to each one.

It helps to understand how this works before reading the numbers: the client sends a request to a test URL (typically http://www.gstatic.com/generate_204) through each node, measuring the full round trip — local machine → node → test URL → back again. That means it's not the same as pinging the node directly; it's the real latency of the entire proxy chain, which is a much better indicator of what you'll actually experience.

Read the results in three tiers:

• Under 300ms: normal — plenty fast for browsing and streaming;
• 300–800ms: usable but noticeably slow, with perceptible lag on web pages — consider switching to a faster node;
• timeout: this node is currently unavailable — skip it.

If every node in the group times out, don't jump to blaming the client — troubleshoot in order: first, go back to "Profiles" and update the subscription, since expired subscriptions or changed node info are the most common cause; second, switch networks (e.g., try your phone's hotspot) and test again, to rule out local network interference with the node's entry point; third, check your system clock — TLS handshakes are sensitive to time drift, and a system clock that's badly off can cause handshake failures.

Step 3: Enable the System Proxy

Picking a node doesn't route your traffic yet — there's one more switch to flip. Go back to the client's home screen (or the "General" settings page), find "System Proxy," and turn it on.

It's worth spelling out what this actually does: turning it on points your OS's proxy settings to the local loopback address plus a port number, e.g. 127.0.0.1:7890. The default port varies by client — Clash Verge Rev defaults to 7897, Clash for Windows to 7890 — check the "Port" section in your client's settings for the exact value. From then on, any app that respects the system proxy (which includes almost every browser) sends its requests through the local Clash core first, which decides whether to connect directly or route through a node based on the rules.

To confirm this with your own eyes: on Windows, open Settings → Network & Internet → Proxy, and you'll see "Use a proxy server" turned on with the address set to 127.0.0.1 and the matching port. On macOS, check System Settings → Network → Details → Proxies for the same change. Turning off the system proxy switch in the client reverts these settings automatically.

Step 4: Confirm the Proxy Is Actually Working (Two Methods)

A switch being "on" doesn't mean it's working — don't skip verification. There are two methods: one checks the result, the other checks the process. Do both once.

Method 1: Check Your IP (Result-Based)

1. With the system proxy off, open ip.sb in your browser and note the IP address and location shown — this is your real, unproxied exit.
2. Turn on the system proxy, confirm a node is selected and the mode is set to Rule, then refresh the ip.sb page.
3. If the IP now shows the node's address and location, your browser traffic is going through the proxy.

To also confirm rule-based routing is working, open a second IP-lookup site that's based in mainland China alongside a site hosted elsewhere: the site outside mainland China should show the node's IP, while the mainland-based one still shows your real IP — proof that "proxy for outside sites, direct for domestic ones" is working as intended.

Method 2: Check the Connections Page (Process-Based)

Your client has a "Connections" page that lists every connection passing through the core in real time: the destination domain or IP, which rule chain it matched, and the final exit. Open your browser, visit a few sites, then check this page:

• When visiting a site outside mainland China, its domain shows up in the list with a node name at the end of the rule chain — confirming the traffic really went through the proxy;
• When visiting a site in mainland China, the rule chain ends in DIRECT — a direct connection with no detour, exactly as the rules intend.

The Connections page is the most reliable evidence you'll get: an IP-lookup site only proves the proxy works for your browser, while the Connections page shows exactly which route every single app and connection is taking. Whenever you need to check "is this app actually using the proxy," this is the page to check.

Common First-Connection Issues

Every node times out during the latency test — what now?

Update the subscription first, then try testing again on a different network, and finally check your system clock. If all three steps still leave everything timing out, the subscription itself has likely expired — you'll need to grab a fresh subscription link and re-import it.

System proxy is on, but the browser isn't routing through it?

Check for any proxy-management browser extensions you've installed (like SwitchyOmega) — they take over the browser's proxy settings. Disable the extension, or switch its mode back to "use system proxy," then restart the browser.

Getting a "port already in use" error on startup?

The default port is being used by another proxy app or a leftover process that didn't fully quit. Close the conflicting app, or change the mixed port in the client's settings to a different value (e.g., 7895), then turn the system proxy back on.

Connected, but speeds are slow?

Low latency doesn't guarantee high bandwidth. Heavy node load during peak evening hours is common — try another node in the same region, or test again at a different time. You can also switch from "Auto" back to manually selecting a node to avoid a congested route.

That covers the full flow, from picking a node to confirming it works. This sequence — check, pick, test, enable, verify — is the same routine you'll run through every time you set up a new device or switch to a new subscription.