Skip to main content

Getting Started

Woops! You fell for the marketing hype. Let's try and get through this together.


The easiest way to install Benthos is with this handy script:

curl -Lsf | bash

Or you can grab an archive containing Benthos from the releases page.


If you have docker installed you can pull the latest official Benthos image with:

docker pull jeffail/benthos
docker run --rm -v /path/to/your/config.yaml:/benthos.yaml jeffail/benthos


On macOS, Benthos can be installed via Homebrew:

brew install benthos


If you use the asdf tool version manager you can install Benthos with:

asdf plugin add benthos
asdf install benthos latest
asdf global benthos latest


For information about serverless deployments of Benthos check out the serverless section here.


A Benthos stream pipeline is configured with a single config file, you can generate a fresh one with:

benthos create > config.yaml

The main sections that make up a config are input, pipeline and output. When you generate a fresh config it'll simply pipe stdin to stdout like this:

stdin: {}

processors: []

stdout: {}

Eventually we'll want to configure a more useful input and output, but for now this is useful for quickly testing processors. You can execute this config with:

benthos -c ./config.yaml

Anything you write to stdin will get written unchanged to stdout, cool! Resist the temptation to play with this for hours, there's more stuff to try out.

Next, let's add some processing steps in order to mutate messages. The most powerful one is the mapping processor which allows us to perform mappings, let's add a mapping to uppercase our messages:

stdin: {}

- mapping: root = content().uppercase()

stdout: {}

Now your messages should come out in all caps, how whacky! IT'S LIKE BENTHOS IS SHOUTING BACK AT YOU!

You can add as many processing steps as you like, and since processors are what make Benthos powerful they are worth experimenting with. Let's create a more advanced pipeline that works with JSON documents:

stdin: {}

- sleep:
duration: 500ms
- mapping: |
root.doc = this
root.first_name = this.names.index(0).uppercase()
root.last_name = this.names.index(-1).hash("sha256").encode("base64")

stdout: {}

First, we sleep for 500 milliseconds just to keep the suspense going. Next, we restructure our input JSON document by nesting it within a field doc, we map the upper-cased first element of names to a new field first_name. Finally, we map the hashed and base64 encoded value of the last element of names to a new field last_name.

Try running that config with some sample documents:

echo '{"id":"1","names":["celine","dion"]}
{"id":"2","names":["chad","robert","kroeger"]}' | benthos -c ./config.yaml

You should see (amongst some logs):


How exciting! I don't know about you but I'm going to need to lie down for a while. Now that you are a Benthos expert might I suggest you peruse these sections to see if anything tickles your fancy?