Official Manual · Technical Reference

Shadowrocket Protocol Guide

protocol-reference

Choose a protocol that matches your existing server, then verify its transport, authentication, and encryption settings. Protocol names are not speed rankings.

Use this page as a technical reference to understand protocol design and what happens when settings do not match. For instructions on adding an existing subscription, connecting, and checking the result, start with the user guide. Return here to review fields and troubleshoot edge cases once you're familiar with the basics.

Shadowsocks VMess / VLESS Trojan HTTP / SOCKS5 WireGuard Hysteria2

The guidance below assumes you already have your own subscription or server details. Shadowrocket offers clients for multiple platforms; the examples focus mainly on iPhone and iPad. Available options depend on the device interface and server configuration.

Understand protocols, transport, and routing separately

layers

A connection involves several independent decisions

In Shadowrocket, choosing a server protocol determines which communication method the client and server use. The original connection from an app, the connection encapsulated by the client, and the connection the server makes to the destination are related but distinct parts of the route. Opening a webpage also involves DNS lookup, connection setup, authentication, and data transfer—none of which can be summed up by a protocol name alone. Understanding these layers helps you identify which setting to change instead of switching protocols repeatedly when a connection fails.

Start by separating a configuration into four layers: the protocol defines authentication and data encapsulation; the transport carries the encapsulated data between endpoints; TLS encrypts the relevant connection and verifies identity; and rules determine whether a destination uses this server. Not every protocol exposes all four layers as options. WireGuard, for example, has its own encryption and handshake mechanisms, so TLS-based server fields do not apply. With VLESS, pay particular attention to the security mechanism used by the outer layer. A field missing from a control panel is not necessarily a missing security feature.

There is no fixed relationship between protocol, speed, and battery use

Protocols are designed for different needs: lightweight application proxying, flexible combinations of authentication and transport, general-purpose proxy negotiation, network-layer tunnels, or carrying data over high-latency and lossy links. These goals are not interchangeable. A low-overhead protocol can still be slow on a congested route, while a more complex one may respond faster when the server is closer and the network is more stable. For a fair comparison, keep the server location, local network, destination, and test period consistent; otherwise, you are mostly measuring differences in the route.

On iOS, battery use depends on more than encryption and decryption. It also includes keeping the wireless radio active, handshake retries, background keepalives, DNS lookups, and frequent reconnections. Many short-lived failed connections can use more power than one sustained transfer. The system's VPN indicator only shows that the tunnel interface is active; it does not confirm that a remote server authenticated successfully or that all traffic is using the proxy. Check destination requests, connection logs, and the rules that matched instead of treating the icon as a complete connectivity report.

LayerQuestion to answerCommon checks
ProtocolHow do the endpoints authenticate and encapsulate data?Type, password, UUID, key
TransportHow is protocol data carried?TCP, UDP, WebSocket, path
SecurityHow is data encrypted and the peer verified?TLS, Server Name, certificate
RoutingWhich destinations use this server?Global Routing, rule order

Check server capabilities before choosing a protocol

Changing Type in Add Server does not make an existing server switch protocols. The protocol, port, and credentials must match a configuration the server actually supports. Even the same address may host entirely different services on different ports. If an imported subscription contains an unfamiliar type, check the settings for the service you already use instead of guessing from the name. Save the original configuration before editing and note why you made each change so you can return to a known, testable state.

Here, choosing a protocol means selecting among configurations you already have, based on your needs, and knowing which fields cannot be interchanged. A one-time client purchase is not a service plan. The iPhone / iPad version costs $2.99 as a one-time App Store purchase; the Mac (Apple silicon and Intel), Windows, Android, and Linux versions are free. The client price does not include remote server service. See the download page for all platforms. This page does not provide or recommend subscriptions or servers.

Shadowsocks: lightweight encapsulation and matching settings

shadowsocks

Designed to encrypt and relay proxy traffic

Shadowsocks encrypts application proxy requests and sends them to a remote server for processing. Its core settings typically include the server address, port, encryption method, and password. Compared with combinations that require more outer-layer negotiation, its basic structure is relatively simple, making it useful for understanding the basic relationship between a client and server. But simple does not mean every configuration labeled Shadowsocks is interchangeable: the encryption method, protocol generation, extensions, and server implementation all affect whether the client can communicate correctly.

Modern configurations typically use AEAD (authenticated encryption with associated data) to protect confidentiality and detect tampering. The encryption method is not a speed setting; it must exactly match the server. If the server uses one method and the client another, communication will fail even if the password is correct. Different devices may process algorithms at different speeds, but only test alternatives when the server explicitly supports them. Do not swap algorithms arbitrarily within a single configuration.

Check the password, port, and extensions separately

When adding a server manually, first confirm Type is set to Shadowsocks, then check the server address, port, encryption method, and password one by one. Password capitalization, symbols, and leading or trailing spaces all matter for authentication; watch for line breaks and full-width characters when copying. Some newer configurations use key material with specific encoding and length requirements. Do not replace it with an easy-to-remember phrase or remove trailing characters because it looks like an ordinary password. Keep the complete value exactly as provided by your service.

A port is simply the number used to reach a service on a server; it does not inherently identify a protocol or indicate whether TLS is enabled. Using a common web port does not mean you should enable TLS. If a configuration includes extensions or additional transport settings, confirm that the client recognizes them and understands their meaning. Seeing only an address and password after import does not prove that every setting was preserved. Compare the entry with the original instructions, paying particular attention to options hidden under advanced settings.

Working TCP does not mean UDP works

Shadowsocks can carry different types of traffic, but whether UDP works depends on server support, network conditions, and the relevant client settings. A webpage may load while a real-time app fails because they use different transport capabilities. Enabling a UDP option cannot give UDP support to a port that only serves TCP. First confirm what your existing configuration supports, then test with a specific app or controlled request rather than relying on a single result in the server list.

On iPhone, basic Shadowsocks encapsulation is generally lightweight, but reconnects and weak signal can still increase battery use. If an app stops responding after switching from Wi-Fi to cellular, first check whether the connection was re-established and whether the remote address is reachable; then review the protocol settings. Don't conclude that another protocol uses less power just because switching to it restored service. The switch itself may clear stale connection state and trigger a fresh DNS lookup and handshake.

Keep a baseline you can restore

If a Shadowsocks entry in your subscription connects successfully, keep it as a baseline for troubleshooting, but avoid casually editing core fields in a managed subscription. A subscription update may overwrite manual changes, making a working setup revert later. To compare settings, create a separate test entry and label its purpose, but only if you understand what the credentials are for. When sharing troubleshooting details, hide passwords, keys, and the full subscription URL; include only the protocol type, field names, and stage where the error occurs.

Choose Shadowsocks when you already have a complete configuration supported by the server and have confirmed the TCP and UDP capabilities you need. Fewer options do not make it a better fit for every situation, and missing TLS fields are not a reason to add them manually. Verify the basic connection first, then check routing, the target app, and behavior when switching networks. This helps distinguish a routing issue from a protocol problem.

VMess and VLESS: authentication and outer-layer security

vmess-vless

Similar names, different security responsibilities

VMess incorporates user authentication and data protection into its protocol design. Common configurations identify users with a UUID and use a specific transport. VLESS has a more minimal protocol layer and does not provide the same payload encryption mechanism as VMess; it typically relies on an outer security mechanism to protect the transport. These are not the same setting under different names. Do not switch between them in Type just because the server address and UUID look alike.

VLESS is minimal in terms of protocol responsibilities; that does not necessarily make the whole connection lighter or provide encryption by default. If the outer layer uses TLS, the corresponding handshake and certificate verification still apply. If the server uses another supported security combination, configure the complete combination as specified. A displayed encryption value of none does not by itself prove that traffic is sent in plaintext, just as choosing VLESS does not automatically make a connection secure. Check both the protocol and the outer layer.

A UUID is neither a server address nor a certificate name

A UUID identifies a user and must be kept intact, including its hyphens and every character. It usually corresponds to a user configured on the server; do not put it in a password, path, or Host field. The server address tells the client where to connect, while the UUID identifies the user. The TLS Server Name is used for peer verification and service selection. Even if all three come from the same subscription, they serve different purposes and cannot be swapped for compatibility.

VMess authentication depends on the correct time. If the device clock is significantly wrong, a connection may fail even when the address and UUID are correct. First make sure the system date and time zone are accurate instead of blindly changing server fields. Older configurations may also include additional authentication parameters, whose relevance depends on what both endpoints support. Do not copy fixed values from outdated guides. Prefer importing and retaining all the parameters provided in your existing subscription.

CheckVMessVLESS
IdentityCheck the UUID and authentication settingsCheck the UUID and any server-required options
Protocol-layer encryptionHas its own data protection mechanismDoes not provide the same payload encryption
Outer-layer settingsConfigure for the actual transport and security setupConfirm the complete outer-layer security setup
Switching requirementsThe server must provide the corresponding endpoint; changing Type alone is not enough

Treat the WebSocket path and TLS name as separate settings

If the configuration uses WebSocket, the path usually needs the correct leading slash and must match the server's routing requirements. A path is not an arbitrary note; changing it can send the request to a different handler. Host is the hostname in an application-layer request, while Server Name is the name-related setting in the TLS handshake. They may be the same, but deployment details can make them different. Check each field separately rather than combining them because both look like domain names.

The connection address can be a domain name or an IP address. Connecting by IP does not mean the certificate should be verified against that IP; if the certificate is issued for a domain, you generally still need the Server Name that matches the deployment. For a certificate name error, check the name, certificate validity period, and system time rather than leaving verification disabled. Disabling verification changes the trust boundary and cannot fix an incorrect UUID, WebSocket path, or server route.

Compare complete configurations, not protocol names

To compare VMess and VLESS on iOS, record the complete setup—protocol, transport, security method, and server location—not just the two names. Loading a webpage once may mostly measure DNS and TLS handshake time, while a sustained download is more affected by bandwidth and congestion. Short-lived connections, long-lived connections, and network switches create different workloads, so results may vary. Run a controlled comparison only if your existing service provides equivalent endpoints.

For more detail on how the fields differ, see VMess and VLESS: encryption, overhead, and settings. In practice, a complete configuration you can verify matters more than a theoretically simpler protocol name. If an imported entry contains unrecognized options, first check what the current client and server support. Don't delete unfamiliar fields and assume what remains is equivalent.

Trojan, HTTP, and SOCKS5: different proxy boundaries

proxy-boundary

TLS is a key part of Trojan

A common Trojan design places the proxy protocol inside a TLS connection and uses a password for authentication. Check both whether the outer TLS connection works and whether the inner authentication details match. A reachable server address only means a connection attempt is possible; successful certificate verification does not prove the password is correct. Because these layers are separate, failures may appear as a handshake error, a disconnect after authentication, or an unreachable destination. Identify the stage where the failure occurs.

When configuring Trojan manually, the address, port, password, and Server Name may all be required. The server certificate must be valid, its name must match the expected identity, and the device clock must be correct. Preserve special characters in the password exactly; do not substitute a UUID. If the service also provides transport extensions, confirm that this endpoint requires them. Do not copy a WebSocket path from another server just because both use TLS.

HTTP proxy capabilities depend on the endpoint

An HTTP proxy handles requests using HTTP semantics; the CONNECT method is commonly used to create a tunnel to a destination. Distinguish between accessing an HTTPS website through a proxy and using TLS between the client and the proxy endpoint. The first describes the destination connection; the second describes the client-to-proxy hop. A standard HTTP proxy can forward an HTTPS CONNECT request, but that does not mean TLS automatically protects authentication details and negotiation with the proxy endpoint.

If an existing service requires a proxy endpoint protected by TLS, configure the matching client type and security options. Do not omit endpoint security just because the destination website uses HTTPS. The username and password are proxy credentials, not the website's login details. Standard HTTP CONNECT is mainly for TCP tunnels; do not assume it carries every UDP app. A working webpage only confirms that the capabilities needed for that particular request were available.

SOCKS5 is a negotiation protocol, not a transport encryption method

SOCKS5 defines how a client requests a connection from a proxy server. This can include username-and-password authentication, and the protocol also defines a UDP association mechanism. SOCKS5 itself does not encrypt the entire proxy connection. Any additional protection depends on the network environment or an outer tunnel. Configuring a SOCKS5 username and password only sets up the corresponding authentication parameters; it does not prove that authentication or payload data is encrypted in transit.

Whether UDP association works also depends on the server implementation, firewall, and address reachability. An endpoint may not enable UDP even though the SOCKS5 specification supports it. Check where domain names are resolved, too: the client may resolve a domain to an IP address before making the proxy request, or pass the domain to the remote side. The configuration and request path determine which occurs. Different resolution results can change the destination address and access behavior; they do not simply mean SOCKS5 is slower than another protocol.

TypeKey settingsWhat not to assume
TrojanPassword, TLS, Server NameA valid certificate means authentication succeeded
HTTPEndpoint address, authentication, endpoint securityAccessing HTTPS means the proxy hop is encrypted
SOCKS5Authentication, UDP support, resolution pathHaving a password means transport is encrypted

Choose a general-purpose proxy in a controlled environment

HTTP and SOCKS5 are commonly used for internal organizational proxies, test environments, or services you maintain. If an existing environment specifies an endpoint type, Shadowrocket's role is to configure that endpoint correctly—not replace it with a protocol that has a more complex name. Before sending sensitive authentication details over an untrusted network, confirm what protects the client-to-proxy hop. End-to-end TLS between the app and destination website remains important; it and proxy endpoint security protect different layers.

On iPhone, Trojan's TLS handshake and connection reuse, as well as the responsiveness of a general-purpose proxy, can affect speed and battery use, but they do not establish a fixed ranking. For real-time audio or video, or other UDP apps, first confirm that the entire path supports them. For an organization-mandated TCP proxy, keep the configuration simple, authentication clear, and certificate verification enabled. Avoid repeatedly changing settings to enable capabilities the endpoint does not provide.

WireGuard: keys, addresses, and tunnel scope

wireguard

Understand the configuration as a network-layer tunnel

WireGuard is designed to create an encrypted IP tunnel, not just define proxy requests for a single app. Its settings therefore include network-layer concepts such as interface addresses, peer public keys, Endpoint, and routing ranges. Its fields look different from those in Shadowsocks and Trojan by design. Do not enter a private key as an ordinary password or treat an interface address as a server address. First distinguish the local interface, remote peer, and destination address inside the tunnel to understand how traffic is routed.

WireGuard uses a fixed set of modern cryptographic algorithms, reducing the number of options the endpoints need to negotiate. You generally do not need to choose from a long list of encryption algorithms, but you must still manage keys correctly. Private Key is the local private key and should only be used for its corresponding device or configuration. The peer's Public Key identifies the other endpoint; it is not your own public key. If the configuration includes a Preshared Key, both sides must use the same value. It is additional shared key material, not a replacement for the public/private key pair.

Endpoint and interface address are not interchangeable

Endpoint describes how to reach the remote peer on the current network, usually with a hostname or IP address and port. The interface address is the local address inside the tunnel and often includes a prefix length. Putting an internal address in Endpoint may make the client try to reach a destination that is unavailable before the tunnel is established. If the configuration includes both IPv4 and IPv6, confirm that both the current network and server support them. The presence of one address family does not prove the other is available.

In WireGuard, Allowed IPs relates to routing and the address ranges permitted for a peer; it is not simply a website allowlist. When using WireGuard as an outbound type in Shadowrocket, distinguish its tunnel address ranges from the client's outer rule system. The outer rules first decide whether a connection uses this outbound; the outbound is then constrained by its own capabilities and settings. Do not assume that routing conclusions from a complete tunnel configuration apply to the entire client.

Keepalive and MTU solve different problems

Mobile networks often use NAT to maintain address mappings. Persistent Keepalive can periodically send data to maintain a mapping when needed, but more frequent is not always better. Unnecessary keepalives increase wireless activity and may affect standby battery life; intervals that are too long may fail to keep a mapping alive on some networks. Set this according to the server and network requirements, not a fixed value copied from elsewhere. If the first request fails only after a long idle period, check the mapping, keepalive, and handshake behavior.

MTU sets the maximum size of packets carried inside the tunnel. Encapsulation adds overhead, and mishandling path limits can make small requests work while large downloads stall. Lowering MTU may help, but not every slow connection is an MTU problem. First check whether the issue relates to packet size or a particular destination, then change one setting at a time. Setting MTU arbitrarily low increases packet count and overhead and may not improve throughput or battery life.

Check handshakes and routing when switching mobile networks

WireGuard has mechanisms for handling changes to the peer address, but recovery on iOS still depends on whether the current network can send UDP, the client tunnel state, and the server's response. If Wi-Fi works but cellular does not, compare UDP reachability and address families before generating new keys. Incorrect keys generally do not fail only on one local network. If the connection is established but internal resources remain unreachable, check the tunnel addresses, allowed ranges, server forwarding, and return path.

WireGuard is suited to users who already have a complete configuration and need to access a network they maintain or an existing tunnel service. Importing a configuration file does not mean every network-layer requirement is met, especially server forwarding and internal DNS. Evaluate battery use under realistic conditions: sustained transfer and long idle periods are different tests. Recording network type, keepalive settings, and reconnection behavior is more useful than comparing total client runtime alone.

Hysteria2: QUIC, congestion, and bandwidth expectations

hysteria2

Designed for data transport over UDP

Hysteria2 uses QUIC-related mechanisms to carry proxy traffic, with a design focused on high-latency and lossy network conditions. QUIC runs over UDP, but that does not mean application data lacks reliability or that all traffic is handled like ordinary real-time datagrams. This design can provide reliable transport, flow control, and encryption. Distinguish the use of UDP underneath from whether the destination app itself uses TCP or UDP; these are separate questions.

This design may improve performance on some routes, but the benefit depends on server capabilities, path conditions, and network policies. If UDP is blocked on the current network, a protocol designed for lossy links cannot establish a connection out of nothing. Even when UDP is reachable, path limits, device resources, and server bandwidth still impose a ceiling. Hysteria2 is not a switch that automatically boosts speed, and one fast download does not prove it will work better on every network.

Configure authentication and TLS fields completely

When entering settings manually, check the protocol type, server address, port, authentication details, and Server Name. Hysteria2 is not interchangeable with other protocols that have similar names; confirm that the client recognizes the imported type correctly. Certificate verification still identifies the peer. A connection timeout and a certificate error point to different troubleshooting steps: check network reachability and the port for a timeout; check the name, certificate chain, validity period, and device clock for a certificate error.

If your existing configuration includes obfuscation settings, they must match the server, including whether obfuscation is enabled and the corresponding authentication material. Obfuscation is not an encryption-strength slider, and not every endpoint needs it. Adding it arbitrarily can prevent the endpoints from understanding each other. When an existing subscription imports correctly, keep its complete settings. For a manual setup, list the fields first and verify the source of each non-default option. Do not treat example values in online articles as requirements for your server.

A bandwidth setting is not a speed test result

Some configurations include expected upload and download bandwidth values. Depending on the setup, these may affect congestion control or sending behavior. Entering a larger number will not increase your ISP bandwidth and may cause queueing, packet loss, or interfere with other traffic on the same network. Upload and download values are not interchangeable: many home connections have asymmetric capacity. Use the sustainable capacity of the current link, not the server's advertised port speed.

If the client interface or an existing configuration uses bandwidth values, first check the units: bits per second and bytes per second are different. Confirm whether the server accepts client bandwidth hints and how congestion control works for the current setup. A value that suits Wi-Fi may not suit cellular data. Without measurements, keep the values provided by your service or use supported defaults. Do not keep raising them just to increase a number in the interface.

On iOS, consider throughput and wireless activity together

On iPhone, high throughput means more data processing and wireless transmission. Finishing a download faster does not by itself mean a setup uses more or less power. Consider the total time and data needed for the same task, retry frequency, and signal strength. Continuous speed tests create a substantial load and are not suitable for estimating standby performance. Keep screen brightness, network, and test content consistent, and avoid large transfers from other apps during testing.

Different tasks need different performance measures. Sustained downloads depend on throughput, while real-time interactions depend more on latency variation and recovery from packet loss. A setup that saturates the link can make interactive requests on the same network wait in a queue. If webpages and calls degrade during downloads, check whether the upload expectation is too aggressive or the uplink is saturated instead of looking only at download speed. For a layered approach, see How to troubleshoot slow speeds: local network, protocol, and route.

Connect protocol settings with routing rules

fields-and-rules

Addresses, names, paths, and credentials each have a distinct purpose

Across protocols, the most common mistake is not missing an advanced option but putting a field from one layer into another. The server address locates the remote endpoint; the port selects a service. Server Name is used for TLS name negotiation and identity verification. Host may belong to a particular application-layer transport, while Path identifies a transport handler route. These fields may happen to contain the same text, but their meanings differ. When something fails, interpret each field by its purpose instead of copying values that merely look alike.

Credentials also come in different forms: passwords, UUIDs, Private Keys, and Public Keys are not interchangeable. A subscription URL may itself contain credentials and should be protected like a password; never include it in a public screenshot or troubleshooting note. If some imported fields are blank, the cause may be the source format, encoding, or client support. Do not immediately fill them with values from another server. Compare the entry with your original information and confirm that it imported completely; if necessary, obtain the correct representation of the same configuration.

Global Routing determines whether a protocol is used

In Global Routing, Config, Proxy, and Direct are client traffic-handling modes, not server protocols. Config follows the rules in the configuration file, Proxy handles connections in proxy mode, and Direct handles connections directly. A protocol entry can test successfully while an app still bypasses the proxy under Direct. Conversely, switching to Proxy may restore access simply by bypassing a faulty rule; it does not necessarily fix the protocol.

Interface optionHow it worksWhat to check
ConfigChoose a policy using the current configuration rulesSee which rule matches the destination
ProxyHandle connections in proxy modeCheck whether rules cause the difference
DirectHandle connections in direct modeCheck basic reachability on the local network

Rules are checked from top to bottom; the first match wins

The following Shadowrocket .conf snippet illustrates rule order. It contains no server details and is not a working proxy configuration on its own. DOMAIN-SUFFIX matches domain suffixes; IP-CIDR matches IP address ranges; GEOIP uses an IP geolocation database; FINAL catches connections that did not match an earlier rule. PROXY selects the proxy policy, and DIRECT selects direct access. example.com is a reserved example domain, not a real service to visit or configure.

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

If a broader rule comes first, a more specific rule below it may never match. For example, if an earlier rule already matches the destination's IP range, adding a domain rule later may not change the result. Domain rules also depend on the client having domain information when it processes the connection; a connection known only by IP may not match a domain suffix as expected. GEOIP does not identify a visitor or determine a website's language; it relies on the destination IP and the local database.

Verify DNS and UDP separately

DNS determines which addresses a domain resolves to, while the protocol determines how the selected connection reaches the remote endpoint. They can affect each other, but they are not the same thing. Changing a UUID usually will not fix a failed domain lookup; resolving to an unreachable address can prevent a webpage from loading even when the server handshake succeeds. Record whether resolution succeeds, which address family it returns, and which policy the connection matches. Avoid describing every failure as “the server is down.”

Check UDP support all the way through the app, client, outbound protocol, server, and network path; a single enabled option is not enough. On Demand controls when the client connection is established or maintained based on conditions. It does not change the server protocol or fix authentication settings. If an issue occurs only during automatic connection or a network switch, investigate the trigger conditions separately from the protocol handshake. For everyday use of the three modes, see the Global Routing guide; look up rule keywords in the glossary.

Choose based on what you have, then troubleshoot by layer

decision-and-check

List your constraints before ranking protocols

Start with a configuration you already know works. Note whether the current network allows UDP, which protocols and transport combinations the server provides, whether the target app needs UDP, and whether you care most about sustained downloads, interactive response, or recovery from idle. If there is only one supported server configuration, the right client choice is the matching type; there is no need to assemble fields around a protocol name. If you have multiple endpoints, first rule out combinations the current network cannot carry, then compare performance for your actual tasks.

If you need access to an internal network you maintain and already have complete addresses and keys, review WireGuard's routes and interface scope. For an existing lightweight encrypted proxy, first confirm the Shadowsocks method and UDP support. If the deployment depends on a specific transport, check the complete outer-layer setup for VMess or VLESS. For an existing TLS proxy endpoint, verify the Trojan certificate and password. For an organization-mandated general-purpose proxy, use HTTP or SOCKS5 according to its actual capabilities. Hysteria2 requires, among other things, a network path that can carry the necessary UDP traffic.

Break “can't connect” into observable stages

First, confirm that the local network works and choose a destination request you can verify. Next, check that the server domain resolves, the address family is suitable, and the port matches the transport. Then check the handshake: for TLS configurations, verify the name, certificate, and system clock; for configurations using keys or UUIDs, check that the credentials are complete. After authentication, check whether the destination request succeeds and whether the rules selected the intended outbound. Change one variable at a time and record what failed.

Connectivity Test can provide a clue, but a single result is not a substitute for testing the app you actually use. A test may cover only one kind of request, destination, or connection stage. A response in the list does not prove that target videos, UDP apps, or internal domains work; a failed test does not prove that every service is unreachable. Use a destination and traffic type that match the issue as closely as possible, and note whether the failure is immediate, times out, or occurs after data transfer begins.

SymptomCheck firstDon't change yet
Handshake never startsAddress, port, TCP / UDP reachabilityRule details, throughput expectations
Certificate verification failsServer Name, system time, certificate chainUUID, rule order
Webpages work, but real-time apps failEnd-to-end UDP path, destination rulesRebuilding every server entry
Only large downloads stallSigns of path, packet loss, or MTU issuesSwapping credentials at random
Only fails in ConfigMatched rule, policy, DNSProtocol type

Track speed and battery use separately

For speed tests, keep the destination, local network, and server endpoint consistent. Measure time to first response and sustained transfer separately, and do not mistake a cache hit for a protocol improvement. For battery tests, keep the task, screen use, and signal conditions consistent, and note whether the connection repeatedly reconnects while idle. System battery statistics reflect combined activity over time; they cannot isolate the cost of each protocol handshake. Do not rank protocols based on a percentage change over a few minutes, and avoid running repeated high-load tests to investigate standby drain.

If an issue appears only after switching networks, record Wi-Fi and cellular behavior separately, including address family, UDP reachability, and recovery after idle. If only one destination is affected, check its DNS, routing, and service status before changing the entire protocol. If several configurations fail on the same local network but work on another, investigate the local route before assuming multiple credentials failed at once. These checks help avoid unsupported configuration changes.

Keep the original configuration and reproducible steps

If behavior changes after a subscription update, first compare whether the entry was replaced, transport fields changed, or manual edits were overwritten. Do not combine credentials from an old configuration with a new endpoint. For a minimal reproduction, record the device network, protocol combination, failure stage, destination type, and results from checks you have already made. Hide the full subscription URL and credentials. For an empty subscription import, see Subscription import troubleshooting; for other common issues, see Frequently asked questions.

After testing, restore your usual Global Routing mode and undo any temporary relaxed settings. A usable, complete configuration you can explain is a better long-term choice than a one-off speed result you cannot reproduce. To review import and connection steps, return to the user guide. To get the client, visit the download page for all platforms. For iPhone / iPad, check the App Store page for system requirements; protocol names alone do not determine device compatibility.