[Settings]Official guide · Reference

Settings Reference

Start with what each setting affects, then check DNS, On Demand, connectivity tests, logs, sync, and local ports.

For importing an existing subscription, choosing a connection method, and checking access for the first time, start with the Quick Start. This page explains how settings work, when to change them, and how to revert changes. Terms follow the common iPhone and iPad interface; settings locations and available options may differ on other platforms.

8 reference sections Check current defaults first Change one setting at a time

[General]

1. Settings and defaults

Distinguish Settings, Config, and server parameters

A single toggle does not determine how Shadowrocket connects. App-level preferences in Settings, the selected Config, the server selected in Home, and network permissions granted by the system all affect requests. Before troubleshooting, note the state of each. For example, changing a DNS entry in Config does not change the server address; changing the port in Add Server does not change the local Proxy listening port. Identifying which layer a setting belongs to helps avoid repeated changes in the wrong place.

Server parameters describe how the client connects to a server you already have: protocol, address, port, authentication, and transport requirements. A configuration file mainly describes how requests are handled, including DNS, rules, and policies. App preferences may cover tests, display, connection triggers, sync, and logs. Locations and options can vary by platform; similar names do not mean settings override one another. If a change seems to have no effect, first check the active Config and server instead of assuming the setting is broken.

How to interpret defaults

This guide does not treat the state seen on one installation as a universal default. A fresh install, restoring from a backup, importing a remote configuration, or keeping older settings can all produce different starting states. Options with the same name may also depend on system permissions or configuration contents. Recommendations here provide a troubleshooting baseline, not instructions to override everyone's preferences. If the interface offers a default option and there is no clear reason to change it, leave it as is. A field missing from a configuration does not necessarily mean it is disabled.

When checking settings, record the field name, screen, current value, and reason for any change. For toggles, note on or off; for selectors, note the selected option; for configurations, keep the original file. Do not delete active data just to inspect defaults, or reset all settings without a backup. To compare behavior, create a separate test Config and keep the original as a restore point. When testing is done, explicitly switch back to the intended file rather than leaving the test file in everyday use.

LayerWhat it affectsRecord before changing
SettingsApp preferences, tests, connection triggers, and accessibility featuresOption names and current values
ConfigDNS, rule matching, and policy referencesCurrent file and its original contents
Add ServerParameters for connecting to an existing serverProtocol, address, port, and transport requirements
System settingsNetwork permissions, background behavior, and cloud account requirementsPermission status and current network

Use a reversible change process

Before starting, make sure you can still access your existing subscription or server details, and back up the current Config. Choose a test destination you can revisit, then keep the network, server, and Global Routing state the same. Change one setting at a time, reconnect if needed, and test against the same destination. Link a change to a result only after you can reproduce it reliably. If you change DNS, the server, and rules together, neither success nor failure will tell you which change mattered.

You do not need to obtain a connection service to read this guide. Shadowrocket is closed-source commercial software; a one-time client purchase does not include a connection service. This site does not provide, sell, or recommend servers or subscriptions. See the Download page for ways to get the app: the iPhone and iPad version costs $2.99 on the App Store as a one-time purchase; the Mac (Apple silicon and Intel), Windows, Android, and Linux versions are free. Check the App Store listing for iPhone and iPad system requirements. For other platforms, the Download page links to installers hosted at /dl/shadowrocket/ on this site.

The following sections cover purpose, checks, recommendations, and verification. If an option mentioned here does not appear in your interface, first check whether it applies to your platform and configuration; do not substitute a similarly named field. When you see a specific error, keep its original wording and look it up by symptom in FAQ. Avoid changes that could widen the scope of the problem before the cause is clear.

[DNS]

2. DNS and name resolution

DNS handles name lookups

DNS translates domain name lookups into addresses that can be used to connect, but successful resolution does not guarantee a successful connection. A request may first look up an address, then match a rule in the Config, and finally connect to a destination or proxy server. Depending on the route, resolution may happen locally or remotely, so a single lookup result cannot tell you where all traffic goes. Troubleshoot these separately: did the name resolve, which policy matched, and did the connection complete?

If a domain does not load but an address connects, that points to a possible issue in the resolution path; it does not prove that a particular DNS service is unavailable. Accessing a site by address can change TLS verification, virtual host selection, and app behavior, so it is not an equivalent test. Instead, identify the error in the same app and request path: a name lookup timeout, a server connection timeout, and a certificate name mismatch require different checks. Do not treat them all as DNS failures.

Check dns-server and system

In the active Config, find dns-server under [General]. If it is set, note its full contents and source. If it is absent, do not assume a particular public address is in use. The snippet below shows the system-provided resolver setting as a minimal example to illustrate the field; it is not a replacement for your entire configuration. The system resolver may depend on the current network, device management, or other system settings.

[General]
dns-server = system

Start with the system settings that currently work, especially on workplace or campus networks, or where you need access to internal hostnames. External resolvers may not be able to resolve internal domains. Replacing one without checking can leave public sites working while file services, sign-in pages, or internal apps fail. If you need another resolution method, choose one that fits your network and confirm it is reachable over the current connection path. Do not choose by name or advertised performance alone.

Caching, network changes, and address families

Refreshing a page immediately after changing DNS may still use an app connection pool or cached results, so one refresh does not prove the new setting is in effect. Finish related test requests, disconnect and reconnect Shadowrocket if needed, then retest under the same conditions. Avoid rapidly cycling through multiple resolvers: network changes, cache updates, and server response differences can happen at once, making the cause hard to identify. Recording before-and-after behavior is more useful than trying more addresses in quick succession.

A domain may resolve to both IPv4 and IPv6 addresses. If the network path for one address family is incomplete, some apps may work while others take longer to respond. Check connection logs for the address actually attempted rather than disabling IPv6 as soon as it appears. Disabling an address family may bypass a path issue temporarily, but can also affect services that depend on that network. Compare results across networks before deciding whether a targeted change is needed.

Observed symptomCheck firstAvoid this assumption
No response to the lookupResolver reachability, current network, and connection statusAssuming server authentication is at fault
An address is available, but the connection times outRouting policy, destination port, and actual connection pathBlindly replacing DNS again
Only internal names failInternal resolver and intranet access requirementsReplacing internal records with external lookup results

Safe changes and rollback

Encrypted DNS changes how queries are transported; it does not automatically fix server parameters, rule order, or destination outages. A resolver endpoint specified by hostname also depends on how that hostname is resolved initially and which path carries the request. If you cannot account for those dependencies, restore the previous setting, confirm ordinary access works again, and investigate the resolution method separately. Do not add several unfamiliar DNS options at once, especially by combining conflicting entries from different configurations.

A successful check means the actual destination remains available under the expected policy, not merely that a test button reports success. Check at least one everyday public destination, one internal destination you actually need, and behavior after a network change. To roll back, restore the original DNS entry and reconnect using the same steps. If behavior does not change when the setting is restored, investigate other layers. DNS-related rule behavior is covered in the next section; the two are connected, but they are not the same setting.

[Rule]

3. Global Routing and Config rules

What Config, Proxy, and Direct do

Global Routing is the first place to check whether the Config is being used. The interface options are Config, Proxy, and Direct. Config selects a policy using rules in the active configuration file; Proxy routes eligible requests through a proxy path; Direct uses a direct path. These are client-side handling choices, not guarantees that a destination is reachable. They should not be taken to override every type of device traffic or system restriction.

For everyday rule-based routing, use Config after checking the configuration contents. Proxy and Direct can help narrow down an issue by comparing how the same destination behaves on different paths, but they are not substitutes for understanding the rules. If a destination works only with Proxy, check the matching rule, policy reference, and final policy in Config. If it fails in all three modes, check the current network, server, or destination service first rather than looking only at the rule list.

Interface modePrimary useWhat to verify
ConfigRoute traffic according to the Config rulesActive file, matching rule, and policy
ProxyCompare behavior on the proxy pathWhether an existing server can complete the request
DirectCompare behavior on the direct pathWhether the local network can reach the destination

Rule order matters more than rule count

In Config, requests are generally checked against rules in order, and the first match determines the policy. More specific exceptions should usually come before broader rules, with FINAL at the end. A broad rule near the top can prevent later, more specific rules from matching. Adding rules does not automatically improve accuracy; overlapping conditions can make troubleshooting harder. Before editing, check why the existing rules are not producing the expected result.

[Rule]
DOMAIN-SUFFIX,example.com,PROXY
DOMAIN-KEYWORD,example,PROXY
IP-CIDR,192.168.0.0/16,DIRECT
GEOIP,CN,DIRECT
FINAL,PROXY

This snippet illustrates syntax and order only; it is not a complete configuration for every network. DOMAIN-SUFFIX matches a domain and its subdomains. DOMAIN-KEYWORD matches a keyword in a domain and may cover more names than expected. IP-CIDR matches an address range, GEOIP uses an IP geolocation database, and FINAL sets the policy for requests that did not match earlier rules. example.com is included for illustration only, not as service information you can import.

Policies, resolution, and network boundaries

PROXY, DIRECT, and REJECT are different policies. The first needs a working proxy path, direct access depends on the local network, and reject actively blocks matching requests. A custom policy name must also have a corresponding definition in the Config; spelling differences can cause reference errors. Troubleshoot with basic policies you understand before checking more complex choices. Do not infer which server a policy uses from its name alone.

Domain rules and address rules also behave differently. When a request retains its domain name, it can match a domain rule directly; address-based matching may depend on DNS results and the corresponding address range. A service may use multiple domains or dynamic addresses, so a rule for a single address may not cover every request. If only part of a page fails to load, check the specific failing destination rather than just the main domain in the browser bar. Sign-in, images, and APIs may use separate hostnames.

Recommended Config maintenance

Manage remote Configs and local edits separately. If you directly edit a file that is replaced during an update, your changes may be lost on refresh. If you rely indefinitely on a local copy, you may miss important updates to the original. First identify the file's source, update method, and where your edits are stored. Document the purpose of each exception, keep an original copy, and test against a small number of real destinations. For the full workflow, see Quick Start: Config and connection steps.

After updating a Config, check more than whether the file saved: verify Global Routing, the selected file, and your test destination. If rules suddenly behave differently, compare the changes first, then check the server; do not immediately reset all Settings. For rule keyword definitions, see the glossary. The recommendations here are not a fixed rule set. Keep Config behavior aligned with your intent, make every exception verifiable, and preserve a clear rollback path for each change.

[On Demand]

4. On Demand connections

On Demand triggers do not guarantee server availability

On Demand triggers connection behavior based on conditions supported by the system. It controls when a connection attempt is made; it does not fix server authentication, update an existing subscription, or ensure a destination is reachable. A working Config, server, and network permission are still required. If a connection repeatedly comes up and drops, first determine whether the conditions are triggering repeatedly or the connection itself is failing. Blaming every reconnect on an unstable toggle can obscure a network or parameter issue.

Support and presentation for On Demand can vary by system. This section focuses on checking settings on iPhone and iPad; use the options shown in your current interface rather than looking for an identical toggle on other platforms. Check the device's existing state instead of assuming a fresh-install default, especially after migrating from an older device. First verify a manual connection without automatic triggers intervening, then decide whether to enable On Demand.

From a manual baseline to On Demand conditions

In Home, confirm the selected server, check Global Routing and Config, then connect manually and visit a real destination. Proceed to On Demand only after this workflow is stable. If the interface lets you set behavior by network type or name, record what you want each condition to do—for example, whether to connect on a trusted network or on mobile data. Do not copy conditions that do not match your network, or rely solely on a network name you cannot verify.

A Wi-Fi network name alone does not prove that a network is trusted; the same name can appear in different places. If behavior depends on network conditions, consider the actual environment and your requirements rather than treating a name match as authentication. On workplace or managed devices, account for network settings and permissions imposed by the administrator. If other network management policies are in place, understand their effect before changing Shadowrocket to avoid conflicting rules that obscure the final state.

Verification after a network change

Test changes that can occur in real use: switch from Wi-Fi to mobile data, return to the original network, and reconnect after a brief loss of service. After each change, first make sure the device can access the internet normally, then check whether Shadowrocket triggered as expected, and finally test the destination. If the underlying network still requires sign-in or has not received a usable address, a connection failure does not mean the On Demand conditions are wrong. Switching repeatedly before the network is ready only adds uncertainty.

Networks that require a web sign-in need separate handling. Complete the network's sign-in first, then test the proxy connection. Repeatedly triggering a connection before signing in may change the connection status while pages remain inaccessible. During testing, avoid updating a subscription, switching Configs, or changing servers at the same time. Record the order and timing of network changes to distinguish a network that is not ready, an unmatched condition, and a triggered connection that gets no response from the remote side.

Background behavior, battery use, and recovery

On Demand does not keep the app in the foreground, nor does it mean that the client actively maintains every background task. System scheduling, network changes, app activity, and retries after connection failures can all affect battery use. If consumption changes noticeably after enabling it, compare connection counts, network switching, and failures under similar usage conditions instead of relying on a single battery percentage. Restore your usual settings after extended troubleshooting.

Set only conditions whose purpose you can explain, and keep a way to disconnect and connect manually. If repeated triggers are unexplained, pause On Demand and check whether manual connection works. If manual connection fails too, investigate the server, DNS, or rules. If manual connection works but automatic behavior does not, restore conditions one at a time to find the smallest set that reproduces the issue. Deleting all network data is not a substitute; it removes useful evidence.

Final checks should cover the expected behavior after a manual disconnect, triggers after network changes, and whether actual requests use the correct policy. If you only need faster access to connection controls, you may not need automatic triggers; see the Widget section below. To route requests by destination, adjust Config rather than broadening On Demand conditions. Separating when to connect from how requests are handled makes settings easier to maintain and prevents rule issues from being mistaken for connection trigger problems.

[Connectivity Test]

5. Ping and connectivity tests

Know what the test actually checks

Ping, Connectivity Test, and loading a web page are not the same thing. Depending on the method, a test may check whether a host responds, whether a port accepts a connection, or whether a request completes over a specified path. A low time value shows only that the part covered by that test responded quickly; it does not represent the speed of every website. Before troubleshooting, check the selected test method, destination, and whether it uses the selected server. Do not compare list numbers alone.

Some networks do not respond to a particular probe even though app requests work. A server address may also accept a connection while authentication or subsequent data transfer fails. So a failed test alone is not a reason to delete a server, and a successful test does not replace an actual access check. Test three layers separately: whether the underlying network works, whether the server connection succeeds, and whether the app request reaches its destination. Record each result instead of summarizing everything as “timeout.”

Test method, destination, and timeout

If the interface offers Ping or another test method, keep the current or default choice as a baseline and read its description. If you change it, use a fixed destination and a method you understand. Choose a stable destination that allows access and fits your use case. Do not use a page that requires sign-in, redirects frequently, or may restrict probes as a universal standard for all servers. A destination may be unreachable because of its response policy or temporary status.

Choose a timeout that suits the network; shorter is not always better. If it is too short, a slow but usable connection may be reported as a failure; if too long, problems take longer to detect. When the interface does not require a change, keep the current default rather than entering an arbitrary value. If you do change it, record the original and compare results over multiple tests. A single fast or slow result may reflect Wi-Fi signal, a mobile network switch, or concurrent activity—not a lasting pattern.

Test layerWhat it can tell youWhat it cannot establish
Host or port responseA basic path is responding nowAuthentication is correct or the destination website works
Request completed through a serverThis server-and-destination combination worksEvery destination and protocol is working
Actual app activityAccess in the current scenarioThe same behavior on other networks

How to make a useful comparison

When comparing servers you already have, keep the device, network, test destination, and method fixed. Stop unnecessary high-traffic activity, then run several tests with pauses between them. Do not switch Wi-Fi networks while comparing results, or rank values from different test methods as if they used the same standard. For a Config that selects policies automatically, also confirm that each test used the same path; otherwise, you may be comparing different handling methods without realizing it.

If a test passes but an app cannot access its destination, check the app destination and matching rule instead of running more speed tests. If only one protocol configuration fails, verify its authentication, transport, and TLS requirements. The same protocol name does not mean its fields are interchangeable: a server hostname, TLS name, and actual connection address may serve different purposes. To identify an existing configuration type, see Protocols supported by Shadowrocket.

From test results to everyday use

Batch tests generate extra requests, and running them frequently does not necessarily improve diagnosis. In everyday use, test a small number of relevant items when an issue occurs and use logs to narrow it down. If the test destination itself is unavailable, choose another and compare again rather than concluding that the entire subscription is unusable. A successful subscription update confirms only that content was retrieved; it does not mean every server entry works on the current network. Treat updating and connectivity testing as separate steps.

After testing, restore the server and Global Routing mode you normally use so a temporary comparison state is not left in place. For recurring issues, record the network, test method, error type, and failed action; this is more useful than a screenshot with only latency numbers. Confirm the fix with a repeatable real task, such as opening the same page and completing the necessary interaction. If only the test button passes while the actual task still fails, continue investigating Log and rules rather than considering the issue resolved.

[Log]

6. Log, Data, and troubleshooting

Use logs to reconstruct what happened

Logs are useful for identifying which step failed, not for collecting as much data as possible. Before troubleshooting, write down how to reproduce the issue: current network, selected Config and server, Global Routing mode, and the destination that fails. Then open or enable the needed log in the current interface and repeat the relevant action. Do not search for one failed entry while many unrelated requests are being generated, especially by opening multiple apps at once.

Log fields vary by record type and platform. Focus on the destination name and address, matched policy, connection status, and error text. First confirm that an entry corresponds to your test, then interpret it. Records close in time are not necessarily from the same request, and a page may call several endpoints at once. A failed entry does not necessarily explain why the whole page failed, and a successful entry does not mean every page resource loaded. Compare the records with what you actually saw.

Troubleshoot by stage, not by guessing from keywords

For name lookup errors, check DNS and network conditions first. For connection timeouts, verify the destination address, port, and path. For authentication failures, return to the server details you already have. For TLS errors, check system time, names, and certificate requirements. Error text is a clue, not proof of a single cause. A timeout can occur between your device and the server or later on the way to the destination. Use surrounding records to locate it instead of changing every setting at the first mention of “timeout.”

For configurations such as Trojan that use TLS, check the password, SNI, and related options against the server details you already have. Do not treat Allow Insecure as a universal fix: skipping verification changes the connection's security boundary and may hide an incorrect name or server-side issue. For field definitions, see Trojan parameters explained. Use logs to identify the relevant field, not to justify loosening settings unrelated to the problem.

ClueNext stepInformation to keep
Name lookup failedCheck DNS and current networkLookup destination and exact error text
Rule result is unexpectedCheck Config and rule orderMatched rule and actual policy
Authentication or TLS failureCheck server details field by fieldProtocol type and related error
Connection dropped mid-sessionCompare network changes with disconnect timingConditions and reproduction steps

Data statistics are not service billing

Interpret traffic or connection statistics in Data according to the scope shown in the current interface. A statistic may cover a time period, a single connection, or app records; it is not necessarily the same as usage reported by a service provider. Tunnel overhead, retries, measurement periods, and server-side accounting can all cause differences. If a value looks unusual, check its scope and whether it was reset, then review background activity. A single cumulative figure cannot show that one destination generated all the traffic.

For a before-and-after comparison, note the starting value, perform one action, and observe the change. Before clearing local statistics, make sure you will not lose diagnostic information you still need. Do not keep resetting the counter to make the number look normal; that removes useful timing evidence. If requests seem to continue in the background, temporarily reduce activity in other apps and compare the readings before and after connections stop. Use request records as well to identify the source.

Privacy and retention of diagnostic records

Logs may contain visited domains, server and local network addresses, subscription details, or request parameters. Before sharing, copy only the relevant lines and remove authentication fields, full subscription URLs, and information that could identify your account. Keep the error type and necessary context. Partially obscuring sensitive text may still reveal its structure; a safer approach is to replace it with clearly identified example values so no one mistakes them for real connection details.

After troubleshooting, return log detail to the level you need for everyday use and clear local records if appropriate. Keeping extensive logs for a long time uses storage and increases the amount of information that needs protection. If the issue remains, prepare a minimal reproduction: before-and-after settings, one action, the corresponding error, and branches already ruled out. Then compare it with the FAQ. This is easier to review and less likely to bury key details than an unfiltered log bundle.

[iCloud]

7. iCloud, backups, and Widget

Sync, backup, and export serve different purposes

iCloud sync transfers app data between eligible devices. A backup preserves a recoverable state, while an export gives you a specific file copy. These are not interchangeable. Deletions and edits made during sync may propagate, so seeing an item on another device does not mean you have a separate historical backup. Before important changes, keep an identifiable local copy and check what it actually contains rather than merely confirming that a sync toggle is on.

The iCloud guidance here applies to devices that support the feature and meet its system requirements; it does not mean every platform uses the same sync mechanism. When setting it up, check the system account, iCloud availability, network, and remaining storage, then review which items the app actually offers to sync. Use the state shown on the device rather than assuming a default, especially after migration or signing in again. Confirm what data is included, then test the sync with a recognizable, non-sensitive change.

What to keep separately

A Config file usually contains rules and related settings, but do not assume an export includes every server, subscription, app preference, and system permission. Check the current interface description before exporting, then inspect the resulting file. Store files containing authentication details securely; do not put them in a public directory or share them publicly. If the file contains only rules, a new device may still need your existing subscription or server details. A successful import does not mean everything has been restored.

Protect the update URL for an existing subscription as well. It may contain account identifiers, and anyone who can access it may be able to retrieve the associated configuration. When backing up, keep the Config copy and subscription source clearly identified, and do not paste the real URL into public feedback. Use example values to explain a format; when restoring, retrieve the actual details from your own protected records. This site explains importing and saving but does not provide subscriptions or server details.

Changes across devices and restore checks

Editing the same Config on two devices at once can lead to overwritten changes, timing differences, or conflict copies. If this happens, stop editing on both devices and save the current contents from each. Choose one as the baseline and merge the needed changes one by one. Do not assume files match because their names are the same, or keep toggling sync in hopes that it will resolve a conflict. Decide which rules, server references, and notes to keep before merging to reduce the risk of losing changes.

When moving to a new device, export a separate copy on the old device first. On the new device, check sync or import results, then test the connection. Check the server list, subscription updates, current Config, Global Routing, On Demand, and necessary system permissions in order. Some system permissions must be granted again on the new device and cannot be restored by copying a file. For detailed steps, see Config backup and device migration. Do not erase the old device until verification is complete.

Widget behavior and limits

Widget displays or provides quick access to app features in locations supported by the system. It is a convenience, not a separate connection engine, and does not replace selecting a server in Home or checking Config. Displayed content, refresh timing, and available actions depend on system and app support. If the status does not update immediately, open Shadowrocket to check the actual connection. The Widget may simply be stale; do not assume the connection has stopped based only on its display.

After adding a Widget, run a controlled check: note the state in the app, use the Widget action, then return to the app and verify the result against a test destination. If On Demand is also enabled, a manual disconnect may meet a trigger condition again, making it seem as if the Widget action did not hold. Check the On Demand conditions before repeatedly removing and re-adding the Widget. The two features have different roles: quick access and condition-based triggers.

You do not need to add a Widget if you do not need quick access; this does not change the rules. If it displays a server name or status you do not want others to see, change what is shown or where the Widget is placed. Keep routine controls straightforward: use the app to check status, the Widget for convenience, and On Demand for condition-based triggers. Each should have a clear purpose; overlapping control methods can make the final state hard to explain.

[Proxy]

8. Proxy ports and local network boundaries

A local port is not a server port

Proxy ports refer to local listening endpoints that provide proxy access to other apps. They are not the same as the remote server connection port in Add Server. Changing a local port does not fix remote authentication or transport settings, and changing the remote port does not change local proxy settings for apps. Before troubleshooting, identify where the request starts, which address and port it connects to, and what protocol that endpoint accepts. Recording all three is more reliable than noting a port number alone.

Not every platform offers the same local listening options; use the names and locations shown in your interface. HTTP and SOCKS5 endpoints are not interchangeable just because their ports are similar. The proxy type set by the client app must match the listening endpoint. Check the current Settings for the default port; this guide does not prescribe one value for every device. If you only use connections on the device and no other app needs an explicit proxy, leave these settings unchanged rather than opening an extra network endpoint for troubleshooting.

The address range determines who can connect

Local-only access and allowing devices on the LAN to connect are different security boundaries. A loopback address is for communication on the same device; a LAN address enables communication with other devices on the network. Listening on a broader address can expose the service to other devices on that network. If the interface offers an option to allow LAN access, identify the users and network first, and check the system firewall and the app's access controls. Being on the same Wi-Fi network does not mean every device should be allowed access.

Limit access to the smallest range you need. If another device you own needs to connect, first confirm both devices are on a network that allows them to communicate. Note the service device's current LAN address, then enter the correct protocol and port on the client device. Do not enter this address as the remote server address, or enter a loopback address on the other device. There, the loopback address points to itself, not to the device running Shadowrocket.

Port conflicts and proxy loops

If the listener will not start, another process may be using the port, the address may be unavailable, or a system permission may be missing. Check the error and existing local services before choosing another port. After changing it, update every app that uses this endpoint; otherwise, they will still connect to the old port. Do not try large numbers of ports or disable unfamiliar system services. Identify the conflict and its purpose before changing anything that could affect other features.

Explicit proxy settings can also create a loop: a request goes to a local endpoint, while the Config routes traffic related to that endpoint back through the same path. This may cause repeated connections, requests that never finish, or unusual resource use. First undo the most recently added proxy setting on the client app and restore a single connection path, then add back only what you need, one layer at a time. Do not enable system proxies, app proxies, and LAN forwarding in multiple places and then guess where requests went.

CheckRecommended baselineFirst step if it fails
Listening addressExpose only the range you needDistinguish local and LAN addresses
Proxy typeMatch the client app to the listenerCheck HTTP or SOCKS5
Listening portKeep the current working value unless there is a conflictCheck port use and the value entered by the client app
Connection pathKnow where each layer leadsRemove the new layer to rule out a loop

Remote resolution and client differences

Whether an app resolves domain names through a proxy depends on the client app's implementation and proxy type. Changing DNS in Shadowrocket does not mean every app using an explicit proxy will follow the same resolution path. If an app on the device works but a client on the LAN does not, check separately whether that client resolves names itself, whether the network allows devices to communicate, whether the listener is running, and whether the request appears in the logs. If there is no corresponding request in the logs, check the entry path before changing the remote server.

A mobile device's network address, background conditions, and connection state can change, so do not treat a temporarily working LAN endpoint as a permanent service. After a network change, check the new address and current connection before testing the client app. When the task is done, disable LAN access if it is no longer needed and remove any proxy settings left in the client app, so it does not keep trying to connect to an unavailable endpoint. On an untrusted network, do not expose access beyond what you need.

Review settings in dependency order

Review settings in dependency order: check the underlying network and your existing server details first, then Config, Global Routing, and DNS; next check On Demand; finally verify auxiliary entry points such as Widget or local Proxy. Use the same destination at each step and return to the most recent working state when something fails. To reinstall the client or check system permission steps, see the Download page installation guide. For everyday operation, return to the Quick Start—there is no need to reset all settings.