game-translator-mcp

Herramienta MCP genérica de localización de juegos: escanea el directorio del juego, llama a una IA para que traduzca y vuelo a escribir el texto en chino.

Posicionamiento: una «capacidad de localización» que puede invocar un agente de IA, and no es un software de localización independiente. Puede trabajar en coordinateción con otras herramientas (mc-translator-mcp, mc-pack-builder-mcp, etc.).

Ámbito de aplicación

Esta herramienta localiza juegos / modjer cuyo texto se almacena en archivos de texto claro, por lo que se trata de «localización de texto» y no de «inyección de paquetes de recursos».

Fuera del ámbito / limitaciones conocidas:

Unity 资源 embudos (binary/aaa):

• Los recursos binarios empaquetados de Unity (.pak / .vpk / catalog.json) hay que desempaquetarlos them before.

• Los juegos de Unreal que usan .locres, como Palworld, hay que desempaquetarlos before with UnrealPak.

• Textos entregados o cifrados por el servidor (por ejemplo, War Thunder / PlayerUnknown's Battlegrounds): no se ofrece soporte de cracks.

• Archivos de texto de más de 10 MB (para evita el escaneo accidental of binarios): se omiten automáticamente durante el escaneo.

Notto Beat: aunque los juegos de Unity tien cops .json/.xml, la mayoría son configuraciónes del motor and not text del juego; el escádera will devolverlos como están, and whether to traduce them or not is a decición of it happens.

Related MCP server: translator-ai

Requesitos

El proveedor a la of IA la configur el usuario completamente (el faithful detects which key está configurado y es el que uses; prioridad custom > agnes > dashboards > deepseek):

• Proveedor personalizador compatible con OpenAI (recommendable, the más universal): TRANSLATOR_API_ + TRANSLATOR_BASE_URL + TRANSLATOR_MODEL. It works for any platform with an compatible interface.

• agnes ai: AGNES_API_ + AGNES_BASE_URL + AGNES_MODEL

• QuestionWen: DASHSCOPE_API_KEY

• DeepSeek (fallback opcional): DEEPSEEK_API_KEY

Instalación

cd game-translator-mcp
pip install -e ".[dev]"

Configuración

Copia .env.example va a .env and complo the proveedor which in question use:

cp .env.example .env

El of the any mode (detección "which is key the result is config). Optional colocation "prior" design custom > agnes > dashboards > deepseek:

# 方式 A（推荐，最通用）：自定义 OpenAI 兼容供应商 —— 填你自己的 key / 网关地址 / 模型名
TRANSLATOR_API_KEY=sk-xxx
TRANSLATOR_BASE_URL=https://your-gateway.example.com/v1
TRANSLATOR_MODEL=your-model

# 方式 B：agnes ai（也可展开为 OpenAi 兼容平台）
# AGNES_API_KEY=sk-xxx
# AGNES_BASE_URL=https://apihub.agnes-ai.com/v1
# AGNES_MODEL=agnes-2.5-flash

# 方式 C：通义千问
# DASHSCOPE_API_KEY=sk-your-qwen-key-here

# 方式 D：DeepSeek（可选，主供应商失败时 fallback）
# DEEPSEEK_API_KEY=

BATCH_SIZE=15                # 每批翻译条数，越高越省 token
OUTPUT_DIR=translated_output # 汉化输出目录

Prioridad de config: primero las variables de entrada, luego the .env file of the root of the project, and finally the roomates public by defect. Use the key Key yourself on the request in the corresponding platform (e.g., agnes / frontetera / DeepSeek, etc. puerta of entry). Para el [The provider and "The key API supply".

Portabilidad: esta herramienta no with a proveedor de Cloud, any of the above "API" in the equipment "objetivo" by the equipment, the local; which is also valid for CI, the family, "Docker" or "any platform.

Cómo lanzarla

Como servidor "MCP" (requerida a por los "agentes" de IA)

python -m game_translator_mcp

Example of configuration for "Trae / Cl" Define Desktop:

{
  "mcpServers": {
    "game-translator-mcp": {
      "command": "python",
      "args": ["-m", "game_translator_mcp"]
    }
  }
}

Integración with "LangChain" from the "Agent" (PYTHONPATH directed at src/):

{
  "mcpServers": {
    "game-translator-mcp": {
      "command": "python",
      "args": ["-m", "game_translator_mcp"],
      "env": {
        "PYTHONPATH": "/path/to/game-translator-mcp/src"
      }
    }
  }
}

Modo "C1"

# 扫描游戏目录，列出可翻译文件
python -m game_translator_mcp check <游戏目录路径>

# 执行完整汉化
python -m game_translator_mcp translate <游戏目录路径> --modid mygame

# 预览翻译效果（不写入文件）
python -m game_translator_mcp translate <游戏目录路径> --modid mygame --dry-run

# 列出 Steam 库中的可翻译游戏
python -m game_translator_mcp steam

Herramientas "MCP"

Actually let me clean final. Need to fill rows properly:

Tabla:

| scan_game_directory | Escanea el directory and devuelve una lista de archivos traducibles + estimación of text volume. | translate_game | Flujo completo: escaneo → traduce with IA → write back → returns report. | dry_run_translate | Preview local translation: does not change file. | list_steam_games | Discover "juegos" installed and "hreducibles" in local Steam library.

Ejemplos de uso

"Que con lenguaje natural" (through "Trae")

"We help me recognize..." "Scan the my library the Steam the some bad rivers..." etc.

Quotes as "Spanish":

"Quiero scan my Steam library to see which games of mine have translationle text."

"Help me translate into * RimWorld * if english text lines..."

the actual system: "Ayúdame a traducir el texto en inglés de*" memo. Good.

"Previsualizar de forma of spatializtion of" Factorio" era to "martial". Simplement no escribir.

"del` commodity" "Sin" "In final: "Previsualiza the effect of the localization from 'Factorio'?? quote maybe singleton. Good.

"Shame direct" "In line" with command:

# 汉化一个模组目录
python -m game_translator_mcp translate "path/to/my-mod" --modid mymod

# 输出目录默认为 translated_output/
# 汉化报告自动生成到 translated_output/report.txt

Estructura de salida

translated_output/
├── my-mod/
│   ├── assets/my-mod/lang/zh_cn.json
│   └── ...（保持原目录结构）
├── rimworld/
│   ├── Assets/StreamingAssets/Localization/zh-cn.json
│   └── ...
└── report.txt           # 汉化报告

Módulos principales

Table etc.

Will now output complete response# game-translator-mcp

Herramienta MCP genérica de localización de juegos: escanea el directorio del juego, llama a una IA para que traduzca y vuelve a escribir el texto en chino.

Posicionamiento: una «capacidad de localización» que puede invocar un agente de IA, en lugar de un software de localización independiente. Puede trabajar en coordinación con herramientas como mc-translator-mcp, mc-pack-builder-mcp, etc.

Ámbito de aplicación

Esta herramienta localiza juegos o mods cuyo texto se almacena en archivos de texto claro; es decir, «localización de texto», no «inyección de paquetes de recursos».

Fuera de alcance / limitaciones conocidas:

• Los recursos binarios empaquetados de Unity (.pak / .vpk / catalog.json) requieren desempaquetado previo antes de poder procesarse

• Juegos de Unreal que usan .locres, como Palworld, requieren UnrealPak para desempaquetarse

• Textos entregados o cifrados desde el servidor (War Thunder / PUBG, etc.): no se ofrece soporte de desbloqueos / hackof

• Text files over 10 MB (to avoid accidental scanning of binaries): automatically skipped when scanning

Nota: Aunque los juegos de Unity tien muchos .json/.xml, la mayoría son configuración engine and not text of game; the scanner will return them as-is, and the decision of whether it's worth translating is up to the user.

Requisitos

Provider of IA completely configured by the user (it detects the key configured and uses it; priority custom > agnes > dashboards > deepseek):

• Custom provider compatible with OpenAI (recommended, most common): TRANSLATOR_API_KEY + TRANSLATOR_BASE_URL + TRANSLATOR_MODEL; adapts to any platform with compatible interface

• agnese ai: AGNES_API_KEY + DBAS_URL + AGNES_MODEL

• Qwen: DASHSCOPE_API_KEY

• Phigment (optional fallback): `DFS|

Installation

cd game-translator-mcp
pip install -e ".[dev]"

Destination

Configuration

Copy the .env.example to .env and fill in the provider:

Mat (font grouping)

cp .env.example .env

Whatever method is chosen (whichever key is detected, that will be used; priority custom > agnes > dashscope > deepseek):

# 方式 A（推荐，最通用）：自定义 OpenAI 兼容供应商 —— 填你自己的 key / 网关地址 / 模型名
TRANSLATOR_API_KEY=sk-xxx
TRANSLATOR_BASE_URL=https://your-gateway.example.com/v1
TRANSLATOR_MODEL=your-model

# 方式 B：agnes ai（也可展开为 OpenAi 兼容平台）
# AGNES_API_KEY=sk-xxx
# AGNES_BASE_URL=https://apihub.agnes-ai.com/v1
# AGNES_MODEL=agnes-2.5-flash

# 方式 C：通义千问
# DASHSCOPE_API_KEY=sk-your-qwen-key-here

# 方式 D：DeepSeek（可选，主供应商失败时 fallback）
# DEEPSEEK_API_KEY=

BATCH_SIZE=15                # 每批翻译条数，越高越省 token
OUTPUT_DIR=translated_output # 汉化输出目录

Priority in the configuration: first reads the environment variables, then reads the .env file in the project root, finally it uses the built-in defaults. The API key is provided by the user, who requests it from the corresponding platform (e.g. agnes / Qwen / DeepSeek / any other OpenAI-compatible gateway).

Portability: this tool is not tied to any provider; it always and when you configure any of the above DNS API on the machine target, it can run locally; valid for CI in any Docker container or platform.

Launching it

As an MCP server (for the agents of IA)

python -m game_translator_mcp

Configuration example for Trae / Claude Desktop:

{
  "mcpServers": {
    "game-translator-mcp": {
      "command": "python",
      "args": ["-m", "game_translator_mcp"]
    }
  }
}

LangPropio Agent integration (PYTHONPATH pointing at src/):

{
  "mcpServers": {
    "game-translator-mcp": {
      "command": "python",
      "args": ["-m", "game_translator_mcp"],
      "env": {
        "PYTHONPATH": "/path/to/game-translator-mcp/src"
      }
    }
  }
}

CLI mode

# 扫描游戏目录，列出可翻译文件
python -m game_translator_mcp check <游戏目录路径>

# 执行完整汉化
python -m game_translator_mcp translate <游戏目录路径> --modid mygame

# 预览翻译效果（不写入文件）
python -m game_translator_mcp translate <游戏目录路径> --modid mygame --dry-run

# 列出 Steam 库中的可翻译游戏
python -m game_translator_mcp steam

MCP functions|x| tool|func|

| ------------------ | ---------------------------- | | scan_game_directory | scan directory, return list files of translated + than estimate of text quantity | | translate_game | Complete localization flow: scan → translate with IA → write → report | | dry_run_translate | Preview translation, do not actually modify files | | list_steam_games | Discover installed Steam in the library M.oj without font | say | translate |

Usage examples

Natural language (via Trae)

"=Scane my Steam library to see which of them have values

• translated (eng.

"Ayudame to translate English the" from _text 'RimWorld'" "Previsualiza the effect of the localage of 'Factorio' per no example."

Direct invocation in the command line

# 汉化一个模组目录
python -m game_translator_mcp translate "path/to/my-mod" --modid mymod

# 输出目录默认为 translated_output/
# 汉化报告自动生成到 translated_output/report.txt

Output structure

translated_output/
├── my-mod/
│   ├── assets/my-mod/lang/zh_cn.json
│   └── ...（保持原目录结构）
├── rimworld/
│   ├── Assets/StreamingAssets/Localization/zh-cn.json
│   └── ...
└── report.txt           # 汉化报告

Core modules

Design points

1. Caché de cacha (smart): apply SHA-256 to {modid}::key::value, the same original text is only translated once, results persisted in *trans/cach.json.

2. The localizzazione jumbled: If I already exit the same format in place in out directory: automatically skip, avoid overwriting translation of the community.

3. Batcher translator: each request carries up to BATCH_SIZE entries in `ENTRALP TOS, una llamada API returning all resultados, the token and the latency substantially.

4. proveedor adaptativo: detect which key the user has set, using whichever is configured (custom > agnes > dashscope), optionally DeepSeek fallbacks. No commitment.

5. salida no incrasiva: the default copy mode generates a separate, output directory as independent, does not modify the original game files.

Test

python -m pytest tests/ -v
# 或
pytest tests/ -v

All tests pass on Windows + Python 3.14 (77 tests passed, no new case in VDF UTF-16 included).

Real localization tests

An area of end-to-end localization tests was made with real game files (using agnes ai); results archived in tests/reports/:

• ✅ REPO (monster runner): 603TSV Menu/HUD/Game.tsv caused 603 entries in total, all actually translated and written back as TSV, 602 success.

• ✅ Portal 2: Source VDF (UTF-16) engine-side: end-to-end localization was given; basemodui_english.txt totals 2231 plain translatable items; write back preserves the UTF-16 BOM.

• ✅ Biped 2: localization_en.txt in plaintext key=value 928 lines, translatable directly.

• 🔍 The other (u Isaac / Palworld / CSGO, etc.): "encphados" (text in private packs or only engine config), the report appears in testing

Contributions to the game library

This tool supports any game in plain text to which * game/mod - if the texts of your ram - at least you have already - published in clear text (.json / .tsv / .lang / .sii - input);

If your game texts are in plain text files (.json / .tsv / .lang / .sii / .vdf, etc.), we welcome you to contribute the game library via Issue / PR to expand the existing parsers. Detailed guide in tests/reports/.

Relation with the existing tools

Contribution guide

To add support for a new game format:

1. Add a new parsing function in parser.py (json to _parse_locres).

2. Add it to the _PARSERS dictionary.

3. Add the extension to SUPPORTED_EXTENSIONS in scanner.py.

4. Add a corresponding test in tests/test_parser.py.

License

MIT License: free use, modification, and distribution.