inlets

README

Inlets Go

A high-availability Go implementation of the inlets tunnel system, including both client and server components. It establishes long-lived WebSocket connections to securely expose local HTTP/TCP services to the public internet.

Architecture

flowchart LR
    subgraph Local["Local Environment"]
        CLI["inlets client CLI"]
        Tunnels["HTTP/TCP Upstream"]
    end

    subgraph Edge["Tunnel Server"]
        WS["WebSocket Server"]
        Router["Request Router"]
        TCPHub["TCP Relay"]
    end

    Internet((Internet))

    Tunnels <--> CLI
    CLI -- WebSocket + Heartbeat --> WS
    WS --> Router
    Router --> Internet
    WS --> TCPHub
    TCPHub --> Internet

Data Flow

1. After the CLI starts, it establishes a connection with the cloud WebSocket server and completes signature authentication (token/credentials/public).
2. After a successful connection, two data channels are created:
   • HTTP: The server sends requests through WebSocket, and the client forwards them locally and writes back responses.
   • TCP: The server listens on a public TCP port, and after a user connects, it calls back to the client to establish the actual data stream.
3. The client maintains heartbeat (ping/pong + server @@CONFIG dynamic delivery) and automatic reconnection to ensure tunnel stability.

Module Structure

internal/client/
├── client.go       // Connection management, reconnection, message distribution
├── handlers.go     // HTTP/TCP data plane logic
├── heartbeat.go    // Heartbeat and authentication timeout
├── types.go        // Configuration and DTOs
└── utils.go        // HMAC, address utilities

internal/server/
├── server.go       // Server main logic
├── protocol/       // Protocol handling (new/legacy protocol adapter)
├── channels/       // Data channel management
├── tunnel/         // Tunnel handling (HTTP/TCP)
└── ...

Features

• HTTP & TCP dual tunneling
• Three authentication methods: Token / Credentials / Public
• Automatic reconnection, heartbeat keepalive, drift timeout prevention
• End-to-end TCP HMAC verification
• IPv4/IPv6 compatible net.JoinHostPort address assembly
• Protocol version negotiation support (2.0.0+ supports new protocol, auto-downgrade for legacy compatibility)
• Server supports hot-reload configuration files
• Server supports bandwidth limiting
• Server supports multiple notification methods (DingTalk, Feishu, WeCom, Slack)

Stability Update (2026-03)

To address cases where some HTTPS requests stayed pending under higher concurrency, the following fixes are now in place:

• Callback race fix: HTTP tunnel now registers response callbacks before sending requests, preventing lost fast responses.
• Request timeout fallback: Server-side tunnel requests now have a timeout guard and return 504 Gateway Timeout instead of hanging indefinitely.
• Atomic callback consumption: Added Take(tcpId, requestId) semantics to fetch-and-remove callbacks in one step.
• Client HTTP response parsing fix: Client no longer waits for EOF to finish a response; it now parses responses via HTTP protocol semantics, which works with keep-alive upstreams.

New tests added for this change:

• internal/server/container/callback_test.go
• internal/server/channels/monitor/auth_test.go
• internal/server/tunnel/http_test.go

Building

# Build the complete program (includes client, server, forward commands)
go build -o inlets ./cmd/inlets

# Or specify the full path
go build -o inlets cmd/inlets/inlets.go

Command Line Usage

Client

HTTP Tunnel

# Public HTTP tunnel (public mode; control plane unauthenticated: session may be time-limited by server)
inlets client http 127.0.0.1:9000

# Specify subdomain + token
inlets client --sub-domain myapp -t token http 127.0.0.1:9000

Public monitor session: If you do not pass --token or --credentials, the server classifies the client as a “public” (temporary) control-plane user and may close the WebSocket after a default period (e.g. 10 minutes, configurable in server YAML as publicHTTPNoAuth). This is independent of HTTP tunnel or public-URL (edge) auth. See docs/features/PUBLIC_MONITOR_SESSION.md.

TCP Tunnel

# Using token
inlets client -p 20100 -t token tcp 127.0.0.1:22

# Using credentials
inlets client --credentials clientId:clientSecret -p 20100 tcp 127.0.0.1:22

Version Information

# Print version information
inlets --version
# or
inlets -V

Protocol Version

# Use latest protocol version (default v2, supports capability negotiation)
inlets client --server https://tunnel.example.com http 127.0.0.1:9000

# Use legacy protocol version (legacy mode, v1)
inlets client --legacy --remote tunnel.example.com:443 http 127.0.0.1:9000

Protocol Version Notes:

• Default (v2 / 2.0.0): Use --server (HTTP/HTTPS URL). The client sends capabilities for negotiation and expects the v2 monitor endpoint.
• Legacy (v1 / 1.2.0): Use --legacy with --remote and --remote-tcp-port. This mode is compatible with older server versions.

Client Common Parameters

Parameter	Description	Default
type	Tunnel type http / tcp	Required
upstream	Local upstream, port or host:port	Required
--sub-domain	HTTP custom subdomain (http subcommand only)
-p, --port	TCP tunnel port
-t, --token	Token authentication
--credentials	clientId:clientSecret
--server	v2 server URL (http:// or https://, optional path)	https://inlets.zcorky.com
-r, --remote	Legacy server address (host:port)	inlets.zcorky.com:443
--remote-tcp-port	Legacy server TCP callback port	8443
--healthcheck-interval	Authentication timeout / health check interval (ms)	30000
--legacy	Use legacy protocol version (v1)	false (default v2)
--report-url	Error report webhook

Client Environment Variables:

All parameters can be configured via environment variables. Environment variables have lower priority than command-line arguments:

• TUNNEL_PORT: TCP tunnel port
• SUB_DOMAIN: HTTP custom subdomain
• TOKEN: Token authentication
• CREDENTIALS: Authentication credentials (clientId:clientSecret)
• SERVER: v2 server URL (default: https://inlets.zcorky.com)
• REMOTE: Legacy server address (default: inlets.zcorky.com:443)
• REMOTE_TCP_PORT: Legacy server TCP callback port (default: 8443)
• HEALTHCHECK_INTERVAL: Health check interval (ms, default: 30000)
• REPORT_URL: Error report webhook
• LEGACY: Use legacy protocol version (set to true, 1, or yes to enable)

Server

# Start server (domain required)
inlets server -d example.com -t your-token

# Use config file
inlets server -c /path/to/config.yml

# Specify ports
inlets server -d example.com -p 8080 --tcp-port 8443

# Disable HTTPS (enabled by default)
inlets server -d example.com -t your-token --secure=false

Server Common Parameters

Parameter	Description	Default
-d, --domain	Server domain (required)
-p, --port	WebSocket service port	8080
--tcp-port	TCP service port	8443
-s, --secure	Enable HTTPS (only for URL)	true
-t, --token	Authentication token
-c, --config	Config file path	$HOME/.config/inlets.yml
--notification-provider	Notification provider (dingtalk, feishu, wecom, slack)
--notification-url	Notification webhook URL

Server Environment Variables:

• DOMAIN: Server domain
• SERVER_PORT: WebSocket service port (default: 8080)
• SERVER_TCP_PORT: TCP service port (default: 8443)
• SECURE: Enable HTTPS (default: true)
• TOKEN: Authentication token
• NOTIFICATION_PROVIDER: Notification provider
• NOTIFICATION_URL: Notification webhook URL

Server Configuration File

The server supports YAML configuration files. The default path is $HOME/.config/inlets.yml, and it supports hot-reload.

Configuration file example:

domain: example.com
port: 8080
tcpPort: 8443
secure: true
token: your-token

clients:
  - clientId: client1
    clientSecret: secret1
    config:
      version: "2.0.0"
    bandwidthLimit:
      upload: 1024000    # 1MB/s
      download: 1024000  # 1MB/s

bandwidthLimits:
  global:
    upload: 512000      # 512KB/s
    download: 512000    # 512KB/s
  clients:
    client1:
      upload: 1024000
      download: 1024000

notification:
  provider: dingtalk
  url: https://oapi.dingtalk.com/robot/send?access_token=xxx

Forward

# TCP port forwarding
inlets forward -s 0.0.0.0:8080 -t 127.0.0.1:3000

Examples

Client Examples

# Development environment connecting to local server
inlets client --server http://127.0.0.1:8080 http 127.0.0.1:9000

# v2 server URL with base path prefix
inlets client --server https://tunnel.example.com/base http --sub-domain myapp 127.0.0.1:9000

# Production SSH tunnel
inlets client --credentials prod:secret -p 20100 tcp 127.0.0.1:22

# HTTP tunnel with custom subdomain
inlets client --sub-domain myapp -t token http 127.0.0.1:9000

# Legacy mode example
inlets client --legacy --remote tunnel.example.com:443 --remote-tcp-port 8443 http 127.0.0.1:9000

Server Examples

# Basic startup
inlets server -d tunnel.example.com -t your-secret-token

# Using config file
inlets server -c /etc/inlets/config.yml

# Custom ports
inlets server -d tunnel.example.com -t token -p 9000 --tcp-port 9443

Documentation

Index

• Variables

Variables

var Version = "1.28.6"

Version is the version of lighthouse