Video calls are unusually sensitive to small network problems. A page may load normally while a Zoom meeting stutters, Slack messages arrive late, or Google Meet keeps reconnecting. Clash can help by giving work apps a predictable route while leaving ordinary browsing on a direct connection. The key is to choose a reliable node, place narrow app-related rules before the final catch-all rule, and test the real call path rather than relying on a single latency number. This guide covers a practical setup for Zoom, Slack, and Meet, with notes on client and core differences.

Plan Work Traffic Before Editing Rules

Start by deciding which traffic should use the proxy. For a focused remote-work profile, send the services you need for calls and collaboration through a dedicated proxy group, and leave the rest of your traffic direct. This avoids an all-or-nothing setup: a work call can use a stable route without forcing local banking, printers, company intranet sites, or routine regional browsing through a distant server.

Clash rules are evaluated from top to bottom. The first matching rule determines the route. That makes rule placement just as important as the rule itself. Put work-service rules above broad rules that could match first, and keep a final rule such as MATCH,DIRECT at the bottom if your intended default is direct access. If your profile already has country, private-network, or rule-provider entries, inspect their order before adding anything. A broad rule placed above the work rules can silently take precedence.

ServiceUseful starting domainsImportant limitation
Zoomzoom.us, zoom.comMeeting media can use separate hosts and UDP; a domain list is a starting point, not a guarantee that every deployment is covered.
Slackslack.com, slack-edge.com, slack-msgs.com, slack-files.comWorkspace features and file delivery may involve additional hosts or content delivery networks.
Google Meetmeet.google.comCall setup and media delivery may use other Google domains; routing all Google traffic can affect unrelated services.

These entries are deliberately conservative. Service infrastructure changes, and organizations may use managed accounts, custom domains, or regional endpoints. Begin with the domains you can verify in connection logs. Expand the list only when you have evidence that a required service connection is missing or taking the wrong route. In particular, avoid adding a broad rule for every google.com or googleapis.com hostname just to fix one Meet call: that may send unrelated Google traffic through the proxy too.

Keep a baseline: Before changing the profile, note whether a call works on the current connection, which node is selected, and whether the problem affects video, audio, screen sharing, or sign-in. A clear baseline makes it much easier to tell whether a new rule actually helped.

Choose a Node for Call Stability, Not Just Ping

A low ping is useful, but it does not prove that a node is suitable for real-time meetings. A ping test measures a small request at one point in time. Video calls also depend on packet loss, jitter, sustained upload capacity, route consistency, and whether the proxy protocol can carry the traffic the application uses. A node that wins a latency test may still perform poorly during a busy afternoon or fail to handle a particular UDP path.

If your configuration has several nodes, create a work-specific group rather than changing the global proxy selection every time a meeting starts. A manual select group is a good first choice because it lets you keep a known-good node selected during an important call. A url-test group can be helpful for ordinary use, but automatic switching during a live meeting may interrupt existing connections. Treat automatic selection as a convenience, not a substitute for a stable fallback plan.

When comparing nodes, test them at the time of day you normally work. Try a short call with the camera enabled, speak while another participant shares a screen, and check whether the connection remains stable for several minutes. If possible, repeat the test with a second node from a different location or provider. For a controlled comparison, change only one variable at a time: keep the device, network, meeting, and Clash mode the same while switching nodes.

  • Prefer consistency: A node with slightly higher latency but little packet loss may feel better than one with a very low but fluctuating result.
  • Check the upload path: Screen sharing and camera video both send data from your device. A download-only speed test does not measure this adequately.
  • Consider distance: A geographically closer node often has a shorter route, but peering and congestion can matter more than the map distance.
  • Keep a backup: Record a second node that works well enough for calls, and switch to it manually if the primary node becomes congested.
  • Avoid unnecessary switching: Do not change the selected node repeatedly during a meeting unless the current route is clearly failing.

For a simple configuration, a work group can contain a manual choice of your tested nodes. Replace the example names with the names used in your profile. If your client manages proxy groups through a subscription or profile interface, make the change in the supported override or profile-editing area rather than editing a generated file that the client may overwrite.

Configure Narrow Rules for Zoom, Slack, and Meet

The example below creates a manual Work-Proxy group and sends selected service domains to it. It assumes the referenced proxy names already exist in the proxies section. If your profile already has a suitable group, reuse it instead of creating a duplicate. Likewise, preserve the rest of your profile’s DNS, provider, and routing settings; this snippet is an illustration of the relevant sections, not a complete replacement configuration.

Work Proxy Group and Domain Rules
proxy-groups:
  - name: Work-Proxy
    type: select
    proxies:
      - node-work-primary
      - node-work-backup
      - DIRECT

rules:
  - GEOIP,LAN,DIRECT,no-resolve
  - DOMAIN-SUFFIX,zoom.us,Work-Proxy
  - DOMAIN-SUFFIX,zoom.com,Work-Proxy
  - DOMAIN-SUFFIX,slack.com,Work-Proxy
  - DOMAIN-SUFFIX,slack-edge.com,Work-Proxy
  - DOMAIN-SUFFIX,slack-msgs.com,Work-Proxy
  - DOMAIN-SUFFIX,slack-files.com,Work-Proxy
  - DOMAIN,meet.google.com,Work-Proxy
  - MATCH,DIRECT

Rule order is intentional: private-network traffic is kept direct, the selected work domains are sent to the work group, and everything else follows the direct default. If your existing rules include a final MATCH or a broad rule that appears before these entries, place the service rules above it. Do not add a second catch-all rule without checking what the profile already contains; multiple broad rules can make the result confusing and conceal an ordering mistake.

The DOMAIN-SUFFIX form matches a base domain and its subdomains, which is convenient when a service uses multiple hostnames beneath one domain. The exact Meet entry is intentionally narrower than a suffix for all Google services. Some meetings may still contact additional infrastructure, so use the client’s connection or rule log to identify the actual hostname when the call setup succeeds but media does not. Add only the specific host or service domain that is necessary, then test ordinary Google browsing to confirm it remains direct.

Some Mihomo-based cores also support process-name rules, which can be useful when an application contacts many changing domains. However, process matching is not equally available in every Clash client, operating system, or core mode. Executable names can differ between platforms, and an application may launch helper processes for calls or updates. If you use a process rule, verify that the running connection is attributed to the expected process in the logs. Do not assume that copying a Windows process name into a macOS or Android profile will work.

Be careful with port rules: Meeting media may use UDP, but a destination-port rule can also match unrelated applications using the same port. Add a port-based rule only when your core supports the required traffic path and you have confirmed the port in logs. A broad port rule is not automatically an app-specific rule.

Apply the Profile and Test Each Application

After editing the configuration, reload or update the active profile using the controls in your client. Do not assume that saving a file means the running core has adopted it. Check the client’s configuration or log view for YAML errors, missing proxy names, or an invalid group reference. A typo in Work-Proxy or in a node name can prevent the intended rule from working even if the rest of the interface appears normal.

  1. Confirm the group works: Select a tested node in Work-Proxy, then verify that the group shows the expected active choice. If the group contains DIRECT, make sure a manual selection has not accidentally left it there.
  2. Open the rule or connection log: Start Zoom, Slack, or Meet and inspect the connections created during sign-in and a live call. Check the matched rule and selected proxy group, not just whether the application window opened.
  3. Test the complete workflow: For Zoom, join a meeting and check audio, video, and screen sharing. For Slack, test messages, file previews, and a call if your workspace uses them. For Meet, test joining, microphone and camera access, and screen sharing.
  4. Check non-work traffic: Open a local or regional site and confirm it matches your intended direct route. Also verify that company intranet resources, local printers, and other LAN devices still work.
  5. Repeat on the backup node: If the primary route fails, select the backup manually and repeat the same test. This gives you a practical fallback instead of a group that has never been tested.

On desktop clients such as Clash Verge or Clash Verge Rev, rule and connection views may expose the matched domain and selected group, though the exact labels vary by version and core. Other clients may present the same information in a different panel. Look for the active profile, core type, rule mode, current group selection, and connection details. A mode such as global proxy can override the assumptions in this guide, so confirm that the client is actually operating in the routing mode your rules expect.

Testing should include both a new connection and an already-open application. Many apps reuse connections, and changing a rule does not necessarily move an established call to a new route. Leave the meeting, close or reconnect the application if appropriate, then start a fresh test. For work that cannot be interrupted, make configuration changes between meetings and keep a copy of the previous working profile so you can restore it quickly.

Troubleshoot Call Quality Without Guesswork

If an application launches but the call has no audio or video, first separate a routing issue from an application permission or device issue. Confirm that the microphone and camera work outside the meeting, then check whether the connection log shows the relevant service traffic and which rule it matched. If the hostname is missing from your work rules, add a narrow domain rule only after confirming that the connection belongs to the service. If it matches the work group but fails, compare the primary and backup nodes before changing DNS or adding broad rules.

When calls connect but suffer from stuttering, compare the connection on the same node with and without the proxy, if your network policy allows a safe direct test. A direct test is diagnostic, not a recommendation to bypass workplace security requirements. If direct is smooth and the proxied call is not, the likely causes include node congestion, packet loss, an unsuitable route, or a proxy protocol that does not handle the media path well. If both paths stutter, investigate Wi-Fi signal, local network congestion, device CPU load, and other active uploads before expanding the Clash rules.

If chat messages work but calls fail, remember that messaging and real-time media may use different destinations and transport behavior. A domain-only rule can route the web interface correctly while missing a media endpoint. Use the connection log during the failed call to identify what is new, and check the core’s support for the transport being used. A domain rule cannot make a node support a protocol it does not carry. Some environments may also restrict UDP; switching to another node or an application-supported TCP fallback may help, but the appropriate option depends on the client and network.

If the wrong traffic is proxied, inspect rule order before adding exceptions. A broad service suffix may be wider than intended, and a broad provider rule may match before the work entries. If local resources stop working, check LAN handling, system proxy behavior, and whether the application is connecting by a private IP address rather than a domain name. Keep changes small: modify one rule, reload the profile, create a fresh connection, and verify the match. Changing DNS, TUN mode, ports, and domain rules all at once makes the source of improvement or failure difficult to identify.

Once your work apps use a stable route and everyday browsing remains direct, save the tested profile or record the changes in a clearly named override. Revisit the setup only when your network, client core, or service behavior changes. A small, verified set of rules is easier to maintain than a long list of guessed domains—and a known-good backup node is often more valuable than another layer of routing complexity.

Get Started

Take Full Control of Your Traffic with Clash

Available on Windows, macOS, Linux, Android, and iOS. Flexible rules, simple setup, ready to use.

Download Free View Setup Guide →