i18n-mcp
An MCP server for managing Rails-style localization files (YAML/JSON), enabling coding agents to read, search, write, and audit translations without hand-editing files.
README
i18n-mcp
An MCP (Model Context Protocol) server that gives coding agents tools to manage
Rails-style localization files (YAML or JSON). Point it at a Rails app's
config/locales directory and the agent can read, search, write and audit
translations without hand-editing files.
Supported layouts
Both common Rails layouts are detected automatically, and can be mixed:
config/locales/
en/ # per-locale directories (nesting allowed)
common.yml # rooted at the locale code: `en: ...`
admin/users.yml
th/common.yml
legacy.yml # flat files whose top-level keys are locale codes
Files may be .yml, .yaml or .json. Writes go through the YAML AST, so
existing comments and key order survive edits. Duplicate keys are tolerated
with last-wins semantics, matching Rails (Psych). Files that cannot be parsed
are reported in locales output under problems instead of aborting
the server.
Install and build
Run directly from npm without installing:
npx -y mcp-i18n --locales-dir /path/to/rails-app/config/locales
Or build from source:
npm install
npm run build
Usage with an MCP client
Add the server to your MCP client configuration (Kimi Code, Claude Desktop, etc.). Via npx (no local checkout needed):
{
"mcpServers": {
"i18n": {
"command": "npx",
"args": [
"-y", "mcp-i18n",
"--locales-dir", "/path/to/rails-app/config/locales"
]
}
}
}
Or from a local build:
{
"mcpServers": {
"i18n": {
"command": "node",
"args": [
"/path/to/i18n-mcp/dist/src/index.js",
"--locales-dir", "/path/to/rails-app/config/locales"
]
}
}
}
Options:
--locales-dir <path>— locales directory. Falls back to theI18N_LOCALES_DIRenvironment variable, then<cwd>/config/locales.--default-file <name>— file name (relative to the locale directory) for brand-new keys. Default:common.yml(orcommon.jsonfor JSON-only locales).
Tools
| Tool | Description |
|---|---|
locales |
List locales with file/key counts and parse problems. |
keys |
List translation key paths; filter by locale and/or prefix. |
get_key |
Get a key's values across locales, with the file each lives in. |
search |
Substring search over translation keys (optionally values). |
set_key |
Create or update a key per locale. Writes to the key's current file, the nearest prefix's file, or the locale default. Warns on %{placeholder} mismatch. |
delete_key |
Delete a key from locales (default: all); prunes empty parents. |
missing |
Audit keys: missing per locale, empty values, %{placeholder} mismatches vs base locale (default en). |
Examples
// get_key
{ "key": "roles_scope.super_admin" }
// set_key
{ "key": "features.reports.title", "values": { "en": "Reports", "th": "Reports (th)" } }
// missing, scoped to one subtree
{ "prefix": "dashboards", "limit": 50 }
All responses are JSON. Errors (unknown locale, missing key, write failures) are returned as tool errors with a human-readable message.
Development
npm test # vitest unit/handler tests
npm run coverage # tests + v8 coverage report (enforced thresholds: 85% lines/statements, 80% branches/functions)
npm run lint # eslint (zero errors required)
npm run typecheck # tsc --noEmit, strict (zero errors required)
npm run build # compile to dist/
npm run smoke # end-to-end test over real stdio transport
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。