SETTINGS REFERENCE
Shadowrocket Settings Guide
First check what each setting affects, then change its value. This page is an official, issue-focused reference. For your first import of an existing subscription, node selection, and initial connection, start with the user guide, then return to the relevant section here.
Shadowrocket is available for iPhone, iPad, Mac, Windows, Android, and Linux. This page uses the iPhone and iPad Settings interface as its reference. For other platforms, consult a setting only if the corresponding option is available; do not assume platform-specific features are shared across all versions.
The iPhone and iPad version costs $2.99 as a one-time purchase on the App Store. The Mac (Apple silicon and Intel), Windows, Android, and Linux versions are free. See the downloads for all platforms for download options and system permission steps. Check the App Store listing for iPhone and iPad system requirements. The app is a one-time purchase; this does not include a service plan. Get subscription or server details from your provider. This site does not provide, sell, or recommend such services.
Defaults are not recommendations
An existing installation may have restored a backup, imported a configuration, or retained older settings. This guide does not invent universal defaults for unverified switches. “Keep the default” means leave the setting in its current, unmodified initial state and note what the interface shows. Change one variable at a time, test the result, then decide whether to keep the change.
01 / SCOPE
Scope, current defaults, and change order
Distinguish Settings, Config, and system permissions
Before troubleshooting settings, sort the issue into one of three layers. Settings control client behavior, such as test methods, logs, and app preferences. The active Config determines rules and some aspects of network handling. System permissions determine whether the client can establish a system network connection, access the local network, or use cloud features. The same symptom can have different causes: a connection switch that will not stay on does not necessarily indicate a DNS problem, and a page that will not load does not automatically mean a rule failed. Note the screen, selected configuration, and exact message before opening the relevant setting. This helps avoid repeated changes without knowing which one took effect.
“Checking the default state” means carrying out a check; it does not mean restoring every switch to a fixed combination. On your first check, note each field’s current value, whether it is blank, whether a Config supplies it, and whether the system shows a permission prompt. Values on a migrated device do not represent a fresh installation’s defaults, and restored cloud data may not suit the current network. Do not delete an active configuration or reinstall the app just to reach a supposedly standard set of values. That changes several conditions at once and makes the issue harder to reproduce.
Global Routing determines whether rules are evaluated
The Global Routing modes Config, Proxy, and Direct correspond to Config, Proxy, and Direct. Config processes requests using the rules in the active configuration. Proxy routes applicable traffic through the proxy path; Direct sends it directly. None of these modes fixes server settings or guarantees that a destination is reachable. Use Config to troubleshoot rules. Temporarily switching modes only compares how the same destination behaves along different paths. Restore the original mode after testing, or later rule changes may appear to have no effect.
| Mode | What to check | Questions the test can answer |
|---|---|---|
| Config | Active Config, rule order, and policy | Which rule matched the request? |
| Proxy | Selected node and its connection settings | Is the proxy path itself working? |
| Direct | Local network and destination reachability | Does the issue change without the proxy path? |
Keep a rollback record instead of toggling settings repeatedly
Record four basics: whether you are on Wi-Fi or cellular, which Config and node are active, the current Global Routing mode, and whether the failure occurs before or after connecting. Then note the one field you plan to change and the expected result. For example, when changing DNS, define the test as “Can this domain resolve and load?” rather than vaguely checking whether the network feels faster. Keep the test target consistent so you can link the result to the change.
After making a change, send a new request. If the setting affects tunnel setup, disconnect and reconnect before testing. Apps with long-lived connections may continue using an existing session, so an old page still showing an error does not necessarily mean the new value failed to take effect. Close the affected page and open it again; this is more informative than repeatedly refreshing it. If nothing improves, restore the value you recorded before trying another setting. Do not switch nodes, change Config, edit DNS, and enable On Demand all in one test. Even if service returns, you will not know why.
Principles for choosing initial values
Leave fields that work as they are. Enter protocol settings exactly as specified by your provider. Enable additional behavior, such as automatic reconnection or external listening, only after the basic connection has been verified. If a field is unfamiliar, keep its current initial state rather than guessing from its name.
For precise troubleshooting, write a one-line conclusion for each test: what changed, how the request was made, what happened, and whether the setting was rolled back. You do not need to include a subscription URL or password. Keep separate notes for each device, since system permissions, networks, and copies of Config can differ. Look up interface terms in the glossary. The remaining sections apply the same method to each setting.
02 / DNS
DNS: Trace resolution issues and pages that will not load
DNS resolves names; it does not fix connectivity or rules
DNS resolves a domain name to an address, a common step before visiting a website. A request may also involve rule matching, connecting to a server, TLS negotiation, and transferring content. So “latency looks normal, but the page will not load” is not enough to conclude that DNS is at fault. Check whether the log shows a destination address, then see whether the connection stage failed. If an address was resolved but the connection timed out, switching resolvers usually will not fix an unreachable route or incorrect protocol settings. If even the selected node’s server hostname cannot be resolved, first troubleshoot basic name resolution before establishing the proxy connection.
DNS behavior may also depend on related settings in the active Config. Check both Settings and Config before making changes; otherwise, one setting may still control behavior while you edit another. A blank field does not necessarily mean “no DNS”—it may indicate that the system or Config supplies the resolver. Use the current installation and selected Config as your reference for defaults. If the network works, keep the existing resolution setup. Change it only when you can reproduce a resolution failure and compare the results.
Check the required format before entering an address
First check whether the field expects a resolver address, an endpoint with a protocol, or a line of configuration text. Do not copy values between fields just because both mention DNS. Check for leading or trailing spaces, typographic punctuation, and line breaks. If you use an encrypted DNS endpoint, make sure the selected field accepts that format and that the endpoint is reachable. A correctly written address does not prove the full path works; reaching the resolver also depends on the current network, certificate validation, and device time.
Prefer the resolution method already specified by your Config and reachable on the current network. If you need to switch, keep one known-working baseline and avoid filling the field with multiple addresses from unclear sources. More resolvers do not guarantee greater reliability: one slow or faulty response can make results inconsistent. Keep the network and node fixed during testing. Try one previously failing domain and one that normally works, and note whether each resolves to an address rather than relying on a single page loading by chance.
| Symptom | Check first | Next step |
|---|---|---|
| Node hostname will not resolve | System network and basic name resolution before connecting | Restore basic connectivity before connecting |
| Destination domain will not resolve | Current DNS settings and related Config entries | Change only the resolution method, then retry |
| Address resolved, but connection times out | Selected policy, node, and destination reachability | Check connection logs instead of blindly switching DNS |
| Only some networks fail | Network restrictions, address families, and resolver reachability | Record Wi-Fi and cellular results separately |
How caches, address families, and rules can interact
The system and apps may cache resolution results, so refreshing an old page immediately after a change may still use an old address. Reconnect and make a new request. If the current interface offers a DNS cache clearing option, note the current state before using it, but do not treat clearing the cache as a lasting fix. If the issue goes away briefly and returns, check upstream responses or network changes. If only one app is affected, consider whether it uses its own resolution path. A working browser does not prove every app uses the same settings.
A domain may resolve to IPv4 and IPv6 addresses, and either address family may be unreachable on the current network. Do not disable IPv6 just because it appears, or mistake a failure in one address family for a complete DNS failure. Use the logs to confirm which address the connection actually used, then compare results on another network. For rules such as IP-CIDR or IP-CIDR6, also check whether a changed resolved address affected the match. Successful resolution and correct routing are separate things to verify.
Do not use a fixed address as a substitute for testing every domain
Some services depend on domain names, TLS hostnames, and dynamic addresses. Failure to reach a particular IP does not prove DNS is faulty; success at that IP does not prove the complete domain-based request works.
After changing a setting, retry both a working domain and the domain that failed, then check the resolution and connection stages in Log. If only a particular site is affected, note the time and failure stage instead of widening the scope of changes. For a layered approach to “the test responds, but the page still fails,” see DNS troubleshooting: pages not loading.
03 / ON DEMAND
On Demand: Automatic connection conditions and manual control
Verify a manual connection before enabling automatic conditions
On Demand triggers connection behavior when specified conditions are met; it does not fix server availability. Set automatic conditions only after you have selected a node manually, connected, and confirmed that the destination is accessible. Otherwise, automatic retries can overlap with the original issue, causing repeated reconnects or attempts whenever the network changes. When troubleshooting for the first time, record the existing conditions and temporarily disable On Demand. Start each connection manually so you can pinpoint when the failure occurs.
Check the default state on the On Demand screen of the current installation, including the main switch and each condition listed below it. After restoring a configuration or device backup, the main switch alone does not tell you whether every condition has returned to its initial state. Start with the narrowest scope: set behavior only for a specific network situation you need, test both entering and leaving that situation, then add other conditions gradually. Do not copy someone else’s network names or assume Wi-Fi networks with the same name offer the same access.
Check network conditions and system permissions together
If the interface lets you set conditions by network type or Wi-Fi name, check each one to see whether it means connect or disconnect, and whether a broader condition overrides it. Verify the network name as the device identifies it; spaces, letter case, and similar-looking characters can affect matching. Access to network information may also require system permission. If the app cannot get the information needed to match a condition, check system permissions and on-screen prompts rather than repeatedly adding conditions with the same name.
Start testing on the current Wi-Fi network: note the manual connection state, enable one condition, leave the network, then rejoin and check for the expected behavior. Test cellular separately. The system needs time to establish a network after switching, so avoid switching back and forth rapidly and checking only the final switch position. Wait for the network to stabilize at each step, make a new request, and compare the trigger time with the connection logs. This helps distinguish a condition that did not match from a connection that failed to establish.
If a manual disconnect is followed by a reconnect, check what triggered it
While On Demand is active, a manual disconnect may not permanently stop automatic connection attempts. The system may reapply existing conditions after a network event or request. If you need to stay disconnected, check the On Demand main switch and relevant conditions instead of repeatedly tapping the connection switch. After changing them, check the system VPN status again to make sure no other automatic condition remains active. Base your conclusion on the actual conditions and system behavior rather than assuming every repeated connection means the switch is broken.
Also distinguish between a VPN configuration saved by the system and an active connection. Permission to add a configuration means the system has the required connection profile; it does not mean traffic is flowing at all times. On Demand cannot bypass a previously denied permission or a deleted system configuration. Connect manually first, follow the system authorization prompt, then return to the conditions. For more on authorization and system configuration, see Understanding the Add VPN Configuration prompt.
Rollback steps for network-switching issues
Record the conditions → pause On Demand → reconnect manually → test on a stable network → restore conditions one at a time. Do not delete the system configuration and edit node settings at the same time, or you will not know which change affected the result.
Automatic connection does not automatically select a working route
Triggering a connection, selecting a node, and confirming that a destination is reachable are separate tasks. Even if On Demand triggers successfully, the selected node may be outdated or have mismatched settings. A successful connection also does not guarantee that every destination follows the expected route. Keep a repeatable set of test targets: one for the local network, one for the proxy path, then check the active rule match. When something fails, locate the issue across these three layers instead of continually broadening automatic conditions just to keep the switch on.
If you need to sign in or confirm a captive portal on a public network, first make sure basic network access works, then test automatic connection. A network that cannot provide basic access has not met the prerequisite; do not mask that with more aggressive reconnection conditions. After testing, note why you use each condition and which networks it applies to. When network names, permissions, or habits change, review the conditions against your notes rather than importing a complete set again.
04 / PING
Ping and Connectivity Test: What test results can and cannot tell you
Check the test method before interpreting latency
Ping is a troubleshooting tool, but not every test is the same kind of probe. Results depend on the selected method, test target, and current network. A TCP connection, a basic probe, and an application-layer request through a proxy do not test the same parts of the path. Note the current method before running a test. Check the displayed value for the default; do not assume an unverified method is the universal default. To compare results, keep both the method and target the same, or the numbers are not directly comparable.
A node responding usually means only that the part of the path covered by this probe responded. It does not prove that authentication settings are correct, a website is reachable, or data can be transferred reliably over time. Conversely, a timeout does not always mean the server is completely offline: the test target may be unreachable, that type of probe may be restricted, or the timeout may be too short. Treat results as clues that narrow down the issue, then verify with a real request. Do not treat a single number in the node list as the final verdict.
Choosing a test target, timeout, and retry count
If the interface lets you specify a test address, choose a target you are authorized to use that is reliably reachable and has a simple response. A target that requires sign-in, redirects, or restricts access introduces extra variables. Do not keep changing addresses in search of a lower number. First keep the original target as a baseline and confirm that the failure is repeatable; then change only the target for comparison. If a different target works, check whether the difference is due to the target or whether the path itself recovered.
If you can set a timeout, keep its initial value first. A short timeout may classify a slow but successful request as a failure. A longer one makes each test take longer but cannot improve server reachability. Adjust it slightly only after confirming that the current network genuinely needs more time to connect, then test again. For retries, a few spaced tests are easier to interpret than rapid repeats and less likely to mistake a brief network fluctuation for a persistent problem.
| Test layer | What it can indicate | What it cannot prove on its own |
|---|---|---|
| Server port connection | Whether the address and port accept a connection | That protocol authentication and destination access work |
| Request through a proxy | Whether this proxy request completed | That every website and app works |
| Connectivity Test | Connection behavior under the current test conditions | Ongoing availability after future network changes |
What to check after a timeout
First confirm that the device has basic network access. Then check the server address and port, followed by the protocol type and authentication fields. Match the Shadowsocks encryption method and password, VMess or VLESS identity fields, and Trojan authentication and TLS settings against the original server details. WireGuard and Hysteria2 have their own parameters; do not reuse values just because some fields look similar. If your provider’s details are unclear, ask them for the correct settings rather than changing values at random.
Check the device clock as well. An incorrect time can affect certificate validation or connections that require accurate time, but correcting the clock is not a universal fix for every timeout. Set the system time and time zone correctly, then repeat the same test. If only one node fails, keep the network unchanged and inspect that node. If all nodes fail at once, check shared conditions such as the network, resolution, and test target instead of editing node settings that may already be correct.
After a test succeeds, verify with a real request
Select the node you plan to use, confirm the Global Routing mode, connect, and make a new request. Check in Log that it followed the expected policy. Do not rely only on a long-lived connection that was already active in the background; opening a new page makes it easier to match the request to its log entry. For uploads, streaming, or long sessions, test your actual use case too. A brief probe cannot substitute for these scenarios.
For troubleshooting notes, keep the test method, network type, failure stage, and sequence of events. You usually do not need to share full node details. For a step-by-step approach to node timeouts, see Node timeouts and connection failures: troubleshooting steps. Once the test passes, return to your normal testing frequency; repeatedly probing just to watch the numbers is unnecessary. Keep a setting only if it reliably supports the requests you need.
05 / LOG
Log and Data: Follow requests to find clues
Treat logs as a record of events, not a list of conclusions
Log helps reconstruct the stages of a request: receiving the destination, resolving it, matching a policy, establishing a connection, and completing or failing. Details may vary by protocol and log level; not every request needs to produce identical lines. Start with entries close to the time you performed the test, then follow later events for the same destination or session. A single error line can lack context. For example, if resolution failed first, a later connection error may simply be a consequence.
Check the current Log screen for whether logging is enabled by default, how much it retains, and which levels are available. For everyday use, keep enough detail for basic troubleshooting, but do not leave the most detailed level enabled indefinitely. Increase detail temporarily when diagnosing an issue, reproduce it once, then restore the previous level. Detailed logs may contain destinations, network information, and configuration clues, and can be harder to read. Identifying the stage you need to investigate is more effective than enabling everything and searching line by line.
Create a clean reproduction with a matching timestamp
Before testing, note the current network, Config, node, and Global Routing mode. If you plan to clear old logs, save any useful findings first; clearing logs only removes records, not network problems. Reduce network activity in other apps, open just one failing destination, and note the time. Then check Log. Background system requests may still appear, so the logs will not necessarily contain only that one action. Identify the request using the destination, timestamp, and event sequence instead of assuming the latest line belongs to your test.
After finding the target, check whether the client handled the request, then inspect its policy. If the rule should use PROXY but the log shows DIRECT, check the rules and Global Routing first. If the expected policy was used but authentication failed, check the node settings. If the connection succeeded but the app still reports an error, consider the destination’s response, the app session, or related requests. A web page may contact several domains; the main page loading does not mean its images, sign-in, or API requests also succeeded. Follow the request sequence as needed.
| Log stage | Check first | Avoid this misdiagnosis |
|---|---|---|
| Resolution failure | DNS and basic connectivity | Re-entering every authentication setting |
| Unexpected policy match | Global Routing, active Config, and rule order | Mistaking a routing issue for an offline server |
| Authentication or TLS failure | Protocol fields, hostname, and system time | Assuming a reachable port rules out incorrect settings |
| Connection drops after it is established | Network changes, destination response, and session stability | Treating one interruption as a failure for every request |
Data statistics do not replace per-request logs
Data statistics show traffic over a particular scope; they are not a per-request success rate. Traffic only indicates that data was active. It may come from a background service or data exchanged before a failure. Totals also depend on the measurement period and when they were reset, so they are not a direct measure of the efficiency of two configurations. To check whether a particular page followed a rule, inspect its request in Log rather than relying on whether the total traffic increased.
Likewise, a connection indicator in the system status bar does not mean every app request succeeded. Tunnel status, rule matching, and destination responses are separate layers. First use the status to check whether the connection was established, then use Log to check the request path, and finally test whether the app works as expected. Only when all three agree do you have a complete picture. If the statistics change but you cannot find the destination in Log, check how you reproduced the issue, whether logging was enabled, and whether you are viewing data from the current connection.
Remove sensitive information before sharing troubleshooting details
Do not share subscription links, authentication fields, full QR codes, or cloud configuration URLs. Keep only the failure stage, policy type, and necessary context. You can replace domains and addresses with clearly marked examples, but say which fields you changed.
When troubleshooting is complete, restore any temporary log-level changes and keep a concise summary: how to reproduce the issue, which field changed the result, and the final value. If a change did not help, note that you rolled it back so the same test is not repeated later. A clear record is easier to maintain than a large, unorganized log. For rule order, see Custom rules and FINAL.
06 / WIDGET
Widget: Shortcuts, status, and system refreshes
A Widget is a shortcut, not a standalone connection engine
A Widget provides shortcuts or status information in locations supported by the system. Its actual capabilities depend on the components available on your device. It cannot replace first-time authorization, a valid configuration, or a working node. Before using one, connect manually inside Shadowrocket. Then add an available Widget through the system interface and check whether it opens the app or performs an action offered by the interface. If the system does not show a particular Widget, do not assume a line is missing from the configuration or look for a network rule to fix an unrelated system component.
Check the Widget’s default state in two places: related settings in the app and whether a component has actually been added to the system Home Screen or Widget page. An available Widget option in the app does not mean the system has added one; an added Widget does not guarantee its content will refresh in real time. Keep the app’s initial options and add only the shortcuts you use. Choose a size and number of components that make controls easy to identify; you do not need every layout to troubleshoot a connection.
Displayed status may lag behind the connection
The system schedules Widget refreshes, so its text may briefly differ from the app’s current status. After tapping a connection control, return to the app and check the system VPN status, then make a real request. Do not keep tapping just because the Widget still shows an old status. Repeated actions may toggle the connection back and forth, making it harder to tell which action took effect. Complete one action and verify it before moving on.
If the status remains out of date, open the app to confirm the current connection, then return to the system’s Widget screen and check again. If it still does not match, remove and re-add the Widget without changing the app configuration. Re-adding it refreshes the shortcut; it does not fix a node or DNS. If the connection also fails inside the app, leave the Widget troubleshooting path and check basic connectivity, permissions, and node settings. Do not delete all Config files to troubleshoot a Widget, or you will turn a display issue into a configuration recovery problem.
Lock Screen, system permissions, and visible information
Whether a Widget can be added to the Lock Screen or another system location depends on the device’s OS and available Widget layouts. Use only the add controls your system provides; layouts from other OS versions are not requirements for your device. If an action needs to open the app, follow the system’s prompts; this does not mean the connection feature has failed. For actions that require an unlocked device, unlock it before testing so a system access restriction is not mistaken for an unresponsive app.
Consider whether Widget information is appropriate for a location others can see. Node labels may reveal personal details about a device, region, or purpose, and configuration names may expose usage habits. Use names you can recognize without making them overly specific, and show only what you need on the Lock Screen. After renaming something, check whether the Widget updates. A name is only a label; do not change server addresses, authentication fields, or subscription URLs to make the display look neater. Those fields serve entirely different purposes.
Distinguish a shortcut failure from a connection failure
If the app connects but the Widget does not work, check the system component and its refresh behavior. If the app itself cannot connect, check permissions, network access, and the node. Do not change settings in both branches at once.
When it interacts with On Demand, check the final trigger conditions
If a connection returns after disconnecting from a Widget, an active On Demand condition may be responsible; the Widget may not have repeated its action. Check the automatic conditions in the app, temporarily disable them, then test the Widget again. If it behaves as expected, adjust the automatic connection conditions instead of repeatedly removing the Widget. Keep the network fixed during testing so a Wi-Fi switch does not confuse the effect of a manual action with a system network event.
As a final maintenance step, test the Widget only after basic connectivity works, the app connects, rules behave as expected, and automatic conditions have been checked. That way, you still have an in-app control path if the Widget is temporarily unavailable. For the complete first-connection and system-authorization workflow, return to the user guide. This section covers shortcuts, status, and refresh behavior; a Widget is not a network performance setting.
07 / ICLOUD
iCloud sync: Scope, conflicts, and recovery
Check what syncs; syncing is not a prerequisite for connecting
iCloud features save or sync data within the limits of the system and app. They are not required to establish a proxy connection. First confirm that the configuration works on this device, then consider syncing it to another. Check the items and actions listed on the current screen to see what the sync feature includes. Do not assume nodes, Config files, preferences, logs, and system permissions are all copied together. In particular, system VPN authorization must be confirmed on each device. Seeing a configuration file in the cloud does not mean another device can connect.
Check the default state on this device as well. In-app switches, system iCloud settings, and the account in use can all affect the result; one enabled switch does not prove sync has completed. Keep a usable local copy, then test with one identifiable configuration. If the interface offers separate actions for importing, exporting, or selecting a cloud file, determine whether each copies data once or maintains an ongoing sync relationship. Similar action names do not guarantee the same update behavior.
Create a recoverable local baseline before enabling sync
Before taking action, note the name and source of the active Config and the last changes you made manually. If you need a backup, use the export method available in the app and save the file somewhere you control. A backup may contain subscription URLs and authentication details, so treat it as sensitive and do not put it in a public folder or shared screenshot. Continue with the cloud copy only when you know what the backup contains and how to restore it. Sync is not a substitute for an independent backup; an incorrect change may be synced to other devices too.
For an initial test, make a recognizable change on one device, such as renaming an item, that will not affect connectivity. Wait for the expected change on the other device, then test in the opposite direction. If the available action only imports a file, treat it as a one-time copy rather than expecting future changes to appear automatically. Do not repeatedly switch accounts, import the same file, or delete local configurations to force sync. This can create more copies and make it unclear which one is active.
Handling edits made on two devices at once
If there is a conflict, stop editing on the other device and save a copy of each version. Do not use the filename alone to decide which is newer; identical names can contain different settings. A displayed timestamp also does not necessarily show when each setting was changed. Compare server fields, rule order, policy names, and the parts you edited most recently. Once you identify the version to keep as your baseline, use an explicit import or save action to make both sides consistent. Do not choose an overwrite direction until you know which copy you want to keep.
Subscription updates and cloud sync are separate paths. Subscribe updates normally fetch content from an existing provider URL; cloud actions handle data between devices or file locations. If nodes have not updated, check the subscription update result rather than turning off iCloud. If local rules revert to an older version, check the import source and cloud copy. Also check whether a subscription update replaces parts you edited. Keep important custom changes in a separate record instead of relying on memory to re-enter them.
| Symptom | Check first | Recommended action |
|---|---|---|
| Content is missing on another device | Account, system cloud permissions, and what is actually synced | Test with one Config; do not repeatedly import batches |
| Multiple similar copies appear | Check whether a one-time import was repeated | Organize names only after confirming the working baseline |
| Node details have not refreshed | Subscribe update result | Troubleshoot the subscription path first |
| Config is present but will not connect | Local authorization, network, and server settings | Follow the first-connection steps on the new device |
Verify the operating environment after restoring
A successful restore means more than seeing a file. You must be able to select the expected Config, check node details, connect, and match the correct rule. Different devices may be on different networks, so check DNS and On Demand conditions separately. Do not treat a permission prompt on another device as unnecessary just because the first one works. Follow the system’s authorization steps on that device, then verify it using the layered method in this guide.
If you are concerned that sync may overwrite recently fixed settings, keep two copies and pause further edits. Confirm the direction of the current sync action before resuming normal use. For subscription URLs, get accurate details from your provider and do not share links publicly when asking for help. For differences between Subscribe, manual entry, and QR code import, see Ways to import a subscription. The import method also determines where to look for future updates.
08 / PROXY AND CONFIG
Proxy ports, rule interactions, and final rollback
A local listening port is not the remote server port
First distinguish the directions of Proxy-related ports. The port under Add Server is a remote server connection setting and must match your provider’s details. A local proxy listening port is used by other programs to connect to an entry point on your device; the two do not have to match. If you use only the system connection feature, you do not need to enable a local listener just to “complete” the setup. Configure one only if the current platform and interface provide it and you have a specific need. Do not assume that a listening feature available on one platform exists on every device.
Check the current interface for the default port, listening address, and whether local network access is allowed. This guide does not specify universal starting values across platforms. If you have no clear need, leave them unchanged. If another program needs to use the local proxy, test on the same device first and make sure the proxy type matches what that program supports. HTTP and SOCKS5 are not interchangeable labels. Selecting the wrong type can prevent requests from working even when the port is reachable.
Check the listening address and access scope separately
Listening on a loopback address makes the proxy available only to the local device. Listening on other addresses can expand access. Do not open it to the entire local network just to troubleshoot a port. First check whether another process is using the port, whether the target program has the correct local address, and whether the proxy is actually running. If a device on the local network needs access, also check system permissions, firewall settings, network isolation, and any access controls in the app. Test only on a network you manage; do not expose a listening port on an untrusted network.
If there is a port conflict, choose an available port permitted by the current system and update every program that uses it. Changing only one side creates a new failure. Afterward, reconnect and restart the relevant listener if needed. Close any temporary access you opened once testing is complete. Local network addresses can change, making an old address saved on another device invalid. When cross-device access fails, check the current address before changing the remote node.
Rules choose a route; they do not fix protocol settings
In Config mode, rules are generally evaluated from top to bottom. A matching rule applies its associated policy. DOMAIN-SUFFIX matches domain suffixes; IP-CIDR and IP-CIDR6 match address ranges; GEOIP relies on the corresponding location lookup; and FINAL acts as a fallback. PROXY, DIRECT, and REJECT indicate proxy, direct, and reject policies. Correct rules cannot compensate for an incorrect password, certificate hostname, or server port. Verify routing and connectivity separately.
The rule block below illustrates ordering; it is not a complete configuration. The example domains and address ranges demonstrate how matching works and should not be copied into your current file without checking. Before use, confirm that the policy names are available and keep a copy of the original file. The FINAL policy in particular determines where other requests go. Do not assume a policy shown in an example is right for every network.
[Rule]
DOMAIN-SUFFIX,example.com,PROXY
IP-CIDR,192.168.0.0/16,DIRECT
GEOIP,CN,DIRECT
FINAL,PROXY
To verify a rule, make a request that should clearly match it, then check the match in Log. If the rule seems ineffective, check in order: whether Global Routing is set to Config, whether the edited Config is selected, whether a broader rule appears earlier, and whether the request reused an existing connection. To test a domain rule, send a new request for that domain instead of relying only on an existing session. For a full explanation of rule order, see Rule priority and FINAL.
Finish troubleshooting with the fewest changes
Finally, restore temporary test settings one at a time: turn off any local listener you no longer need, reset the log level, return Global Routing to its normal mode, and keep only verified On Demand conditions. Then reconnect on a stable network, revisit the destination that failed, and check the rule and logs again. If restoring a setting immediately brings the issue back, that change is a clear focus for the next round. You do not need to adjust every option again.
Completion checklist
You can establish the connection repeatedly, requests follow the expected policy, behavior after network changes matches the settings, and you have a recoverable copy of the configuration. A switch showing as enabled or one successful test is not enough on its own.
If the issue persists, prepare a minimal reproduction: platform, network type, current mode, failure stage, and the settings you tried and rolled back. Do not include full subscription or authentication details. For first-connection issues, return to the user guide and check each step in order. To download the client again or review system permission steps, see the installation instructions on the downloads page. Keep settings that explain the result and undo changes without supporting evidence; this gives you a stable starting point for future maintenance.