What file formats does i18n Agent support?

i18n Agent supports 9+ file formats including JSON, YAML, CSV, XML, Markdown, Properties, and more. The service preserves structure and formatting automatically.

How much does i18n Agent cost?

i18n Agent costs €0.01 per word, which is significantly more affordable than traditional translation services that typically charge €0.15-0.25 per word.

What is MCP integration?

MCP (Model Context Protocol) integration allows i18n Agent to work directly with AI development tools like Claude Code, Cursor, and other AI-powered development environments for seamless translation workflows.

What is pseudo-translation and why should I use it?

Pseudo-translation replaces your source strings with modified versions that simulate properties of real translations — longer text, different character sets, reversed reading direction — without requiring actual translation. It catches layout bugs, hardcoded strings, encoding issues, and RTL problems early in development, before you pay for real translations.

Does it break my ICU placeholders and variables?

No. i18n-pseudo parses ICU message syntax and only transforms translatable text segments. Placeholders like {name}, interpolation variables, and plural expressions ({count, plural, one {# item} other {# items}}) are preserved exactly as-is.

Which preset should I start with?

Start with the default preset (accents + brackets) to find hardcoded strings. Then use the layout preset to catch text truncation and overflow. Use charset if you support CJK languages, and rtl if you support Arabic or Hebrew.

How do I integrate pseudo-translation into CI?

Generate pseudo-translated files with 'i18n-pseudo en.json --preset layout -o locales/pseudo/', then run your visual regression tests (Playwright, Chromatic, Percy) with the pseudo locale. Any layout that breaks with expanded text will show up as a visual diff.

Pseudo-Translation Tool

Name: i18n Agent
Price: 0.01 EUR
Rating: 4.8 (127 reviews)
Author: i18n Agent

Test your i18n UI before launch. Generate fake translations that stress-test layout, encoding, and bidirectional rendering — without waiting for real translations.

The Problem

Your UI looks perfect in English. Then German arrives — 40% longer strings break every button and truncate every table header. Arabic flips the entire layout and your icons point the wrong way. Japanese needs different line heights and your carefully spaced cards collapse. You find all of this after paying for translations.

The average cost of fixing i18n layout bugs after translation is 5-10x higher than catching them during development. Pseudo-translation catches these issues for free, before you spend a dollar on real translations.

The Solution

i18n-pseudo generates fake translations that stress-test your UI without waiting for real translators. It applies 7 transformation strategies — accents, CJK characters, text expansion, RTL markers, and more — while preserving ICU placeholders, plural syntax, and interpolation variables.

Installation

i18n-pseudo is available as an npm package, a Homebrew formula, or you can build from source with Cargo.

Terminal

npm install -g @i18n-agent/i18n-pseudo

The npm package includes a global CLI binary. No runtime dependencies required.

7 Transformation Strategies

Each strategy targets a specific class of i18n bug. Use them individually with flags or combine them with presets.

Strategy	Flag	Input	Output
Accents	`--accents`	Hello	Ĥéľľó
CJK	`--cjk`	Hello	你好世界你好
Special Chars	`--special`	Hello	Hello
Expansion	`--expansion N`	Save	Save~~~~
Brackets	`--brackets`	Hello	[Hello]
RTL	`--rtl`	Hello	‮Hello‬
Unicode Stress	`--unicode-stress`	Hello	H́ȩl̈l̀õ

Accents are the most popular strategy for detecting hardcoded strings. If a string appears without diacritics in your UI, it was never passed through the translation function.

Presets

Presets bundle strategies for common testing scenarios. Use --preset to apply a group of strategies at once.

Preset	Strategies	Use Case
`default`	accents + brackets	Quick visual check that all strings are externalized
`layout`	expansion (1.4x) + brackets	Catch truncation and overflow before real translations
`charset`	CJK + special + brackets	Test font rendering and character encoding
`rtl`	RTL markers + brackets	Verify bidirectional layout without switching locale
`full`	All strategies combined	Maximum stress test — everything at once

Usage Examples

Run i18n-pseudo against your source locale file. It reads JSON, YAML, PO, and other i18n formats, applies the selected strategies, and writes the transformed output.

Terminal

# Default (accents + brackets)
i18n-pseudo en.json -o output/

# Layout testing (text expansion + brackets)
i18n-pseudo en.json --preset layout -o output/

# Specific strategies
i18n-pseudo en.json --accents --expansion 1.5 -o output/

# Kitchen sink
i18n-pseudo en.json --preset full -o output/

Before & After

Here is what the default preset (accents + brackets) does to a typical translation file. Notice that ICU plural syntax and interpolation placeholders are preserved exactly.

Before & After

// Input (en.json):
{
  "greeting": "Hello, {name}!",
  "save": "Save",
  "items": "{count, plural, one {# item} other {# items}}"
}

// Output (default preset — accents + brackets):
{
  "greeting": "[Ĥéľľó, {name}!]",
  "save": "[Šáṽé]",
  "items": "[{count, plural, one {# íťéɱ} other {# íťéɱš}}]"
}

Placeholders like {name}, {count}, and ICU plural expressions are never modified. Only translatable text segments are transformed, so your app renders correctly with pseudo-translated files.

CI Integration

Add pseudo-translation to your CI pipeline to catch i18n regressions automatically. Generate pseudo-translated files, then run your visual regression tests against them.

CI Pipeline

# Generate pseudo-translations and run visual regression
i18n-pseudo locales/en.json --preset layout -o locales/pseudo/
NEXT_PUBLIC_LOCALE=pseudo npx playwright test --project=visual

Combine the layout preset with Playwright or Chromatic visual regression testing. Any component that overflows or truncates with 1.4x expanded text will fail the visual diff.