Official Guide · Protocol Reference

Shadowrocket Protocol Guide
Settings and Use Cases

First confirm which protocol the server provides, then check the identity, transport, and routing settings. A protocol name alone does not determine connection speed or provide all the required settings.

This reference is for looking up settings, understanding protocol trade-offs, and diagnosing problems. If you’re new to Shadowrocket, follow the Getting Started guide to import a subscription and connect, then return to the relevant section. The instructions assume you already have your own subscription or server details. This site does not provide, sell, or recommend proxy servers or subscriptions.

Before choosing a protocol: connection layers

Start with the server configuration

The protocol configured in Shadowrocket must match the one running on the remote server. A subscription usually includes the protocol type, address, port, and authentication details. When adding a server manually, enter each field in the corresponding Add Server setting. Don’t switch an existing entry to a different protocol just because its name sounds newer. If the server isn’t running that protocol, the client may save the entry but won’t be able to complete the handshake. Choose among the connection options you actually have; changing a protocol label won’t make a server support a different one.

A working connection has several layers. The server address identifies which host to contact; the port specifies which service to reach; the protocol defines how requests and identity are exchanged; and the transport carries the data. TLS and similar mechanisms provide encryption and certificate verification at their respective layers. Routing rules then determine which requests use the connection. Keeping these layers separate explains why an address can resolve and a port can accept a connection while webpages still fail to load.

Three separate checks

First, compatibility: what the server supports and whether the client entry is complete. Second, network conditions: whether the current network can carry the required TCP or UDP traffic, and what bandwidth and packet loss look like along the route. Third, your needs: short requests, sustained downloads, interactive apps, and private network access have different requirements. If any one layer is missing, repeatedly changing encryption methods or increasing concurrency won’t fix it. Find the specific missing requirement first.

LayerTypical settingsCheck first
DestinationServer、Port、EndpointAddress, port, DNS resolution
Identity and protocolPassword、UUID、PublicKeyProtocol type, credentials, server configuration
Transport and securityTCP、UDP、TLS、WebSocketTransport settings, certificates, network path
Traffic handlingConfig、Proxy、DirectMatching rules and routing policy

Subscriptions and configuration: what to expect

A subscription is a way to distribute connection details in bulk, not a separate proxy protocol. Its entries may use different protocols or provide multiple endpoints for the same protocol. A configuration file can also include rules, DNS settings, and other behavior. After updating a subscription, check whether the selected entry is still there and whether any settings have changed. The same subscription name doesn’t guarantee that the underlying settings are unchanged, and a successful update doesn’t verify that every connection works.

Importing a link or scanning a QR code saves manual entry, but doesn’t prove that the information it contains is valid. If a field is unclear, keep the original details and ask the provider of your existing service what it means and which client setting it maps to. Don’t post full subscription links, passwords, private keys, or other access credentials in public forums. Troubleshooting usually requires only the protocol, transport, failure stage, and redacted field structure—not your credentials.

The client’s one-time purchase is separate from any service plan. The iPhone and iPad app costs $2.99 as a one-time App Store purchase. The Mac, Windows, Android, and Linux versions are free. Client access costs and fees for any server or subscription you already have are separate.

This guide distinguishes protocol capabilities from the service configuration you actually have. A protocol’s ability to carry UDP doesn’t mean the current server has enabled it. A transport that can reduce connection steps in theory won’t necessarily be faster on your network. First set up a working entry with matching fields using the chapters below. Then compare it under the same conditions, using the same destination and network, to get useful results.

Shadowsocks: shared keys and a lightweight connection

Background and connection model

Shadowsocks is designed to encrypt and forward proxy traffic through a relatively lightweight connection. It sends client requests to a remote server for handling. It isn’t a full network-layer tunnel, and a server entry alone doesn’t determine where all your traffic goes. In Shadowrocket, Global Routing and configuration rules determine which requests use the connection. This distinction helps avoid mistaking a rule that didn’t match for a protocol failure.

Traditional deployments use a password and encryption method agreed on by both sides. The server uses these to identify and process traffic; the client doesn’t go through the same kind of login as a website account. Newer protocol variants may differ in key format, data encapsulation, and replay protection. As a result, an older Shadowsocks entry may not migrate by simply copying its password. Preserve the exact encryption method name when importing, and don’t assume similar-looking options are interchangeable.

Address, port, password, and encryption method

Server is the proxy server’s domain or address, not the website you want to visit. Port is the port the proxy service listens on. Password must match the remote server; watch for spaces, line breaks, or substituted characters when copying it. Method specifies how both sides protect the data, and must match the server’s actual configuration—not whichever option appears first on your device. Some methods require an encoded key of a specific length; an arbitrary short passphrase won’t work.

If a subscription includes additional transport or extension settings, the main password alone may not be enough to connect. Check whether the current server requires those settings and whether they were imported into the right fields. Don’t combine settings from other entries. If your service details list only the server, port, password, and method, start with those known values. If required settings are missing, ask the provider of your existing service for the complete details rather than guessing.

Speed and battery use on iOS

Lightweight encapsulation can reduce some protocol processing, but actual performance on a phone also depends on processor encryption support, server load, and the wireless network. During sustained downloads, encryption is often not the only bandwidth bottleneck. Short requests are more sensitive to connection setup and remote DNS resolution. A record that downloads quickly may not make webpages load faster the first time. Compare first response time, sustained transfer, and connection stability separately instead of relying on a single speed figure.

UDP availability must be confirmed across the entire connection path. The client must support the required handling, the server must allow forwarding, and the destination must respond. If webpages work but a real-time app doesn’t, first check whether it relies on UDP, then verify server permissions and the network path. A password is rarely the sole explanation for every failure. Likewise, a successful TCP request doesn’t prove that UDP forwarding works.

Troubleshooting checklist

If the connection doesn’t work at all after saving, first check DNS resolution and whether the port is reachable, then compare Method and Password character by character. A connection that starts and drops quickly may indicate a method mismatch, an incorrectly formatted key, or a server rejection; use the error details to narrow it down. Don’t change several settings at once. After each change, retest against the same known-working destination and keep the original entry so you can revert.

Shadowsocks is suitable when you already have a matching service and want to connect using clearly specified shared settings. If you need access to private network ranges, first find out whether your service provides an application proxy or network-layer access. Don’t assume importing an entry gives you routes to the entire remote network. To switch connection types, you’ll need complete settings for another service. Renaming an entry or changing its icon won’t change what the remote server provides.

A useful check order is Server → Port → Method → Password → additional transport settings → UDP availability. After connecting, check Config to see which rule matched the request.

VMess and VLESS: separate identity from transport

Similar fields don’t mean the same protocol

VMess combines identity verification and data handling in the protocol. It commonly uses a UUID to identify an authorized user and has its own authentication and data protection mechanisms. VLESS uses a more lightweight identity and request format; the protocol itself doesn’t provide complete transport encryption, so it’s generally paired with an appropriate secure transport. The two protocols often share similar-looking server fields, but use different handshakes. Changing a VMess entry to VLESS won’t make the server accept it automatically.

When working with these protocols, keep two lists: which identity protocol is used, and how the connection is transported. One protocol may work with different transport settings, and different protocols may both use TLS. The complete combination is what matters—not the protocol name alone. If you change the protocol, transport, and server at the same time, you can’t attribute the result to VMess or VLESS alone; you’re comparing two complete connections.

UUID, time, and legacy settings

The UUID identifies the user and must match the server configuration exactly. It isn’t the server address or certificate domain. Even two entries connecting to the same domain may use different identities. VMess authentication depends on the correct time, so a significantly inaccurate device clock can cause failures. Check that the system date and time are correct rather than adding a guessed offset to the protocol settings. Once the clock is correct, check the credentials and server status.

Older setup information may include additional compatibility fields. Don’t apply a fixed value from an old guide to every entry, or remove settings that still have a clear purpose just to “simplify” things. If the imported entry differs from the original details, confirm which authentication method the service currently uses. For VLESS, preserve any additional mode specified by the server. Authentication only works when the complete combination matches; checking the UUID alone isn’t enough.

TLS, Server Name, and WebSocket

TLS Server Name is used during the handshake and certificate identity check, and usually needs to be the domain specified in the service configuration. The connection address and certificate domain can differ in some deployments, but that difference must be supported by the server configuration. WebSocket Host and Path are used by the application-layer request and routing; the leading slash, path capitalization, and any additional content can all affect matching. Setting Server Name, Host, and Server to the same value isn’t a universal fix.

With a reverse proxy, a reachable port only means the front-end accepted the connection; it doesn’t confirm that the request reached the right backend. If the handshake still fails after TLS succeeds, check Host, Path, and the backend protocol. Receiving an ordinary webpage may mean the request reached the default site, not that the identity is wrong. Repeatedly changing the UUID can break credentials that were already correct. First confirm that the transport request is reaching the intended service.

Field groupPurposeCommon mistake
Type、UUIDProtocol and authorized identityChanging Type while keeping settings for the other protocol
TLS、Server NameSecure transport and server identityMistaking a certificate error for a UUID error
Host、PathTransport request and backend routingLeaving out the path or changing the domain at random

Trade-offs and when to use each

VLESS’s lightweight design doesn’t mean it always uses less battery. Additional handshakes, long keepalives, or frequent reconnects can use more energy than the protocol saves in processing. VMess’s built-in data protection doesn’t mean the overall connection is necessarily slower. For phone users, a clearly configured combination that matches the server and can reuse connections reliably is generally more useful than repeatedly changing protocol names.

For an entry imported from a complete subscription, keep the original transport settings and verify the connection before trying other entries. Adjust both remote and local protocol settings at the same time only if you manage the server and understand the full forwarding path. To narrow down a failure, check in this order: time and identity → TLS → transport path → backend service. Look for evidence at each stage instead of re-entering every field at once.

Trojan, HTTP, and SOCKS5: security and proxy functions

Trojan identity and TLS

Trojan carries proxy requests over a TLS connection and uses a password for authentication. Its design relies heavily on TLS for transport security, so certificate validation, domain matching, and device time are essential connection requirements—not optional extras. Server and Port specify where to connect, Password determines whether access is allowed, and Server Name is used to verify the TLS identity. Although these fields appear on the same screen, they’re checked at different stages.

If certificate validation fails, check for a mistyped domain, an expired certificate, an incomplete certificate chain, or an incorrect system clock. If you connect by address but the certificate is issued for a domain, enter the correct Server Name from your service details. Don’t routinely disable certificate verification: that removes a server identity check without fixing a wrong domain, certificate, or backend. The connection may still fail at another stage.

HTTP proxies and HTTPS destinations

HTTP proxies were designed to forward application-layer requests. For an HTTPS destination, a common method is to use CONNECT to create a tunnel, then let the app establish its own TLS connection to the destination. These are two separate connections: whether the client-to-proxy leg is encrypted, and whether the app-to-destination leg uses HTTPS. An HTTPS website doesn’t mean the connection to the proxy is protected by TLS. Conversely, TLS on the proxy connection doesn’t verify the destination website’s content.

For an HTTP entry, first confirm whether the remote service offers a standard HTTP proxy or a TLS-protected proxy endpoint. Then check the address, port, and whether a username and password are required. Some proxies restrict the destination ports available through CONNECT or allow only certain destinations. That can make some sites work while others fail. These are service-side policies; changing DNS or arbitrarily switching to a common port won’t fix them.

What SOCKS5 does—and doesn’t do

SOCKS5 defines connection requests, optional authentication, and UDP association. It doesn’t provide general-purpose transport encryption. A username and password authenticate you; they don’t protect the entire connection the way TLS does. SOCKS5 can provide a clear proxy interface on a controlled local network or over an additional secure channel. If using it over an untrusted network, check whether the existing setup provides separate protection. A password field alone doesn’t mean the security layer is complete.

SOCKS5 can support UDP association, but that doesn’t mean every server enables it. Check server support, data port reachability, and whether the network allows traffic in both directions. Where a domain is resolved also affects behavior: if the request is converted to an address before it reaches the proxy, the remote side can’t resolve the original domain. If a domain fails but its address works, include DNS resolution in your checks rather than looking only at authentication.

TypeAuthentication fieldsSecurity notes
TrojanPasswordCheck the TLS certificate and Server Name
HTTPOptional username and passwordConfirm separately whether the endpoint uses TLS
SOCKS5Optional username and passwordThe protocol itself doesn’t provide general-purpose transport encryption

Choosing among services you already have

If your existing service explicitly provides Trojan, import it using the supplied certificate and authentication details. Don’t switch to HTTP just to reduce the number of fields. If your workplace or home network provides a standard HTTP or SOCKS5 endpoint, use the matching type and confirm its access scope and security requirements. Choose a protocol that matches the actual service interface. Entering a working SOCKS5 address as a Trojan entry won’t add TLS to that service.

There’s no fixed battery-use ranking for these connection types on iPhone and iPad. The frequency of certificate handshakes, proxy idle timeout, reconnects after network changes, and app request patterns all affect the result. When troubleshooting, note whether the failure happens during connection setup, authentication, or access to the destination, then retest against the same destination. This helps distinguish service policies, certificate problems, and transport failures instead of blaming every issue on the protocol being unsuitable for phones.

WireGuard: network tunnels and key pairing

How it differs from an application proxy

WireGuard is designed to create an encrypted network-layer tunnel using a relatively lightweight protocol. It handles IP packets rather than simply asking a proxy to connect an app to a website. Its configuration therefore includes the remote endpoint and identity as well as a local tunnel address, peer keys, and the address ranges allowed through the peer. Choose it based on the network you need to reach and your routing requirements—not as just another server type that takes a password.

WireGuard usually runs over UDP. The current network must allow UDP traffic to and from the Endpoint, and the server must be configured for the appropriate listening and forwarding. Successful import only means the configuration could be read; it doesn’t prove that the data path is working. If the handshake completes but you can’t reach the destination, check the assigned addresses, peer allowed ranges, server forwarding, and return route. These are separate from checking the keys.

PrivateKey, PublicKey, and Address

PrivateKey is the local private key. Keep it on the device it belongs to or in a secure configuration. In the peer settings, PublicKey usually refers to the remote peer’s public key—not your own public key entered again. The server must have the local public key that corresponds to your private key so both sides can authenticate each other. If you generate a new key for a device without updating the server, the old connection won’t work even if every other setting is unchanged.

Address is the local tunnel address and prefix—not the server’s public address or the address assigned to your device on its current Wi-Fi network. It must come from the existing network plan, not an address picked at random because it looks unused. Reusing the same identity and address on two devices can cause peer state to overwrite or routing conflicts. Assign each device its own identity and address, and confirm the server has the right range configured for each peer.

AllowedIPs, Endpoint, and DNS

In the protocol model, AllowedIPs helps select which destination addresses go through the peer and restricts which source addresses are accepted from it. It’s not a list of website domains, so entering domains won’t enable domain-based routing. Endpoint is the reachable remote address and port; Address is an internal tunnel address. They’re both address fields, but serve different purposes. DNS affects how names are resolved to destination addresses; it doesn’t replace the network ranges specified by AllowedIPs.

Shadowrocket’s traffic handling also depends on both the client’s routing rules and the ranges allowed by the tunnel peer. A rule can send a request through a connection, but that doesn’t guarantee the remote side allows access to every address. To reach a private network, check that the destination is in the authorized range, the server has a route to that network, and return traffic can reach the tunnel. A one-way route may show connection activity while the app waits and eventually times out.

MTU and PersistentKeepalive

MTU sets the maximum packet size the tunnel interface can carry. If it’s too large, fragmentation or path-related issues can result in small requests working while larger responses stall. Test gradually based on the network path and service configuration; don’t assume a commonly suggested value is best for every network. After changing MTU, test both large responses and two-way transfers. Loading one short webpage isn’t enough to confirm that a packet-size problem is fixed.

PersistentKeepalive periodically sends keepalive packets when needed to maintain mappings on stateful network devices. It isn’t a speed boost. An overly frequent interval can wake the mobile network and increase background activity; without it, some networks may become unreachable after a period of inactivity. Set it based on whether the server needs to reach the device, how long mappings stay idle, and the symptoms you’re seeing—not by always enabling the shortest interval for “stability.”

WireGuard is suitable when you have a complete peer configuration and need access to specific network ranges. For more on how the fields map to each other, see the WireGuard settings guide. A standard proxy subscription without the peer’s public key, local address, and authorization settings isn’t enough to set up WireGuard. Use the protocol your service actually provides rather than assembling a tunnel entry that only looks complete.

Hysteria2: QUIC transport and bandwidth settings

Design goals and requirements

Hysteria2 uses a QUIC-based transport designed for data transfer over networks with packet loss or high round-trip times. QUIC runs over UDP and has its own reliable transport and congestion-control mechanisms; it doesn’t send data without confirmation. Its trade-off is a different approach to transport control, not the removal of network capacity limits. Results still depend on server capacity, path quality, and local network conditions.

First confirm that the current network allows the required UDP traffic. If the network restricts UDP, a working TCP service on the same server doesn’t mean Hysteria2 should also work. Conversely, UDP reachability only confirms the first requirement; TLS and authentication must still succeed. Troubleshoot in stages: DNS resolution, UDP path, certificate, authentication, and destination forwarding. Don’t change every setting at once when you see a timeout.

Server, authentication, and certificate settings

Server and Port specify the endpoint. Use the authentication details provided by your existing service; don’t substitute a UUID or password from another protocol. TLS Server Name must match the service configuration and certificate. If your service details include an additional obfuscation setting, make sure its type and password match on both sides. The authentication password and obfuscation password serve different stages, even if both are called passwords; don’t combine them into one value.

Hysteria2 has compatibility limits with the older Hysteria protocol. Similar names don’t mean their handshakes are interchangeable. After importing, confirm that the type is Hysteria2 and that the client recognized all the fields. If the imported entry doesn’t match your service details, keep the original and check the format. Don’t switch to a similarly named type just to clear an error. That may hide a protocol mismatch rather than fix it.

Bandwidth settings aren’t speed test results

Some deployments provide upload and download bandwidth settings for transport control or related negotiation. Their effect depends on the connection configuration and server policy, so follow the service’s instructions. Entering values far above the actual link capacity won’t give the wireless network or server more bandwidth; with some control methods, it may increase queuing, packet loss, or burst traffic. A value that’s too low can also limit throughput. Don’t inflate it at random or copy a value from another device.

Check bandwidth units carefully: bits per second and bytes per second are different, as are upload and download directions. A phone’s connection speed also varies with signal, network load, and movement. A single speed-test peak isn’t a reliable permanent setting. If the service doesn’t require a manual value, keep the imported setting. If you need to adjust it, change one direction at a time and check whether actual throughput and app response both improve.

Packet loss, congestion, and battery use

On a network with packet loss, suitable transport control may help sustained transfers use the remaining capacity. But sending more aggressively won’t fix every kind of packet loss. If loss comes from congestion at the access point or on the carrier network, more traffic may worsen queuing. For interactive apps, faster downloads with longer response times aren’t necessarily an improvement. During tests, track page response, sustained transfer, and recovery after a network change.

Phone battery use depends on encryption, packet processing, and wireless activity over time. Higher throughput may finish a task sooner, but it can also increase processor load and retransmissions. Using UDP doesn’t automatically mean lower battery use. Heat, background sync, and signal strength can affect results, so keep the test duration and workload consistent. For everyday use with few requests, also observe what happens after an idle period instead of testing only continuous downloads.

Hysteria2 is suitable when you already have a matching service and UDP availability is clear. If the current network repeatedly blocks UDP, choose another complete connection that’s available through your existing service rather than continually increasing bandwidth settings. Keep a verified baseline entry to tell whether a problem comes from a network change or a new configuration. Retest after switching networks before deciding whether to adjust the client settings.

iOS performance: speed, battery use, and testing

Measure different kinds of speed

On iPhone and iPad, the experience involves at least four stages: DNS resolution, initial connection, first response, and sustained transfer. A short webpage may depend mostly on DNS and handshake time; a large download is more likely to be limited by path throughput and server capacity. Video, voice, and other interactive apps also depend on jitter and interruptions. Reducing these tasks to one speed ranking can lead you to trade everyday responsiveness for a higher download peak, or mistake destination server load for a protocol difference.

Connectivity Test results reflect a specific method and destination; they don’t replace testing your apps. A shorter response time for one entry only shows that the test completed faster under those conditions. It doesn’t prove higher large-file throughput, working UDP forwarding, or more reliable background recovery. First check exactly what the test measures, then choose a test that matches your everyday use. Don’t draw conclusions beyond what a single result can show.

Set up a repeatable comparison

Keep the device, network, destination, and task consistent. Pause unrelated high-bandwidth sync and try to maintain similar signal strength and device temperature. Change only the connection entry, keep all other settings the same, and alternate between entries over several rounds to reduce the effect of changing server load. If the entries use different servers, record the result as a comparison of two service paths—not as a comparison of the protocols themselves. This matters more than which speed-test tool you use.

Also distinguish a cold start from a reused connection. The first visit may include DNS resolution and a full handshake; later visits may reuse a connection or cache, so response times will naturally differ. A fair comparison uses the same state for both entries, not the first entry’s initial visit against the second entry’s cached visit. To test recovery after idle time, wait the same amount of time before sending a request. After a network change, record both how long the connection was interrupted and whether the task completes after recovery.

A protocol name can’t predict battery use

Battery use includes the display, wireless radio, processor, and background apps. Sustained transfers keep the wireless radio active; frequent small requests, keepalives, and reconnects can also increase wakeups. More complex encapsulation may require more processing, but a more stable path with fewer retransmissions may still use less power overall. Don’t claim that one protocol uses the least battery on every iOS device without accounting for workload, connection state, and network quality.

When checking Battery in Settings, consider activity time, foreground and background use, and what other apps are doing. System statistics for network extensions aren’t a direct measurement of power use by an individual protocol. Be especially cautious about small differences: higher temperature, weaker signal, or background system maintenance may have a greater effect than protocol processing. Without repeatable evidence, describe results as observations for the task tested, not universal rules.

Test taskKeep consistentWhat to observe
First page loadDestination, cache state, networkDNS resolution, handshake, first response
Sustained transferFile, duration, device temperatureThroughput, signs of retransmission, stability
Request after idleWait time, background conditionsReconnects and recovery time
Everyday battery useWorkload, signal, screen useActivity time and recurring patterns

On Demand and connection recovery

On Demand triggers connection behavior based on configured conditions. It doesn’t change the protocol used by the server or fix missing authentication settings. First verify the connection manually, then check that the trigger conditions match your actual network. An unstable network combined with frequent triggers can cause repeated connection attempts. Check the trigger conditions and failure reason instead of assuming that keeping the connection active is the only solution. After changing automatic behavior, test what happens when the device is locked and unlocked.

Switching between Wi-Fi and mobile data can change the source address, network mappings, and available transport options. Some connections may recover, but that doesn’t guarantee that every active request will continue; the app may need to retry too. To test recovery, check both the client connection status and whether app requests succeed. If failures happen consistently after switching networks, compare how each network handles TCP, UDP, and DNS before considering a different protocol combination from your existing service.

Choosing a connection, working with rules, and troubleshooting by layer

Narrow down your options based on what you have

For everyday browsing and common apps, start with a complete, stable entry from your existing service, then compare short requests with sustained transfers. For controlled private networks, check whether you need network-layer access such as WireGuard and whether the remote side authorizes the relevant ranges. If the network clearly restricts UDP, don’t use a UDP-dependent connection as your baseline. If your service offers only one protocol, configuring it correctly is more useful than trying to use a service that doesn’t exist.

After choosing a protocol, confirm that requests are actually using it. In Global Routing, Config applies the rules in your configuration, Proxy sends proxyable traffic through the selected proxy path, and Direct connects directly. System traffic and exclusions may also depend on client and operating system behavior. Temporarily switching modes can help identify a rule issue, but restore your original routing setup afterward so the test state doesn’t become permanent by accident.

Rules are evaluated in order

Rules in the configuration file’s [Rule] section are checked in order, and the first match determines the policy. DOMAIN matches a specific domain; DOMAIN-SUFFIX matches a domain and its subdomains; DOMAIN-KEYWORD matches a keyword in the domain; IP-CIDR and IP-CIDR6 match address ranges; GEOIP matches a location database classification; and FINAL handles requests that didn’t match earlier rules. Domain and IP rules use different criteria, and the actual match depends on the resolution result and the request information available.

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

This example illustrates syntax and ordering; it isn’t a complete configuration for every network. The first line rejects a specific domain, and the suffix rule that follows handles other matching requests. If you reverse the order, the broader suffix condition may catch the request first. Direct connections to an IP range also require the device to be able to reach that network; a rule won’t create a route to a remote local network. Save the original configuration before making changes, change one thing at a time, and check what actually matches.

Check each layer from the entry to the destination

First, check the entry: do the type, address, port, identity, and transport match the original service details? Second, check the endpoint: did DNS resolution succeed, and does the current network allow the required transport? Third, check the handshake: are the certificate, time, authentication, and transport path valid? Fourth, check the request: are DNS resolution, routing policy, server forwarding, and the return path working? A reachable endpoint doesn’t prove that the proxy handshake succeeded, and a successful handshake doesn’t mean every destination is reachable.

If all entries stop working at once, check the current network and device time first, then see whether a subscription update changed the entries. If just one entry fails, compare its settings with the remote server status. If only one domain fails, check the rules and DNS before changing protocols. If the list is empty after importing, follow the subscription import and empty-list troubleshooting guide to check the link format and update conditions; this issue occurs before a connection is established.

Keep evidence and a way to revert

When reporting a problem to the provider of your existing service, include the approximate time, network, protocol, failure stage, and a redacted error message. Don’t send a full subscription, PrivateKey, or password. Keep a verified connection and a copy of the original configuration so you can restore a working setup if changes fail. For more on the three routing modes, see the Global Routing guide. If you’re new to setup, return to the Getting Started guide and follow the steps in order.

For the client, see the download and installation guide. Check the App Store page for iPhone and iPad system requirements. Mac versions are available for Apple Silicon and Intel; the download page covers installation and first-launch steps for Windows, Android, and Linux.

The final configuration should be easy to explain: why you chose the protocol, which settings came from your existing service, which requests the rules send through it, and where to start checking if it fails. Understanding these relationships makes it easier to respond to network changes. Use this protocol reference to identify requirements and limitations, not to assume that newer names perform better or that one successful connection works in every situation.

Download Shadowrocket