Skip to main content

Create a Multi-Build CI Pipeline


The Dagger Python SDK makes it easy to build an application for multiple OS and architecture combinations. This guide provides a working example of a Python CI tool that performs this task.


This guide assumes that:

  • You have a Python development environment with Python 3.10 or later. If not, install Python.
  • You are familiar with the basics of the Python SDK and have it installed. If not, read the Python SDK guide and the Python SDK installation instructions.
  • You have Docker installed and running on the host system. If not, install Docker.
  • You have an application that you wish to build. This guide assumes a Go application, but you can use an application of your choice.

Dagger pipelines are executed as standard OCI containers. This portability enables you to do very powerful things. For example, if you're a Python developer, you can use the Python SDK to create a pipeline (written in Python) that builds an application written in a different language (Go) without needing to learn that language.


Assume that the Go application to be built is stored in the current directory on the host. The following code listing demonstrates how to build this Go application for multiple OS and architecture combinations using the Python SDK.

Create a multi-build pipeline for a Go application.

import itertools
import sys

import anyio
import graphql

import dagger

async def build():
print("Building with Dagger")

# define build matrix
oses = ["linux", "darwin"]
arches = ["amd64", "arm64"]

# initialize dagger client
async with dagger.Connection(dagger.Config(log_output=sys.stderr)) as client:

# get reference to the local project
src =".")

# create empty directory to put build outputs
outputs =

golang = (
# get `golang` image
# mount source code into `golang` image
.with_mounted_directory("/src", src)

for goos, goarch in itertools.product(oses, arches):
# create a directory for each OS and architecture
path = f"build/{goos}/{goarch}/"

build = (
# set GOARCH and GOOS in the build environment
.with_env_variable("GOOS", goos)
.with_env_variable("GOARCH", goarch)
.with_exec(["go", "build", "-o", path])

# add build to outputs
outputs = outputs.with_directory(path,

# write build artifacts to host
await outputs.export(".")

if __name__ == "__main__":
except graphql.GraphQLError as e:
print(e.message, file=sys.stderr)

The build() function does the following:

  • It defines the build matrix, consisting of two OSs (darwin and linux) and two architectures (amd64 and arm64).
  • It creates a Dagger client with dagger.Connection().
  • It uses the client's host().directory(".") method to obtain a reference to the current directory on the host. This reference is stored in the src variable.
  • It uses the client's container().from_() method to initialize a new container from a base image. This base image contains all the tooling needed to build the application - in this case, the golang:latest image. This from_() method returns a new Container class with the results.
  • It uses the Container.with_mounted_directory() method to mount the host directory into the container at the /src mount point.
  • It uses the Container.with_workdir() method to set the working directory in the container.
  • It iterates over the build matrix, creating a directory in the container for each OS/architecture combination and building the Go application for each such combination. The Go build process is instructed via the GOOS and GOARCH build variables, which are reset for each case via the Container.with_env_variable() method.
  • It obtains a reference to the build output directory in the container with the with_directory() method, and then uses the Directory.export() method to write the build directory from the container to the host.