Padrone

Padrone CLI Framework

Padrone is a type-safe CLI framework for Node.js/Bun. It uses any schema library that implements the Standard Schema spec (Zod, Valibot, ArkType, etc.) for argument validation and provides an immutable builder API for defining programs, commands, interceptors, and extensions.

Installation

npm install padrone zod    # or any Standard Schema-compatible library instead of zod

Quick Start

import { createPadrone } from 'padrone';
import * as z from 'zod/v4';

const program = createPadrone('mycli')
  .configure({ version: '1.0.0', description: 'My CLI app' })
  .command('greet', (c) =>
    c
      .arguments(z.object({ name: z.string() }), { positional: ['name'] })
      .action((args) => `Hello, ${args.name}!`),
  )
  .command(['deploy', 'dp'], (c) =>
    c
      .arguments(z.object({
        env: z.enum(['staging', 'production']),
        dry: z.boolean().default(false),
      }))
      .action((args, { runtime }) => {
        runtime.output(`Deploying to ${args.env}...`);
        return { deployed: true };
      }),
  );

program.cli();

Core Concepts

• Immutable builder: Every method returns a new builder/program instance
• Standard Schema validation: Any schema library supporting standard-schema (Zod, Valibot, ArkType, etc.) defines positional args, named flags, defaults, coercion, and validation
• Two entry points: 'padrone' (core) and 'padrone/test' (testing utilities)
• Sync by default: Returns become async only when async schemas or interceptors are used

Builder API Summary

Program API Summary (after builder methods)

Arguments Meta

The second parameter to .arguments() configures positional args, interactive prompts, and field metadata:

.arguments(schema, {
  positional: ['source', '...files'],     // '...' prefix = variadic
  interactive: true,                       // or ['fieldName'] for specific fields
  autoAlias: true,                         // auto kebab-case aliases for camelCase (default: true)
  stdin: 'data',                           // infers text/lines from schema type; use zodAsyncStream() for streaming
  fields: {
    output: { flags: 'o', description: 'Output path', examples: ['./dist'] },
    verbose: { flags: 'v', hidden: true },
    dryRun: { alias: 'dry' },             // multi-char long alias (--dry)
    local: { negative: 'remote' },        // --remote sets local to false, disables --no-local
    old: { deprecated: 'Use --new instead', group: 'Legacy' },
  },
})

Interceptor System

Seven phases in onion/middleware pattern with next():

1. start — before pipeline (root only, not called by parse()/run())
2. parse — command routing (root only)
3. route — after command resolved, before validation (root + command chain)
4. validate — schema validation (root + command chain)
5. execute — handler execution (root + command chain)
6. error — error handling, two layers: command-level first, then root-level (return { error: undefined, result } to suppress)
7. shutdown — cleanup, always runs, two layers: command-level first, then root-level

All phase contexts include context (user-provided context), signal (AbortSignal for cancellation), caller (invocation method: 'cli', 'eval', 'run', etc.), and runtime.

import { defineInterceptor } from 'padrone';

const timer = defineInterceptor({ name: 'timer', order: -10 }, () => {
  let startTime: number;
  return {
    start: (ctx, next) => {
      startTime = Date.now();
      return next();
    },
    execute: (ctx, next) => {
      const result = next();
      console.log(`${ctx.command.path} took ${Date.now() - startTime}ms`);
      return result;
    },
  };
});
program.intercept(timer);

defineInterceptor() returns a factory — each execution gets fresh closure state. Supports .provides<T>() and .requires<T>() for typed context (type-level only).

Extension-First Architecture

Padrone's core is minimal — most features are implemented as extensions composed via .extend(). When you call createPadrone(), built-in extensions are automatically applied:

Each can be disabled: createPadrone('myapp', { builtins: { help: false } }).

Advanced opt-in extensions imported from 'padrone': padroneLogger(), padroneTiming(), padroneProgress(), padroneUpdateCheck(), padroneEnv(), padroneConfig(). Optional integrations live behind subpath imports to keep their dependencies out of the main bundle: padroneInk from 'padrone/ink', padroneMcp from 'padrone/mcp', padroneServe from 'padrone/serve', padroneTracing from 'padrone/tracing', padroneCompletion from 'padrone/completion', padroneMan from 'padrone/man'.

Testing

import { testCli } from 'padrone/test';

const result = await testCli(program).run('greet World');
// result: { command, args, result, issues, stdout, stderr, error }

// With mocks
await testCli(program)
  .env({ API_KEY: 'xxx' })
  .prompt({ name: 'myapp' })
  .run('deploy --env staging');

// REPL testing
const { results } = await testCli(program).repl(['greet Alice', 'greet Bob']);

Progress Indicators

Auto-managed spinners for long-running commands via padroneProgress() context-providing interceptor:

.command('deploy', (c) =>
  c
    .async()
    .extend(padroneProgress({
      message: {
        progress: 'Deploying...',
        success: (result) => `Deployed v${result.version}`,
        error: 'Deploy failed',
      },
      bar: true,
      time: true,
      eta: true,
    }))
    .action(async (_args, ctx) => {
      await deploy();
      ctx.context.progress.update(0.5);
      ctx.context.progress.update('Finalizing...');
      return { version: '2.0' };
    }),
)

• Auto-managed: padroneProgress() starts before execution, calls succeed/fail automatically
• Messages: message accepts a string (progress message) or { validation?, progress?, success?, error? }. Can also be provided from context via progressConfig.message — command-level fields take precedence
• Manual control: Use ctx.context.progress in action handlers — update(string | number | { message?, progress?, indeterminate?, time? }), succeed, fail, stop, pause, resume
• Typed context: padroneProgress() uses .provides<{ progress: PadroneProgress }>() — ctx.context.progress is fully typed
• Dynamic messages: success/error can be callbacks returning string | null | { message, indicator }
• Spinner config: spinner accepts preset name ('dots', 'line', etc.), true (always show), false (disable), or { frames, interval, show } object
• Progress bar: bar: true or bar: { width, filled, empty, animation, show } — renders percentage + bar. Indeterminate animations: 'bounce', 'slide', 'pulse'
• Elapsed time: time: true shows ⏱ M:SS counter. Can be toggled via update({ time: true/false })
• ETA: eta: true shows ETA M:SS based on progress rate. Requires numeric update() calls. Counts down between updates
• Custom renderer: renderer: (message, options?) => PadroneProgress to replace the built-in terminal renderer

Error Classes

• PadroneError — base (exitCode, suggestions, command, phase)
• RoutingError — unknown command
• ValidationError — schema failures (has .issues)
• ConfigError — config file problems
• ActionError — throw from action handlers with structured metadata

Additional Resources

• For the complete API reference with all type signatures, see api-reference.md
• For full working examples covering common patterns, see examples.md