envtrue

Type-safe environment variables. Loads .env files, validates with Zod/Valibot/ArkType, auto-coerces types. Drop-in dotenv replacement.

The problem

• process.env is string | undefined, so you lose types immediately.
• Missing or malformed variables fail at runtime, often far away from startup.
• Client bundles can accidentally expose server secrets if env handling is ad hoc.

Install

npm install envtrue zod

pnpm add envtrue zod

yarn add envtrue zod

Quick start

import { createEnv } from "envtrue";
import { z } from "zod";

const env = createEnv({
  server: {
    DATABASE_URL: z.string().url(),
    PORT: z.number(),        // auto-coerced from "3000" → 3000
    API_KEY: z.string().min(1),
  },
  client: {
    API_BASE: z.string().url(),
  },
  clientPrefix: "NEXT_PUBLIC_"
});

const db: string = env.DATABASE_URL;  // typed as string
await connect(db, { port: env.PORT }); // typed as number
fetch(env.API_BASE);

Why not t3-env?

Feature	envtrue	t3-env
Loads .env files for you	✅	❌
Auto-coerces z.number(), z.boolean(), z.array()	✅	❌
No runtimeEnv boilerplate	✅	❌
Framework-agnostic core	✅	⚠️

envtrue is optimized for the common case: read .env, merge with runtime env, coerce strings into the right primitives, validate once, return typed values. No extra runtime mapping step.

Why not the DIY Zod pattern?

The usual Zod setup is a small pile of repeated glue:

• Load .env yourself
• Merge sources manually
• Remember to coerce strings before validation
• Split public and private variables by convention
• Build readable startup errors yourself

envtrue keeps the schema but removes the glue. It auto-loads .env and .env.local, auto-coerces string inputs, separates client variables by prefix, and throws one formatted error with every invalid variable listed at once.

Framework adapters

Adapter	Import	Notes
Next.js	import { createNextEnv } from "envtrue/nextjs"	Uses NEXT_PUBLIC_, skips server validation in browser bundles, skips validation during Next build phases when needed
Vite	import { createViteEnv } from "envtrue/vite"	Uses VITE_, works with import.meta.env or process.env
Hono	import { createHonoEnv } from "envtrue/hono"	Server-only, supports explicit env bindings such as Cloudflare Workers / c.env

API reference

createEnv(options)

createEnv({
  server,
  client,
  clientPrefix,
  envFiles,
  cwd,
  env,
  skipValidation,
  onError
})

Option	Type	Default	Description
server	SchemaShape	{}	Server-only environment schema
client	SchemaShape	{}	Client-safe environment schema
clientPrefix	string	"NEXT_PUBLIC_"	Prefix required for client variables
envFiles	string | string[]	[".env", ".env.local"]	.env files to load, in merge order
cwd	string	process.cwd()	Base directory used to resolve envFiles
env	Record<string, string | undefined>	process.env	Explicit runtime env source override
skipValidation	boolean	false	Skip schema validation and return coerced raw values
onError	(errors: EnvError[]) => void	throws formatted error	Hook for custom error handling

Validation flow

1. Load .env files from disk.
2. Merge loaded values with runtime env. Runtime env wins.
3. Coerce string values before validation.
4. Validate server and client schemas.
5. Throw one aggregated error if anything is invalid.
6. Return a frozen typed object.

Supported schemas

• Zod raw shapes
• Standard Schema compatible shapes (Valibot, ArkType)

ArkType is supported through Standard Schema compatibility, but is not currently covered by the test suite.

Roadmap

• Monorepo support
• Encrypted .env support with dotenvx compatibility
• VS Code extension for .env autocomplete and intellisense

License

MIT