Run Claude with your real secrets while redacting them from its output.

Prerequisite: install Claude Code to get the claude command.

$ curl -fsSL https://claude.ai/install.sh | bash
$ claude --version

$ echo "HELLO=World" > .env
$ dotenvx spec
$ dotenvx run -- claude -p 'Run dotenvx get HELLO and echo back just Hello VALUE' --dangerously-skip-permissions
Hello [REDACTED]

see Claude redaction guide

Run Codex with your real secrets while redacting them from its output.

Prerequisite: install the Codex CLI to get the codex command.

$ npm install -g @openai/codex
$ codex --version

$ echo "HELLO=World" > .env
$ dotenvx spec
$ dotenvx run -- codex exec 'Run dotenvx get HELLO and echo back just Hello VALUE' --skip-git-repo-check
Hello [REDACTED]

see Codex redaction guide

Run Cursor with your real secrets while redacting them from its output.

Prerequisite: install the separate Cursor CLI. Installing the Cursor desktop app does not necessarily install the agent command.

$ curl https://cursor.com/install -fsS | bash
$ agent --version

$ echo "HELLO=World" > .env
$ dotenvx spec
$ dotenvx run -- agent -p --force 'Run dotenvx get HELLO and echo back just Hello VALUE' --output-format text
Hello [REDACTED]

see Cursor redaction guide

Run with secrets resolved directly from 1Password.

$ echo "HELLO=op://Personal/hello/password" > .env
$ dotenvx run -- sh -c 'echo Hello $HELLO'
Hello World

see 1Password guide

Run with secrets resolved directly from Bitwarden Password Manager.

$ echo 'HELLO="bw://My Hello Login/password"' > .env
$ dotenvx run -- sh -c 'echo Hello $HELLO'
Hello World

Install the Bitwarden Password Manager CLI before running dotenvx.

// package.json
{
  "type": "module",
  "dependencies": {
    "chalk": "^5.3.0"
  }
}

// index.ts
import chalk from 'chalk'
console.log(chalk.blue(Hello ${process.env.HELLO}))

$ npm install
$ echo "HELLO=World" > .env
$ dotenvx run -- npx tsx index.ts
Hello World

Preface Astro scripts with dotenvx run -- and read your env values in Astro.

{
  "scripts": {
    "dev": "dotenvx run -- astro dev",
    "build": "dotenvx run -- astro build",
    "preview": "dotenvx run -- astro preview"
  }
}

export async function GET() {
  return new Response(
    JSON.stringify({
      HELLO: process.env.HELLO,
    }),
    {
      status: 200,
      headers: {
        "Content-Type": "application/json",
      },
    }
  );
}

see astro guide

Preface Expo scripts with dotenvx run --.

{
  "scripts": {
    "start": "dotenvx run -- expo start",
    "reset-project": "node ./scripts/reset-project.js",
    "android": "dotenvx run -- expo start --android",
    "ios": "dotenvx run -- expo start --ios",
    "web": "dotenvx run -- expo start --web",
    "lint": "expo lint"
  }
}

see expo guide

Install Dotenvx and @dotenvx/next-env.

$ npm install @dotenvx/dotenvx
$ npm install @dotenvx/next-env

Override @next/env in your package.json.

{
  "overrides": {
    "@next/env": "npm:@dotenvx/next-env"
  }
}

Encrypt your .env file.

$ npx dotenvx encrypt
◈ encrypted (.env)

Your encrypted secrets are automatically injected and readable in Next.js.

import { NextResponse } from 'next/server'
export async function GET() {
  return NextResponse.json({
    HELLO: process.env.HELLO
  })
}

Set DOTENV_PRIVATE_KEY in production before deploying.

$ dotenvx encrypt -f .env.txt

// src/index.js
import envSrc from '../.env.txt'
import dotenvx from '@dotenvx/dotenvx'
const config = dotenvx.config({ envs: [{ type: 'env', value: envSrc, privateKeyName: 'DOTENV_PRIVATE_KEY' }] })
const envx = config.parsed
export default {
  async fetch(request, env, ctx) {
    return new Response(Hello ${envx.HELLO})
  }
}

"scripts": {
  "deploy": "wrangler deploy",
  "dev": "wrangler dev --var $(dotenvx keypair -f .env.txt --format=colon)",
  "start": "wrangler dev --var $(dotenvx keypair -f .env.txt --format=colon)",
}

$ echo "HELLO=Test" > .env.test
$ echo "console.log('Hello ' + process.env.HELLO)" > index.js
$ bun index.js
Hello undefined
$ dotenvx run -f .env.test -- bun index.js
Hello Test

$ echo "HELLO=World" > .env
$ echo "console.log('Hello ' + Deno.env.get('HELLO'))" > index.ts
$ deno run --allow-env index.ts
Hello undefined
$ dotenvx run -- deno run --allow-env index.ts
Hello World

[!WARNING]

Some of you are attempting to use the npm module directly with deno run. Don't, because deno currently has incomplete support for these encryption ciphers.

>

> $ deno run -A npm:@dotenvx/dotenvx encrypt

Instead, use dotenvx as designed, by installing the cli as a binary - via curl, brew, etc.

$ echo "HELLO=World" > .env
$ echo 'import os;print("Hello " + os.getenv("HELLO", ""))' > index.py
$ dotenvx run -- python3 index.py
Hello World

see extended python guide

$ echo "HELLO=World" > .env
$ echo '<?php echo "Hello {$_SERVER["HELLO"]}\n";' > index.php
$ dotenvx run -- php index.php
Hello World

see extended php guide

$ echo "HELLO=World" > .env
$ echo 'puts "Hello #{ENV["HELLO"]}"' > index.rb
$ dotenvx run -- ruby index.rb
Hello World

see extended ruby guide

$ echo "HELLO=World" > .env
$ echo 'package main; import ("fmt"; "os"); func main() { fmt.Printf("Hello %s\n", os.Getenv("HELLO")) }' > main.go
$ dotenvx run -- go run main.go
Hello World

see extended go guide

$ echo "HELLO=World" > .env
$ echo 'fn main() {let hello = std::env::var("HELLO").unwrap_or("".to_string());println!("Hello {hello}");}' > src/main.rs
$ dotenvx run -- cargo run
Hello World

see extended rust guide

$ echo "HELLO=World" > .env
$ echo 'public class Index { public static void main(String[] args) { System.out.println("Hello " + System.getenv("HELLO")); } }' > index.java
$ dotenvx run -- java index.java
Hello World

$ echo "HELLO=World" > .env
$ echo '(println "Hello" (System/getenv "HELLO"))' > index.clj
$ dotenvx run -- clojure -M index.clj
Hello World

$ echo "HELLO=World" > .env
$ echo 'fun main() { val hello = System.getenv("HELLO") ?: ""; println("Hello $hello") }' > index.kt
$ kotlinc index.kt -include-runtime -d index.jar
$ dotenvx run -- java -jar index.jar
Hello World

$ dotnet new console -n HelloWorld -o HelloWorld
$ cd HelloWorld
$ echo "HELLO=World" | Out-File -FilePath .env -Encoding utf8
$ echo 'Console.WriteLine($"Hello {Environment.GetEnvironmentVariable("HELLO")}");' > Program.cs
$ dotenvx run -- dotnet run
Hello World

$ echo "HELLO=World" > .env
$ dotenvx run --quiet -- sh -c 'echo Hello $HELLO'
Hello World

$ dotenvx run -- next dev
$ dotenvx run -- npm start
$ dotenvx run -- bin/rails s
$ dotenvx run -- php artisan serve

see framework guides

$ docker run -it --rm -v $(pwd):/app dotenv/dotenvx run -- node index.js

Or in any image:

FROM node:latest
RUN echo "HELLO=World" > .env && echo "console.log('Hello ' + process.env.HELLO)" > index.js
RUN curl -fsS https://dotenvx.sh/install.sh | sh
CMD ["/usr/local/bin/dotenvx", "run", "--", "echo", "Hello $HELLO"]

see docker guide

name: build
on: [push]
jobs:
  build:
    runs-on: ubuntu-latest
    steps:
    - uses: actions/checkout@v3
    - uses: actions/setup-node@v3
      with:
        node-version: 16
    - run: curl -fsS https://dotenvx.sh/install.sh | sh
    - run: dotenvx run -- node build.js
      env:
        DOTENV_KEY: ${{ secrets.DOTENV_KEY }}

see github actions guide

# heroku
heroku buildpacks:add https://github.com/dotenvx/heroku-buildpack-dotenvx
docker
RUN curl -fsS https://dotenvx.sh | sh
vercel
npm install @dotenvx/dotenvx --save

see platform guides

// pm2
"scripts": {
  "start": "dotenvx run -- pm2-runtime start ecosystem.config.js --env production"
},

see process manager guides

# alternatively use npx
$ npx @dotenvx/dotenvx run -- node index.js
$ npx @dotenvx/dotenvx run -- next dev
$ npx @dotenvx/dotenvx run -- npm start

$ npm install @dotenvx/dotenvx --save

{
  "scripts": {
    "start": "./node_modules/.bin/dotenvx run -- node index.js"
  },
  "dependencies": {
    "@dotenvx/dotenvx": "^0.5.0"
  }
}

$ npm run start
> start
> ./node_modules/.bin/dotenvx run -- node index.js
[[email protected]] injecting env (1) from .env.production
Hello World

Reference and expand variables already on your machine for use in your .env file.

# .env
USERNAME="username"
DATABASE_URL="postgres://${USERNAME}@localhost/my_database"

// index.js
console.log('DATABASE_URL', process.env.DATABASE_URL)

$ dotenvx run --debug -- node index.js
[[email protected]] injecting env (2) from .env
DATABASE_URL postgres://username@localhost/my_database

Add the output of a command to one of your variables in your .env file.

# .env
DATABASE_URL="postgres://$(whoami)@localhost/my_database"

// index.js
console.log('DATABASE_URL', process.env.DATABASE_URL)

$ dotenvx run --debug -- node index.js
[[email protected]] injecting env (1) from .env
DATABASE_URL postgres://yourusername@localhost/my_database

$ echo "HELLO=local" > .env.local
$ echo "HELLO=World" > .env
$ dotenvx run -f .env.local,.env -- node index.js
[[email protected]] injecting env (1) from .env.local,.env
Hello local

Comma-separate multiple files after a single -f. Subsequent files do NOT override pre-existing variables defined in previous files or env. This follows historic principle. For example, above local wins – from the first file.

$ echo "HELLO=local" > .env.local
$ echo "HELLO=World" > .env
$ dotenvx run -f .env.local,.env --overload -- node index.js
[[email protected]] injecting env (1) from .env.local,.env
Hello World

Note that with --overload subsequent files DO override pre-existing variables defined in previous files.

$ echo "HELLO=production" > .env.production
$ dotenvx run -f .env.production --verbose -- node index.js
[dotenvx][verbose] injecting env from /path/to/.env.production
[dotenvx][verbose] HELLO set
[[email protected]] injecting env (1) from .env.production
Hello production

$ echo "HELLO=production" > .env.production
$ dotenvx run -f .env.production --debug -- node index.js
[dotenvx][debug] configuring options
[dotenvx][debug] {"envFile":[".env.production"]}
[dotenvx][verbose] injecting env from /path/to/.env.production
[dotenvx][debug] reading env from /path/to/.env.production
[dotenvx][debug] parsing env from /path/to/.env.production
[dotenvx][debug] {"HELLO":"production"}
[dotenvx][debug] writing env from /path/to/.env.production
[dotenvx][verbose] HELLO set
[dotenvx][debug] HELLO set to production
[[email protected]] injecting env (1) from .env.production
Hello production

Use --quiet to suppress all output (except errors).

$ echo "HELLO=production" > .env.production
$ dotenvx run -f .env.production --quiet -- node index.js
Hello production

You can also set DOTENV_QUIET=true.

$ DOTENV_QUIET=true dotenvx run -f .env.production -- node index.js
Hello production

Set --log-level to whatever you wish. For example, to suppress warnings (risky), set log level to error:

$ echo "HELLO=production" > .env.production
$ dotenvx run -f .env.production --log-level=error -- node index.js
Hello production

Available log levels are error, warn, info, verbose, debug, silly

Load envs using Next.js' convention or dotenv-flow convention. Set --convention to nextjs or flow:

$ echo "HELLO=development local" > .env.development.local
$ echo "HELLO=local" > .env.local
$ echo "HELLO=development" > .env.development
$ echo "HELLO=env" > .env
$ dotenvx run --convention=nextjs -- node index.js
Hello development local
$ dotenvx run --convention=flow -- node index.js
Hello development local

(more conventions available upon request)

$ echo "HELLO=World" > .env
$ dotenvx encrypt
$ echo "console.log('Hello ' + process.env.HELLO)" > index.js
$ dotenvx run -- node index.js
[[email protected]] injecting env (2) from .env
Hello World

$ echo "HELLO=Production" > .env.production
$ dotenvx encrypt -f .env.production
$ echo "console.log('Hello ' + process.env.HELLO)" > index.js
$ DOTENV_PRIVATE_KEY_PRODUCTION="<.env.production private key>" dotenvx run -- node index.js
[[email protected]] injecting env (2) from .env.production
Hello Production

Note the DOTENV_PRIVATE_KEY_PRODUCTION ends with _PRODUCTION. This instructs dotenvx run to load the .env.production file.

$ echo "HELLO=Ci" > .env.ci
$ dotenvx encrypt -f .env.ci
$ echo "console.log('Hello ' + process.env.HELLO)" > index.js
$ DOTENV_PRIVATE_KEY_CI="<.env.ci private key>" dotenvx run -- node index.js
[[email protected]] injecting env (2) from .env.ci
Hello Ci

Note the DOTENV_PRIVATE_KEY_CI ends with _CI. This instructs dotenvx run to load the .env.ci file. See the pattern?

$ dotenvx set HELLO World -f .env
$ dotenvx set HELLO Production -f .env.production
$ echo "console.log('Hello ' + process.env.HELLO)" > index.js
$ DOTENV_PRIVATE_KEY="<.env private key>" DOTENV_PRIVATE_KEY_PRODUCTION="<.env.production private key>" dotenvx run -- node index.js
[[email protected]] injecting env (3) from .env, .env.production
Hello World

Note the DOTENV_PRIVATE_KEY instructs dotenvx run to load the .env file and the DOTENV_PRIVATE_KEY_PRODUCTION instructs it to load the .env.production file. See the pattern?

Point -f at a directory to load the .env inside it. From a workspace, this makes a shared root .env available without repeating its filename.

my-monorepo/
  .env
  apps/
    web/
      index.js

$ dotenvx encrypt
$ cd apps/web
$ dotenvx get HELLO -f ../..
World
$ dotenvx run -f ../.. -- node index.js
[[email protected]] injecting env (1) from ../../.env
Hello World

With the private key in your OS secret store, dotenvx finds it automatically from any workspace on the same machine.

The directory also becomes the base when using a convention:

$ dotenvx run -f ../.. --convention=nextjs -- node index.js
[[email protected]] injecting env (1) from ../../.env.development.local, ../../.env.local, ../../.env.development, ../../.env
Hello development local

For a workspace with its own encrypted .env, run from that workspace:

$ dotenvx run -- node index.js

$ echo "HELLO=World" > .env
$ dotenvx encrypt --stdout
$ dotenvx encrypt --stdout > .env.encrypted

secp256k1 is a well-known and battle tested curve, in use with Bitcoin and other cryptocurrencies, but we are open to adding support for more curves.

If your organization's compliance department requires NIST approved curves or other curves like curve25519, please reach out at [email protected].

Reference and expand variables already on your machine for use in your .env file.

# .env
USERNAME="username"
DATABASE_URL="postgres://${USERNAME}@localhost/my_database"

// index.js
console.log('DATABASE_URL', process.env.DATABASE_URL)

$ dotenvx run --debug -- node index.js
[[email protected]] injecting env (2) from .env
DATABASE_URL postgres://username@localhost/my_database

Use default values when environment variables are unset or empty.

# .env
Default value syntax: use value if set, otherwise use default
DATABASE_HOST=${DB_HOST:-localhost}
DATABASE_PORT=${DB_PORT:-5432}
Alternative syntax (no colon): use value if set, otherwise use default
API_URL=${API_BASE_URL-https://api.example.com}

// index.js
console.log('DATABASE_HOST', process.env.DATABASE_HOST)
console.log('DATABASE_PORT', process.env.DATABASE_PORT)
console.log('API_URL', process.env.API_URL)

$ dotenvx run --debug -- node index.js
[[email protected]] injecting env (3) from .env
DATABASE_HOST localhost
DATABASE_PORT 5432
API_URL https://api.example.com

Use alternate values when environment variables are set and non-empty.

# .env
NODE_ENV=production
Alternate value syntax: use alternate if set and non-empty, otherwise empty
DEBUG_MODE=${NODE_ENV:+false}
LOG_LEVEL=${NODE_ENV:+error}
Alternative syntax (no colon): use alternate if set, otherwise empty  
CACHE_ENABLED=${NODE_ENV+true}

// index.js
console.log('NODE_ENV', process.env.NODE_ENV)
console.log('DEBUG_MODE', process.env.DEBUG_MODE)
console.log('LOG_LEVEL', process.env.LOG_LEVEL)
console.log('CACHE_ENABLED', process.env.CACHE_ENABLED)

$ dotenvx run --debug -- node index.js
[[email protected]] injecting env (4) from .env
NODE_ENV production
DEBUG_MODE false
LOG_LEVEL error
CACHE_ENABLED true

Complete reference for variable interpolation patterns supported by dotenvx:

# .env
DEFINED_VAR=hello
EMPTY_VAR=
UNDEFINED_VAR is not set
Default value syntax - use variable if set/non-empty, otherwise use default
TEST1=${DEFINED_VAR:-fallback}     # Result: "hello"
TEST2=${EMPTY_VAR:-fallback}       # Result: "fallback"  
TEST3=${UNDEFINED_VAR:-fallback}   # Result: "fallback"
Default value syntax (no colon) - use variable if set, otherwise use default
TEST4=${DEFINED_VAR-fallback}      # Result: "hello"
TEST5=${EMPTY_VAR-fallback}        # Result: "" (empty, but set)
TEST6=${UNDEFINED_VAR-fallback}    # Result: "fallback"
Alternate value syntax - use alternate if variable is set/non-empty, otherwise empty
TEST7=${DEFINED_VAR:+alternate}    # Result: "alternate"
TEST8=${EMPTY_VAR:+alternate}      # Result: "" (empty)
TEST9=${UNDEFINED_VAR:+alternate}  # Result: "" (empty)
Alternate value syntax (no colon) - use alternate if variable is set, otherwise empty  
TEST10=${DEFINED_VAR+alternate}    # Result: "alternate"
TEST11=${EMPTY_VAR+alternate}      # Result: "alternate" (empty but set)
TEST12=${UNDEFINED_VAR+alternate}  # Result: "" (empty)

Key differences:

• :- vs -: The colon makes empty values trigger the fallback
• :+ vs +: The colon makes empty values not trigger the alternate
• Default syntax (-): Use variable value or fallback
• Alternate syntax (+): Use alternate value or empty string

Add the output of a command to one of your variables in your .env file.

# .env
DATABASE_URL="postgres://$(whoami)@localhost/my_database"

// index.js
console.log('DATABASE_URL', process.env.DATABASE_URL)

$ dotenvx run --debug -- node index.js
[[email protected]] injecting env (1) from .env
DATABASE_URL postgres://yourusername@localhost/my_database

Prevent your shell from expanding inline $VARIABLES before dotenvx has a chance to inject it. Use a subshell.

$ dotenvx run --env="HELLO=World" -- sh -c 'echo Hello $HELLO'
Hello World

Dotenvx supports multiline values. This is particularly useful in conjunction with Docker - which does not support multiline values.

# .env
MULTILINE_PEM="-----BEGIN PUBLIC KEY-----
MIIBIjANBgkqhkiG9w0BAQEFAAOCAQ8AMIIBCgKCAQEAnNl1tL3QjKp3DZWM0T3u
LgGJQwu9WqyzHKZ6WIA5T+7zPjO1L8l3S8k8YzBrfH4mqWOD1GBI8Yjq2L1ac3Y/
bTdfHN8CmQr2iDJC0C6zY8YV93oZB3x0zC/LPbRYpF8f6OqX1lZj5vo2zJZy4fI/
kKcI5jHYc8VJq+KCuRZrvn+3V+KuL9tF9v8ZgjF2PZbU+LsCy5Yqg1M8f5Jp5f6V
u4QuUoobAgMBAAE=
-----END PUBLIC KEY-----"

// index.js
console.log('MULTILINE_PEM', process.env.MULTILINE_PEM)

$ dotenvx run -- node index.js
MULTILINE_PEM -----BEGIN PUBLIC KEY-----
MIIBIjANBgkqhkiG9w0BAQEFAAOCAQ8AMIIBCgKCAQEAnNl1tL3QjKp3DZWM0T3u
LgGJQwu9WqyzHKZ6WIA5T+7zPjO1L8l3S8k8YzBrfH4mqWOD1GBI8Yjq2L1ac3Y/
bTdfHN8CmQr2iDJC0C6zY8YV93oZB3x0zC/LPbRYpF8f6OqX1lZj5vo2zJZy4fI/
kKcI5jHYc8VJq+KCuRZrvn+3V+KuL9tF9v8ZgjF2PZbU+LsCy5Yqg1M8f5Jp5f6V
u4QuUoobAgMBAAE=
-----END PUBLIC KEY-----

Unlike other dotenv libraries, dotenvx attempts to unblock you with contextual help.

For example, when missing a custom .env file:

$ dotenvx run -f .env.missing -- echo $HELLO
[MISSING_ENV_FILE] missing file (/Users/scottmotte/Code/dotenvx/playground/apr-16/.env.missing). fix: [echo "HELLO=World" > .env.missing]

or when missing a KEY:

$ echo "HELLO=World" > .env
$ dotenvx get GOODBYE
[MISSING_KEY] missing key (GOODBYE)

Resolve 1Password op:// directly from your .env file.

# .env
API_KEY=op://Personal/my_api_key/password

Install the 1Password CLI and authenticate with op. Dotenvx automatically reads op:// values through op read before injecting them.

$ dotenvx run -- node index.js

Use --no-1password to leave op:// values unresolved.

$ dotenvx run --no-1password -- node index.js

The same flag is available for dotenvx get and dotenvx check.

Resolve bw:// references directly from your .env file through the Bitwarden Password Manager CLI.

# .env
API_KEY="bw://My GitHub Account/password"

The reference format is bw:///. The item can be a human-readable Bitwarden search term or an exact item UUID. Quote the dotenv value when the item name contains spaces.

Supported fields are:

• username
• password
• uri

Human-readable names are convenient, but must identify a single item. If multiple items match, Bitwarden returns an error. Use the item UUID when you need an unambiguous reference.

If Bitwarden cannot resolve a reference, dotenvx reports the error and leaves the original bw:// value unresolved.

Use --no-bitwarden to skip Bitwarden resolution intentionally.

$ dotenvx run --no-bitwarden -- node index.js

The same flag is available for dotenvx get and dotenvx check.

Run a command using the .env file in a directory. This is useful with monorepos.

$ dotenvx run -f ../.. -- node index.js
[[email protected]] injecting env (1) from ../../.env
Hello World

Compose multiple .env files for environment variables loading, as you need.

$ echo "HELLO=local" > .env.local
$ echo "HELLO=World" > .env
$ echo "console.log('Hello ' + process.env.HELLO)" > index.js
$ dotenvx run -f .env.local,.env -- node index.js
[[email protected]] injecting env (1) from .env.local, .env
Hello local

Comma-separate multiple files after a single -f. Subsequent files do NOT override pre-existing variables defined in previous files or env. This follows historic principle. For example, above local wins – from the first file.

Set environment variables as a simple KEY=value string pair.

$ echo "HELLO=World" > .env
$ echo "console.log('Hello ' + process.env.HELLO)" > index.js
$ dotenvx run --env HELLO=String -f .env -- node index.js
[[email protected]] injecting env (1) from .env, and --env flag
Hello String

Run any command with real environment variables while automatically redacting values selected by your Envspec from stdout and stderr.

$ echo "SECRET=super-secret-value" > .env
$ echo "PUBLIC_VALUE=visible-value" >> .env
$ echo "console.log(process.env.SECRET, process.env.PUBLIC_VALUE)" > index.js
$ dotenvx spec
$ dotenvx run --quiet -- node index.js
[REDACTED] visible-value

With an Envspec, all values are redacted by default. Use redacted: false on individual declarations to leave those values visible. spec generates these settings, including public-name exceptions. If an existing environment variable takes precedence, its effective value follows the same rules. Matching is exact, so transformed or derived values are not redacted.

Without an Envspec, dotenvx run --redact -- yourcommand redacts loaded values except recognized public names (containing PUBLIC or starting with VITE, case-sensitive). _PLAIN values are redacted too. Without either an Envspec or --redact, output is not redacted. An Envspec always takes precedence: its rules apply automatically, even when --redact is passed.

When stdin, stdout, and stderr are attached to a terminal, dotenvx preserves interactive behavior on macOS and Linux systems with script available. Piped and redirected commands continue to use normal stdin, stdout, and stderr streams.

Run Claude in print mode with real environment variables while redacting any values it prints.

$ echo "SECRET=super-secret-value" > .env
$ dotenvx spec
$ dotenvx run --quiet -- claude -p 'Print the value of $SECRET'
[REDACTED]

Start a fully interactive Claude session. Claude receives the real values, but any values it prints are redacted.

$ echo "SECRET=super-secret-value" > .env
$ dotenvx spec
$ dotenvx run --quiet -- claude

Run Codex non-interactively with real environment variables while redacting any values it prints.

$ echo "SECRET=super-secret-value" > .env
$ dotenvx spec
$ dotenvx run --quiet -- codex exec 'Print the value of $SECRET'
[REDACTED]

Start a fully interactive Codex session. Codex receives the real values, but any values it prints are redacted.

$ echo "SECRET=super-secret-value" > .env
$ dotenvx spec
$ dotenvx run --quiet -- codex

Override existing env variables. These can be variables already on your machine or variables loaded as files consecutively. The last variable seen will 'win'.

$ echo "HELLO=local" > .env.local
$ echo "HELLO=World" > .env
$ echo "console.log('Hello ' + process.env.HELLO)" > index.js
$ dotenvx run -f .env.local,.env --overload -- node index.js
[[email protected]] injecting env (1) from .env.local, .env
Hello World

Note that with --overload subsequent files DO override pre-existing variables defined in previous files.

When deploying applications in containers or cloud environments, you often need to override specific environment variables at runtime without modifying committed .env files. By default, dotenvx follows the historic dotenv principle: environment variables already present take precedence over .env files.

# .env.prod contains: MODEL_REGISTRY=registry.company.com/models/v1
$ echo "MODEL_REGISTRY=registry.company.com/models/v1" > .env.prod
$ echo "console.log('MODEL_REGISTRY:', process.env.MODEL_REGISTRY)" > app.js
Without environment variable set - uses .env.prod value
$ dotenvx run -f .env.prod -- node app.js
MODEL_REGISTRY: registry.company.com/models/v1
With environment variable set (e.g., via Azure Container Service) - environment variable takes precedence
$ MODEL_REGISTRY=registry.azure.com/models/v2 dotenvx run -f .env.prod -- node app.js
MODEL_REGISTRY: registry.azure.com/models/v2
To force .env.prod to override environment variables, use --overload
$ MODEL_REGISTRY=registry.azure.com/models/v2 dotenvx run -f .env.prod --overload -- node app.js
MODEL_REGISTRY: registry.company.com/models/v1

For container deployments: Set environment variables through your cloud provider's UI/configuration (Azure Container Service, AWS ECS, etc.) to override specific values from committed .env files without rebuilding your application.

Decrypt your encrypted .env by setting DOTENV_PRIVATE_KEY before dotenvx run.

$ touch .env
$ dotenvx set HELLO encrypted
$ echo "console.log('Hello ' + process.env.HELLO)" > index.js
check your .env.keys files for your privateKey
$ DOTENV_PRIVATE_KEY="122...0b8" dotenvx run -- node index.js
[[email protected]] injecting env (2) from .env
Hello encrypted

Decrypt your encrypted .env.production by setting DOTENV_PRIVATE_KEY_PRODUCTION before dotenvx run. Alternatively, this can be already set on your server or cloud provider.

$ touch .env.production
$ dotenvx set HELLO "production encrypted" -f .env.production
$ echo "console.log('Hello ' + process.env.HELLO)" > index.js
check .env.keys for your privateKey
$ DOTENV_PRIVATE_KEY_PRODUCTION="122...0b8" dotenvx run -- node index.js
[[email protected]] injecting env (2) from .env.production
Hello production encrypted

Note the DOTENV_PRIVATE_KEY_PRODUCTION ends with _PRODUCTION. This instructs dotenvx run to load the .env.production file.

Decrypt your encrypted .env.ci by setting DOTENV_PRIVATE_KEY_CI before dotenvx run. Alternatively, this can be already set on your server or cloud provider.

$ touch .env.ci
$ dotenvx set HELLO "ci encrypted" -f .env.ci
$ echo "console.log('Hello ' + process.env.HELLO)" > index.js
check .env.keys for your privateKey
$ DOTENV_PRIVATE_KEY_CI="122...0b8" dotenvx run -- node index.js
[[email protected]] injecting env (2) from .env.ci
Hello ci encrypted

Note the DOTENV_PRIVATE_KEY_CI ends with _CI. This instructs dotenvx run to load the .env.ci file. See the pattern?

Decrypt your encrypted .env and .env.production files by setting DOTENV_PRIVATE_KEY and DOTENV_PRIVATE_KEY_PRODUCTION before dotenvx run.

$ touch .env
$ touch .env.production
$ dotenvx set HELLO encrypted
$ dotenvx set HELLO "production encrypted" -f .env.production
$ echo "console.log('Hello ' + process.env.HELLO)" > index.js
check .env.keys for your privateKeys
$ DOTENV_PRIVATE_KEY="122...0b8" DOTENV_PRIVATE_KEY_PRODUCTION="122...0b8" dotenvx run -- node index.js
[[email protected]] injecting env (3) from .env, .env.production
Hello encrypted
$ DOTENV_PRIVATE_KEY_PRODUCTION="122...0b8" DOTENV_PRIVATE_KEY="122...0b8" dotenvx run -- node index.js
[[email protected]] injecting env (3) from .env.production, .env
Hello production encrypted

Compose any encrypted files you want this way. As long as a DOTENV_PRIVATE_KEY_${environment} is set, the values from .env.${environment} will be decrypted at runtime.

Set log level to verbose. (log levels)

$ echo "HELLO=production" > .env.production
$ echo "console.log('Hello ' + process.env.HELLO)" > index.js
$ dotenvx run -f .env.production --verbose -- node index.js
loading env from .env.production (/path/to/.env.production)
HELLO set
[[email protected]] injecting env (1) from .env.production
Hello production

Set log level to debug. (log levels)

$ echo "HELLO=production" > .env.production
$ echo "console.log('Hello ' + process.env.HELLO)" > index.js
$ dotenvx run -f .env.production --debug -- node index.js
process command [node index.js]
options: {"env":[],"envFile":[".env.production"]}
loading env from .env.production (/path/to/.env.production)
{"HELLO":"production"}
HELLO set
HELLO set to production
[[email protected]] injecting env (1) from .env.production
executing process command [node index.js]
expanding process command to [/opt/homebrew/bin/node index.js]
Hello production

Use --quiet to suppress all output (except errors). (log levels)

$ echo "HELLO=production" > .env.production
$ echo "console.log('Hello ' + process.env.HELLO)" > index.js
$ dotenvx run -f .env.production --quiet -- node index.js
Hello production

You can also set DOTENV_QUIET=true.

$ DOTENV_QUIET=true dotenvx run -f .env.production -- node index.js
Hello production

Set --log-level to whatever you wish. For example, to suppress warnings (risky), set log level to error:

$ echo "HELLO=production" > .env.production
$ echo "console.log('Hello ' + process.env.HELLO)" > index.js
$ dotenvx run -f .env.production --log-level=error -- node index.js
Hello production

Available log levels are error, warn, info, verbose, debug, silly (source)

When an Envspec is present, run and check validate the same requirements: required values, types, enums, ranges, and storage encryption. No validation flag is needed. run performs validation before starting your command.

# Envspec
env "DATABASE_URL", type: "url"
env "API_KEY"
env "SENTRY_DSN", optional: true

$ dotenvx run -- node index.js
[INVALID_ENV] DATABASE_URL is required; API_KEY is required

Envspec validation failures, including unencrypted values, warn by default; run --strict stops before launching the command. A strict true setting in Envspec, at the root or in an active file block, also makes its validation failures fatal. check always exits with failure when validation fails. Strictness changes enforcement, not which rules are checked. Invalid Envspec syntax always stops execution.

Envspec values default to redacted: true: run masks their values in child stdout/stderr and resolved-value debug output, while the child still receives the real values. Set redacted: false for values that may appear in output. This is independent of encrypted: true, which requires an encrypted source.

file ".env.development" do
  env "BASE_URL", type: "url", redacted: false
  env "TOKEN_SECRET", encrypted: true
end

spec generates declarations with encryption and redaction required by default. Encryption is required by default: root and file-level encrypted settings are not supported. Use encrypted: false on individual declarations to permit plaintext. Entirely plaintext inputs generate no encryption exceptions except for keys ending in _PLAIN, so the next encrypt encrypts the remaining values. _PLAIN keys get encrypted: false only when none of their assignments contain ciphertext; code-only references do not infer this exception. For files that already contain ciphertext, spec preserves plaintext values with per-key exceptions. Source files are never modified.

spec infers redacted: false for names starting with PUBLIC, VITE, NEXT_PUBLIC, or NUXT_PUBLIC, or containing PUBLIC anywhere (case-sensitive). This affects visibility only; public names still require encryption unless explicitly exempted. Other names, including those ending in _PLAIN, remain redacted.

If an Envspec already exists, spec leaves it untouched and prints ○ Envspec already exists [edit or run: spec --overwrite]. Use spec --overwrite to regenerate it from the selected inputs, replacing custom rules, comments, and blocks for files absent from this machine. Interactive overwrite runs show the full file-and-code checklist with the prompt Recreate Envspec from .env files and code. Noninteractive runs use .env.example and .env, or the file selected with -f. Input validation finishes before the old Envspec is replaced; cancelled selection leaves it untouched. Symlinked Envspecs cannot be overwritten.

Use dotenvx spec --stdout to print the generated Envspec without writing it, or dotenvx spec --stdout -f .env.production to select an input file. This works even when an Envspec already exists; --stdout leaves it untouched even with --overwrite. Prompts and progress stay on stderr so stdout contains only the generated content.

Root and file-level redacted settings are not supported. Use per-key redacted: false to permit output visibility. When selected file policies disagree about a variable, redaction wins. check shows individual issues or a compact success summary without displaying values.

Use exact file blocks to override rules for selected files:

env "HELLO"
env "PORT", type: "port"
env "STRIPE_SECRET_KEY", optional: true
file ".env.production" do
  env "STRIPE_SECRET_KEY", required: true
  env "PORT", min: 1024
end

Both dotenvx run -f .env.production -- node index.js and dotenvx check -f .env.production use the production rules. Block declarations inherit top-level options and override only the options they specify. A block's encrypted directive applies to all inherited and newly declared variables; a per-variable encrypted: option inside that block overrides it.

Paths in file blocks are relative to the Envspec. They match the selected paths exactly after path normalization (./.env.production matches .env.production); they are not basename matches or globs. Directory inputs and DOTENV_FILE use their resolved file paths.

When no file block matches, top-level rules apply. When one or more blocks match, each matching block's inherited rules must hold for the final resolved environment. Shell values, fallback files, and --overload cannot bypass them. A selected missing file still activates its block. Multiple matching blocks cannot cancel each other's restrictions; conflicting proxy domains are rejected. Blocks cannot be nested, and duplicate declarations within one scope or duplicate file blocks are errors.

Validate the final resolved environment against your Envspec, using the same loader and precedence as run:

$ dotenvx check -f .env.local -f .env
▣ valid (.env.local, .env)
$ dotenvx check -f .env.production
☠ TOKEN_SECRET not encrypted (.env.production:4)

Envspec file blocks define policy; they do not select files to load. By default, check uses the same .env selection as run, including configured paths and private-key-based filename inference. Use repeated -f flags or --convention to select layers. DOTENV_FILE and its aliases are also supported.

All selected sources are merged before validation. The first file wins by default, existing shell values take precedence, and --overload lets later sources override earlier values. Partial files can satisfy the schema together. Encryption requirements apply to the source of the winning value, not to every overridden assignment on disk. Each selected file block's inherited rules must hold for the final environment, just as with run.

Missing files explicitly selected with -f or DOTENV_FILE (and its aliases) fail the check. Missing optional convention layers and implicit default files

... (README truncated for length)