Skip to main content

Build a Custom API Client


This tutorial teaches you how to create a custom client for the Dagger GraphQL API in a programming language of your choice. You will learn how to:

  • Create a custom client for the Dagger GraphQL API
  • Connect to the Dagger GraphQL API and run your custom client with the Dagger CLI


You can create a Dagger GraphQL API client in any programming language. This tutorial demonstrates the process in the following languages:

This tutorial assumes that:

Step 1: Select and install your GraphQL client library

The first step is to identify available GraphQL clients for your chosen programming language and select one that fits your requirements. GraphQL has a large and growing list of client implementations in over 20 languages.

Create a new directory for the project and install the client as follows:

mkdir my-project
cd my-project
cargo init
cargo add gql_client@1.0.7
cargo add serde_json@1.0.89
cargo add tokio@1.22.0 -F full
cargo add base64@0.13.1

Step 2: Create an API client

Once the client library is installed, create an API client as described below.

Add the following code to src/

use base64::encode;
use gql_client::Client;
use serde_json::Value;
use std::collections::HashMap;
use std::env;

async fn main() -> Result<(), Box<dyn std::error::Error>> {
let port = env::var("DAGGER_SESSION_PORT").expect("$DAGGER_SESSION_PORT doesn't exist");
let token = env::var("DAGGER_SESSION_TOKEN").expect("$DAGGER_SESSION_TOKEN doesn't exist");
let query = r#"
query {
container {
from (address: "alpine:latest") {
withExec(args:["uname", "-nrio"]) {

let mut headers = HashMap::new();
format!("Basic {}", encode(format!("{}:", token))),
let client = Client::new_with_headers(format!("{}/query", port), headers);
let data = client.query_unwrap::<Value>(query).await.unwrap();



This code listing initializes the client library and defines the Dagger pipeline to be executed as a GraphQL query. This query performs the following operations:

  • It requests the from field of Dagger's Container object type, passing it the address of a container image. To resolve this, Dagger will initialize a container using the specified image and return a Container object representing the alpine:latest container image.
  • Next, it requests the withExec field of the Container object from the previous step, passing the uname -a command to the field as an array of arguments. To resolve this, Dagger will return a Container object containing the execution plan.
  • Finally, it requests the stdout field of the Container object returned in the previous step. To resolve this, Dagger will execute the command and return a String containing the results.
  • The result of the query is a JSON object, which is processed and printed to the output device.

The API endpoint and the HTTP authentication token for the GraphQL client are not statically defined, they must be retrieved at run-time from the special DAGGER_SESSION_PORT and DAGGER_SESSION_TOKEN environment variables. This is explained in more detail in the next section.

Step 3: Run the API client

To run the pipeline, the API client needs to communicate with the Dagger Engine, which is responsible for accepting the query, executing it and returning the result. The dagger run command takes care of initializing a new local instance (or reusing a running instance) of the Dagger Engine on the host system and executing a specified command against it.

The Dagger Engine creates a unique local API endpoint for GraphQL queries for every Dagger session. This API endpoint is served by the local host at the port specified by the DAGGER_SESSION_PORT environment variable, and can be directly read from the environment in your client code. For example, if DAGGER_SESSION_PORT is set to 12345, the API endpoint can be reached at$DAGGER_SESSION_PORT/query


The Dagger Engine protects the exposed API with an HTTP Basic authentication token which can be retrieved from the DAGGER_SESSION_TOKEN variable. Treat the DAGGER_SESSION_TOKEN value as you would any other sensitive credential. Store it securely and avoid passing it to, or over, insecure applications and networks.

Run the API client using the Dagger CLI as follows:

dagger run cargo run

This command:

  • initializes a new Dagger Engine session
  • sets the DAGGER_SESSION_PORT and DAGGER_SESSION_TOKEN environment variables.
  • executes the cargo run command in that session

The specified command, in turn, invokes the custom API client, connects to the API endpoint specified in the DAGGER_SESSION_PORT environment variable, sets an HTTP Basic authentication token with DAGGER_SESSION_TOKEN and executes the GraphQL query. Here is an example of the output:

buildkitsandbox 5.15.0-53-generic unknown Linux


This tutorial explained how to write a custom client for the Dagger GraphQL API. It provided working examples of how to program and run a Dagger pipeline using this client in two different programming languages. A similar approach can be followed in any other programming language with a GraphQL client implementation.

Use the API Reference and the CLI Reference to learn more about the Dagger GraphQL API and the Dagger CLI respectively.