Custom flows overview

Run custom commands during deployment with env0.yml custom flows, supporting bash, Python, and Ansible hooks before or after Terraform init and apply.

Custom Flows are a way to run your own commands during the deployment of your environment. These will run outside the IaC of your choice and allow you to perform supplemental actions. For example, injecting custom variables into your IaC code before you apply it or calling external APIs with the outputs of your deployment.

Custom Flows allow you to run whatever you want (bash, Python, Gcloud, Ansible, CloudFormation, etc.) at any point in the deployment process– before or after Terraform init/plan/apply, destroy/error and even tasks.

To attach a custom flow to a single template:
Create a file named env0.yml under the template folder. env zero will use this file if it’s found in your git repository.

For sharing a custom flow between templates:
Create a file named env0.yml under the root folder. env zero will use this file if it is defined, and an env0.yml file is not found in the template folder.

For sharing a custom flow for all environments under a particular project:
Create a custom flow file in any repository and configure it as a project policy.

We support the suffixes env0.yml and env0.yaml in the same way.

Working Directory for env0.yml

env zero will run Custom Flow commands with the working directory being the one containing the YAML file. Based on that, you can reference your commands to other scripts/commands by using ./script.sh.

Project level Custom Flows

Only the env0.yaml is “cloned” for project level Custom Flows, this means that scripts and other files referenced by your Custom Flow will either need to be cloned into the working directory, or you must specify the repo path location of your scripts and files.

Below is an example of a custom flow that creates a more “dynamic” Terraform file, based on an environment variable value, and uses Python and Jinja before the Terraform Init (the full example of this template git repo (including the env0.yml file) is here):

version: 2

deploy:
  steps:
    terraformInit:
      before:
        - name: Hydrate Template
          run: |
            pip install --user jinjanator
            echo "Generating \"stage.auto.tfvars\" for \"$STAGE\""
            jinjanate stage.auto.tfvars.j2 > stage.auto.tfvars
            echo -e "Generated \"stage.auto.tfvars\":\n$(cat stage.auto.tfvars)"
            echo "Generating \"elb.tf\" for \"$STAGE\""
            jinjanate elb.tf.j2 > elb.tf
            echo -e "Generated \"elb.tf\":\n$(cat elb.tf)"

You can include the following possible hooks in an _ env0.yml _ file.

OpenTofu backward-compatibility

Terraform’s Custom Flow format is interchangeable with OpenTofu’s.

Task hooks

Task hook uses the stored working directory source code to run, it will not fetch an updated version from you VCS provider. If you need updated files on your working directory to run the task hooks you will need to first run a deployment to update the files.

Hook Stages

Terraform

version:
  type: number
shell: sh (default) or bash

deploy:
  onCompletion:
    - type: string
  onSuccess:
    - type: string
  onFailure:
    - type: string

  steps:
    setupVariables:
      after:
        - type: string

    terraformInit:
      before:
        - type: string
      after:
        - type: string

    terraformPlan:
      before:
        - type: string
      after:
        - type: string

    terraformApply:
      before:
        - type: string
      after:
        - type: string

    storeState:
      before:
        - type: string
      after:
        - type: string

    terraformOutput:
      before:
        - type: string
      after:
        - type: string

destroy:
  onCompletion:
    - type: string
  onSuccess:
    - type: string
  onFailure:
    - type: string

  steps:
    setupVariables:
      after:
        - type: string

    terraformInit:
      before:
        - type: string
      after:
        - type: string

    terraformPlan:
      before:
        - type: string
      after:
        - type: string

    terraformDestroy:
      before:
        - type: string
      after:
        - type: string

Specifying the Shell

Currently, env zero’s Custom Flow features allow you to select the shell: sh (default) or bash
To use bash, add the following snippet into your_ env0.yaml_:

shell: bash

Handling Errors

If any command in your custom flow returns a non-zero exit code, the env zero deployment will stop on a ‘Failed’ status. You can use this to perform validation. For instance, asserting that a certain environment variable was supplied.

In such a case, env zero will display any output from stderr as the deployment error. If no output is found in stderr, it will be shown as an ‘Unknown Error’.

Error Message Handling

Use the 1>&2 notation to redirect stdout messages to stderr. This allows env zero to display the message in its UI (see I/O Redirection)

Here is an example:

env0.yml

version: 2
deploy:
  steps:
    setupVariables:
      after:
        - if [ -z "$MY_VAR" ]; then echo "MY_VAR must be supplied" 1>&2 && exit 1; fi

Exposed env zero System Environment Variables

env zero exposes the following environment variables for you to use:

Variable Name	Value Description
ENV0_ENVIRONMENT_ID	The deployed Environment ID
ENV0_PROJECT_ID	The Project ID of the deployed Environment
ENV0_PROJECT_NAME	The Project Name of the deployed Environment
ENV0_DEPLOYMENT_LOG_ID	The deployment ID
ENV0_DEPLOYMENT_TYPE	The deployment type. One of deploy / destroy / prPlan / driftDetection / task
ENV0_DEPLOYMENT_REVISION	The revision (available only when deployment revision is defined)
ENV0_WORKSPACE_NAME	The Terraform Workspace name used in the Environment
ENV0_ROOT_DIR	The root repository path
ENV0_ORGANIZATION_ID	Your env zero organization ID
ENV0_ORGANIZATION_NAME	Your env zero organization name
ENV0_TEMPLATE_ID	The deployed Template ID
ENV0_ENVIRONMENT_URL	Link to the deployed Environment in the env zero web app
ENV0_DISCOVERY_REPOSITORY_URL	The repository the Environment was discovered from. (available on every deployment of an Environment created by Environment discovery, whatever started it)
ENV0_DISCOVERY_VCS	The VCS provider of that repository. (available only on the PR Plan that a pull request on the discovery repository starts)

See this example for using these variables in our public templates repo.

Note

You can use exposed env zero variables in the UI using the ${…} notation:

env zero System Files

env zero generates a few files during runtime that can be used in a custom flow (and terraform) to help you programmatically access environment variables and Terraform variables.

env0.system-env-vars.json

This file contains all the environment variables generated by env zero through metadata. See above for list of available environment variables and their respective descriptions.

env0.auto.tfvars.json

This file contains all the variables defined through the UI in json format. For example, you can parse the file using jq.

cat env0.auto.tfvars.json | jq -r '.'

env0.env-vars.json

This file is similar to the Terraform variables, and contains all the environment variables defined through the env zero UI. This can also be parsed using jq.

Accessing Environment Variables in Terraform

You can use this file to access environment variables within your Terraform code. Simply read and jsondecode the env0.system-env-vars.json file to get access to the metadata. See example:

locals {
  env_vars = jsondecode(file("env0.system-env-vars.json"))
}

output "env0_environment_id" {
  value = local.env_vars.ENV0_ENVIRONMENT_ID
}

Using Additional CLI Tools in Custom Flows

Custom flows execute your shell commands within a dedicated deployment container. This container comes pre-loaded with a standard set of utilities, including common OS tools and package managers.

However, if your custom flow hooks require specific command-line interface (CLI) tools, such as the AWS CLI or kubectl, you must explicitly configure the deployment to install them.

To add these tools, set the ENV0_INSTALLED_TOOLS environment variable with a comma-separated list of the tools you need.

For example, to make the AWS CLI and kubectl available in your deployment container, you would set the following variable:

ENV0_INSTALLED_TOOLS=aws,kubectl

You can set this variable at any level. For instance, if one project uses kubectl and another uses the aws CLI, it’s best to set the variable at the project level to ensure each environment has only the tools it requires.

Available tools

Tool	Description
kustomize	A Kubernetes native tool for customizing application configuration.
infracost	Shows cloud cost estimates for Terraform, helping to manage budgets.
kubelogin	A kubectl plugin for non-interactive authentication with OIDC providers.
opa	Open Policy Agent, a general-purpose policy engine for cloud native environments.
aws	The official command-line tool for Amazon Web Services.
gcloud	The primary command-line tool for Google Cloud Platform.
az	The official command-line tool for Microsoft Azure.
helm	The package manager for Kubernetes, used to find, share, and use software.
dyff	A diff tool for YAML files, especially useful for Kubernetes manifests.
kubectl	The command-line tool for controlling Kubernetes clusters.
terratag	A CLI tool for managing tags across all resources in a Terraform project.
pwsh	PowerShell, a cross-platform task automation and configuration management framework. (Not supported on the arm64 architecture or when running in strictSecurityContext mode. If you need this functionality, you must extend the image.)
gke-gcloud-auth-plugin	A plugin to handle GKE authentication for kubectl and other clients.

Version Updates

env0 will periodically update tool versions to address security considerations or end-of-life (EoL) versions. Customers will be notified in advance whenever an update introduces breaking changes.

You can always check the currently installed versions under the Init Packages step.

Exporting your own environment variables

To export your own Environment Variable to use in any downstream custom flow you should write a shell command in the format: echo KEY=VALUE >> $ENV0_ENV.
In later steps of the same deployment, you will be able to use $KEY to access the stored value.

Note

For _env0.yml _ files that use schema version 2, when using multiple commands within a given step e.g., terraform:plan-before, all commands are executed inside the same shell. Exported Environment variables like these are only available for the steps that follow. For values within the same shell, using the usual export key=value still works.

Forcing Manual Approvals

Whenever ENV0_REQUIRES_APPROVAL=true occurs, env zero will force deployment to require approval, regardless of the Auto Approval setting in the Environment.

For example, instead of failing a deployment based on some policies, we can use this field to help ensure that others can still review and proceed with the deployment.

Usage:
- echo ENV0_REQUIRES_APPROVAL=true >> $ENV0_ENV

Note

If the project has Skip apply when plan has no changes enabled and the plan reports no resource changes, the deployment finishes without waiting for approval and the apply step is marked as skipped. There is nothing to apply, so there is nothing to approve. This applies to workflow sub-environments as well, through the policies of the project they belong to.

An approval policy that returns a pending result still holds the deployment for approval, whether or not the plan has changes.

Repository custom flows

The Repository Custom Flows organization policy controls whether env0.yml files in a template’s repository run. Custom flows assigned in env zero, such as project-level custom flows, are not affected by this policy.

The Repository Custom Flows policy is rolling out to organizations. If you do not see it in Organization Settings > Policies, it is not yet available for your organization, and repository env0.yml files run as described on this page.

To change the policy, you need the Organization Admin role or the Edit Organization Settings permission. Go to Organization Settings > Policies, select an option under Repository Custom Flows, and click Save.

Option	What runs
Allow (default)	env0.yml files in the template’s repository run, on every deployment.
Verified users only	A changed env0.yml runs only for a user with the Run Tasks permission on the environment, and in a pull request only for a verified VCS user with that permission. See Verified users only.
Block	env0.yml files in the template’s repository never run. Only custom flows assigned in env zero run.

When Block is selected, the Custom Flows Project Policy has no effect, because there is no repository custom flow to override or merge with.

Self-hosted agents must run version 3.0.284 or later to enforce Block, and version 5.5.13 or later to enforce Verified users only. On an older agent, a deployment that env zero must give a custom flow other than the clone’s env0.yml fails with an error that asks you to upgrade the agent to the required version. Upgrade your agents before you select either option.

Verified users only

With Verified users only, env zero checks every pull request plan that changes env0.yml, and every PR comment command. To decide if a pull request changes env0.yml, env zero compares the whole pull request with its target branch, not only the latest push.

The pull request’s env0.yml runs only when the user who did the last relevant operation meets both requirements:

• They connected their VCS account to env zero, so their VCS user is verified. A VCS user ID typed in Personal Settings does not count. See Setting your VCS user in env zero.
• They have the Run Tasks permission on the environment.

The last relevant operation is:

• For a pull request plan, the push to the pull request.
• For a comment command such as env0 plan or env0 apply, the comment. The pull request plan of the latest push must also have run the pull request’s env0.yml. A trusted commenter does not approve an env0.yml that an untrusted user pushed.

If the user does not meet the requirements, env zero runs the env0.yml from the branch the environment tracks instead of the one in the pull request. env zero then comments on the pull request with the environments affected and the reason for each one:

• The user has no verified VCS user connected to env zero.
• The user does not have the Run Tasks permission on the environment.
• The latest push came from a user who is not a verified VCS user with the Run Tasks permission.

Deploying another revision

Verified users only also applies when a user deploys or destroys an environment with a revision other than the one the environment tracks, for example from the env zero UI, the API or the CLI:

• If env0.yml is different in the two revisions, or env zero cannot compare them, the user needs the Run Tasks permission on the environment. Without it, the request fails with an error that names the permission.
• If env0.yml is the same in the two revisions, the deployment runs the env0.yml from the tracked revision.

Changing an environment’s template to one in another repository also needs the Run Tasks permission.

You can connect GitHub, GitLab, Azure DevOps and Bitbucket accounts. For other VCS providers, a pull request that changes env0.yml always runs the tracked branch’s env0.yml.

Add a CODEOWNERS rule for env0.yml files on the branch your environments track, and require code owner review in your VCS provider’s branch protection. A change to env0.yml then reaches the tracked branch, where it runs on every deployment, only after a code owner approves it.

FAQ

Q. How can I use Custom Flows to add new Terraform variables?

A. The recommended approach to adding Terraform variables is to use variable definitions files, specifically the .auto.tfvars or .auto.tfvars.json files, to pass additional variables to your Terraform code.

Note: Please be aware of the Terraform Variable Definition Precedence order, defined here.

Another way is to create new Environment Variables inside the custom flow in the form of TF_VAR_key=value. To do that, please follow our guide for exporting your environment variables.

Q. I noticed that env zero runs init again after approval. How can I run my custom flow only once per deployment?

A. You can take advantage of the ENV0_REVIEWER_NAME field to see if the plan has already been approved.

For example:

deploy:
  steps:
    terraformInit:
      before: # run ./myScript.sh only on pre-approval
        - if [ -z "$ENV0_REVIEWER_NAME" ]; then ./myScript.sh; fi

Q. Do env zero SaaS runners use static IP addresses? I would like to whitelist env zero’s runners to access my network.
A. Yes, env zero uses those IP addresses for all the outbound requests.

Q. How do I export a multi-line string into an environment variable?
A. Wrap the value in double quotes. env zero reads $ENV0_ENV as a dotenv file, which keeps line breaks inside a double-quoted value: printf 'MY_PEM_FILE="%s"\n' "$(cat "$PEM_FILE")" >> $ENV0_ENV

GitHub API limits

When downloading resources from GitHub (e.g., tflint), make sure you use an authenticated call (e.g., curl -u USER:TOKEN). Otherwise, you will likely run into API limits, as unauthenticated calls are pooled with other SaaS users, with a much lower API limit. See GitHub Rate Limiting.

Next steps

• Switch to the v2 step schema so each step shows as its own deployment step
• Pass variables into custom flow commands
• Build and push Docker images in a custom flow on a self-hosted Kubernetes agent