Skip to content
Cloud Native Tunnel for APIs
Go Shell Makefile Dockerfile
Branch: master
Clone or download

Inlets is a Cloud Native Tunnel written in Go

Expose your local endpoints to the Internet or to another network, traversing firewalls and NAT.

Build Status License: MIT Go Report Card GitHub All Releases

Follow @inletsdev on Twitter

English | 中文文档


inlets combines a reverse proxy and websocket tunnels to expose your internal and development endpoints to the public Internet via an exit-node. An exit-node may be a 5-10 USD VPS or any other computer with an IPv4 IP address.

Why do we need this project? Similar tools such as ngrok or Argo Tunnel from Cloudflare are closed-source, have limits built-in, can work out expensive, and have limited support for arm/arm64. Ngrok is also often banned by corporate firewall policies meaning it can be unusable. Other open-source tunnel tools are designed to only set up a single static tunnel. inlets aims to dynamically bind and discover your local services to DNS entries with automated TLS certificates to a public IP address over a websocket tunnel.

When combined with SSL - inlets can be used with any corporate HTTP proxy which supports CONNECT.

Conceptual diagram

Conceptual diagram for inlets

Who is behind this project?

inlets is brought to you by Alex Ellis. Alex is a CNCF Ambassador and the founder of OpenFaaS.

OpenFaaS® makes it easy for developers to deploy event-driven functions and microservices to Kubernetes without repetitive, boiler-plate coding. Package your code or an existing binary in a Docker image to get a highly scalable endpoint with auto-scaling and metrics. The project has around 21k GitHub stars, over 270 contributors and a growing number of end-users in production.

SWAG and merchandise

Head over to the new OpenFaaS Ltd SWAG store to get your very own branded inlets hoodie, t-shirt and mug.

Insiders subscription

Buy an Insiders Subscription to get regular updates on the inlets, openfaas, k3sup, arkade along with hints and tips on Go, Docker and Kubernetes.

License & terms


Developers wishing to use inlets within a corporate network are advised to seek approval from their administrators or management before using the tool. By downloading, using, or distributing inlets, you agree to the LICENSE terms & conditions. No warranty or liability is provided.

Backlog & goals


  • automatically create endpoints on exit-node based upon client definitions
    • multiplex sites on same port and websocket through the use of DNS / host entries
  • link encryption using SSL over websockets (wss://)
  • authentication for server and client
  • automatic reconnect
  • native multi-arch with ARMHF/ARM64 support
  • Dockerfile and Kubernetes YAML files
  • discover and implement Service type of LoadBalancer for Kubernetes - inlets-operator
  • tunnelling websocket traffic in addition to HTTP(s)
  • get a logo for the project
  • TLS certificates when used with a reverse proxy like Nginx, Caddy or Traefik

Inlets PRO

The following features / use-cases are covered by

  • Tunnel L4 TCP traffic in addition to HTTP/s at L7
  • Automated TLS - including via inletsctl/inlets-operator
  • Commercial services & support


Unlike HTTP 1.1 which follows a synchronous request/response model websockets use an asynchronous pub/sub model for sending and receiving messages. This presents a challenge for tunneling a synchronous protocol over an asynchronous bus.

inlets 2.0 introduces performance enhancements and leverages parts of the Kubernetes and Rancher API. It uses the same tunnelling packages that enable node-to-node communication in Rancher's k3s project. It is suitable for development and may be useful in production. Before deploying inlets into production, it is advised that you do adequate testing.

Feel free to open issues if you have comments, suggestions or contributions.

  • The tunnel link is secured via --token flag using a shared secret
  • The default configuration uses websockets without SSL ws://, but to enable encryption you could enable SSL wss://
  • A timeout for requests can be configured via args on the server
  • The upstream URL has to be configured on both server and client until a discovery or service advertisement mechanism is added The client can advertise upstream URLs, which it can serve
  • The tunnel transport is wrapped by default which strips CORS headers from responses, but you can disable it with the --disable-transport-wrapping flag on the server

inlets projects

Inlets is a Cloud Native Tunnel and is listed on the Cloud Native Landscape under Service Proxies.

  • inlets - Cloud Native Tunnel for L7 / HTTP traffic written in Go
  • inlets-pro - Cloud Native Tunnel for L4 TCP
  • inlets-operator - Public IPs for your private Kubernetes Services and CRD
  • inletsctl - Automate the cloud for fast HTTP (L7) and TCP (L4) tunnels

Get started

You can install the CLI with a curl utility script, brew or by downloading the binary from the releases page. Once installed you'll get the inlets command.

Install the CLI

Note: inlets is made available free-of-charge, but you can support its ongoing development through GitHub Sponsors 💪

Utility script with curl:

# Install to local directory
curl -sLS | sh

# Install to /usr/local/bin/
curl -sLS | sudo sh

Via brew:

brew install inlets

Note: the brew distribution is maintained by the brew team, so it may lag a little behind the GitHub release.

Binaries are made available on the releases page for Linux (x86_64, armhf & arm64), Windows (experimental), and for Darwin (MacOS). You will also find SHA checksums available if you want to verify your download.

Windows users are encouraged to use git bash to install the inlets binary.

Video demo

Using inlets I was able to set up a public endpoint (with a custom domain name) for my JavaScript & Webpack Create React App.

Quickstart tutorial

You can run inlets between any two computers with connectivity, these could be containers, VMs, bare metal or even "loop-back" on your own laptop.

See how to provision an "exit-node" with a public IPv4 address using a VPS.

  • On the exit-node (or server)

Start the tunnel server on a machine with a publicly-accessible IPv4 IP address such as a VPS.

Example with a token for client authentication:

export token=$(head -c 16 /dev/urandom | shasum | cut -d" " -f1)
inlets server --port=8090 --token=$token

Note: You can pass the --token argument followed by a token value to both the server and client to prevent unauthorized connections to the tunnel.

inlets server --port=8090

You can also run unprotected, but this is not recommended.

Note down your public IPv4 IP address.

  • Head over to your machine where you are running a sample service, or something you want to expose.

You can use my hash-browns service for instance which generates hashes.

Install hash-browns or run your own HTTP server

export GO111MODULE=off
export GOPATH=$HOME/go/

go get -u
port=3000 $GOPATH/bin/hash-browns

If you don't have Go installed, then you could run Python's built-in HTTP server:

mkdir -p /tmp/inlets-test/
cd /tmp/inlets-test/
touch hello-world
python -m SimpleHTTPServer 3000
  • On the same machine, start the inlets client

Start the tunnel client:

export REMOTE=""    # for testing inlets on your laptop, replace with the public IPv4
export TOKEN="CLIENT-TOKEN-HERE"  # the client token is found on your VPS or on start-up of "inlets server"
inlets client \
 --remote=$REMOTE \
 --upstream= \
  • Replace the --remote with the address where your exit-node is running inlets server.
  • Replace the --token with the value from your server

We now have three processes:

  • example service running (hash-browns) or Python's webserver
  • an exit-node running the tunnel server (inlets server)
  • a client running the tunnel client (inlets client)

So send a request to the inlets server - use its domain name or IP address:

Assuming points to in /etc/hosts or your DNS server.

curl -d "hash this" -H "Host:"
# or
curl -d "hash this"
# or
curl -d "hash this"

You will see the traffic pass between the exit node / server and your development machine. You'll see the hash message appear in the logs as below:

~/go/src/$ port=3000 go run server.go
2018/12/23 20:15:00 Listening on port: 3000
"hash this"

Now check the metrics endpoint which is built-into the hash-browns example service:

curl $REMOTE/metrics | grep hash

You can also use multiple domain names and tie them back to different internal services.

Here we start the Python server on two different ports, serving content from two different locations and then map it to two different Host headers, or domain names:

mkdir -p /tmp/store1
cd /tmp/store1/
touch hello-store-1
python -m SimpleHTTPServer 8001 &

mkdir -p /tmp/store2
cd /tmp/store2/
touch hello-store-2
python -m SimpleHTTPServer 8002 &
export REMOTE=""    # for testing inlets on your laptop, replace with the public IPv4
export TOKEN="CLIENT-TOKEN-HERE"  # the client token is found on your VPS or on start-up of "inlets server"
inlets client \
 --remote=$REMOTE \
 --upstream="," \

You can now create two DNS entries or /etc/hosts file entries for and, then connect through your browser.

Docs & Featured tutorials

inlets and inlets PRO now has a dedicated documentation site:

Official docs:

Other inlets OSS documentation & tutorials:

Advanced usage


Automate inlets exit server creation and tunnel clients with the inlets-operator

Or see the examples in the docs/ file for how to run either part of the tunnel manually.


Docker images are published as multi-arch for x86_64, arm64 and armhf

  • inlets/inlets:2.6.3

Bind to a different adapter, or to localhost

By default the inlets server will bind to all adapters and addresses on your machine.

At times, you may wish to change this, so that you can "hide" the HTTP websocket behind a reverse proxy, adding TLS termination and link-level encryption without exposing the plain HTTP port to the network or Internet.


  • --bind-addr
  • --bind-addr

Bind a different port for the control-plane

You can bind two separate TCP ports for the user-facing port and the tunnel.

  • --port - the port for users to connect to and for serving data, i.e. the Data Plane
  • --control-port - the port for the websocket to connect to i.e. the Control Plane

Strict forwarding policy

By default, the server code can access any host. The client specifies a number of upstream hosts via --upstream. If you want these to be the only hosts that the server can connect to, then enable strict forwarding.

  • --strict-forwarding

This is off by default, however when set to true, only hosts in --upstream can be accessed by the server. It could prevent a bad actor from accessing other hosts on your network.

Tunnelling multiple services

You can expose multiple hosts through the --upstream flag using a comma-delimited list.

inlets client --remote ws://$IP:8080 \
  --upstream ","

You can also forward everything to a single host such as:

inlets client --remote ws://$IP:8080 \
  --upstream "http://nginx.svc.default"


What are people saying about inlets?

You can share about inlets using @inletsdev, #inletsdev, and

inlets has trended on the front page of Hacker News twice.

Official tutorials:

Community tutorials:


Note: add a PR to send your story or use-case, I'd love to hear from you.

See for what companies are doing with inlets today.



For development you will need Golang 1.13 or newer

You can get the code like this:

go get -u
cd $GOPATH/src/

Alternatively, you can get everything setup right in the browser with a single click using Gitpod:

Open in Gitpod

Contributions are welcome. All commits must be signed-off with git commit -s to accept the Developer Certificate of Origin.

Other Kubernetes port-forwarding tooling

  • kubectl port-forward - built into the Kubernetes CLI, forwards a single port to the local computer.
  • kubefwd - Kubernetes utility to port-forward multiple services to your local computer.
  • kurun - Run main.go in Kubernetes with one command, also port-forward your app into Kubernetes.
You can’t perform that action at this time.