git clone https://github.com/whazeted/paraglide-messages-mcp && cp paraglide-messages-mcp/*.md ~/.claude/agents/1 items in this repository
Use when the user wants to manage Paraglide JS / inlang messages or translations: translate missing or existing messages, add or remove locales, inspect/search message keys, rename/delete messages, remove orphan messages, or review i18n state.
Subagents overview
# paraglide-messages-mcp
[](https://www.npmjs.com/package/paraglide-messages-mcp)
[](LICENSE)
[](package.json)
An MCP server that turns AI agents into a parallel translation team for
[Paraglide JS](https://inlang.com/m/gerre34r/library-inlang-paraglideJs) /
[inlang](https://inlang.com) `messages/{locale}.json` files.
The agent calling the tools **is** the translator. The server makes that
safe and fast:
- **Validated writes** — every translation is checked against the source
(placeholders, markup, plural variants) before anything lands on disk;
bad items are rejected individually, never the whole batch.
- **One agent per locale, in parallel** — the translate loop reads and
writes only the source and target locale's files, so subagents can
translate all locales concurrently without conflicts.
- **Fast and offline** — direct JSON file access with a stat-validated
cache; tool calls cost milliseconds regardless of project size, and no
network is ever needed.
- **Toolchain-invisible** — written files are byte-compatible with the
message-format plugin's own output, and external edits (editor, compiler,
git) are always picked up.
## Quick start
No installation — add this to your `.mcp.json` (Claude Code) or
`claude_desktop_config.json`:
```json
{
"mcpServers": {
"paraglide": {
"command": "npx",
"args": [
"-y",
"paraglide-messages-mcp",
"--project",
"./project.inlang",
"--translation-style",
"Concise product UI; informal address; keep brand terms untranslated."
]
}
}
}
```
`--project` is optional: the server finds `project.inlang` in the working
directory (or a single `*.inlang` directory up to one level deep) by itself.
`--translation-style` is optional but recommended: it gives agents the
linguistic brief to use for tone, formality, and terminology instead of
deriving style from existing translations.
Then ask your agent to translate — or use the `translate_project` prompt to
translate every locale at once with one subagent per locale.
## Tools
| Tool | Purpose |
| --- | --- |
| `project_info` | Locales, base locale, `totalKeys` across all locales, `translatableKeys` from non-empty base messages, per-locale translated/missing counts, extra non-source keys, and the startup translation style brief when configured. |
| `get_translation_batch` | Next batch of untranslated messages for a locale (default 50), with source text and required placeholders. Optionally autosaves the previous batch in the same call (pass `translations`), so the loop is one round-trip per batch and the final batch is saved by the call that reports `done`. |
| `get_retranslation_batch` | Cursor-paged batch over *already-translated* messages too — refresh stale entries after source/terminology changes. Same optional autosave as `get_translation_batch`. |
| `save_translations` | Validate and persist translations for one locale; per-item results — overwrites existing values. (The batch tools share this save core for their autosave.) |
| `list_message_keys` | Keys only, filterable by prefix and status, paginated. |
| `get_messages` | Full message content by keys or prefix. |
| `search_messages` | Find messages by text or key substring. |
| `delete_messages` / `rename_message` | Key management across all locales. |
| `remove_orphan_messages` | Delete target-locale keys that are absent from the source locale (base locale by default), optionally scoped by locale or prefix. |
| `add_locale` / `remove_locale` | Locale management in `settings.json`. |
## Prompts
| Prompt | Purpose |
| --- | --- |
| `translate_project` | Translate every locale: the main agent uses the startup translation style brief (or asks the user for one), then fans out one subagent per locale in parallel. |
| `translate_locale` | Translate one locale via the batch loop. |
| `translate_prefix` | Same, scoped to keys starting with a prefix. |
| `retranslate` | Redo existing translations (stale copy, changed terminology) — by key prefix, every target locale by default, one subagent per locale. |
| `review_locale` | Review existing translations against the base locale and fix problems. |
Read-only state is also exposed as MCP resources
(`paraglide://project/info`, `paraglide://locales/{locale}/missing`,
`paraglide://messages/{locale}/{key}`), so clients can pin it as context
without spending tool calls.
## Agent skill
`skill/paraglide-translation/` is an optional installable skill in the open
[Agent Skills](https://agentskills.io) format. The skill simply points the agent to the MCP server.
```sh
# Codex, via the skills CLI
npx skills add whazeted/paraglide-messages-mcp --skill paraglide-translation -a codex
# Or install from the skill folder URL directly
npx skills add https://github.com/whazeted/paraglide-messages-mcp/tree/main/skill/paraglide-translation -a codex
```
For other skills-compatible agents, replace `codex` with that agent's
`skills` CLI target name. You can list the skill before installing:
```sh
npx skills add whazeted/paraglide-messages-mcp --list
```
## Compatibility
Requires the standard Paraglide JS setup: the inlang **message format**
plugin with a single message file per locale (any `pathPattern` location).
Other inlang plugins (i18next, next-intl, ICU) and multi-file namespaces are
deliberately not supported — see [COMPATIBILITY.md](COMPATIBILITY.md) for
the exact criteria and reasoning.
## Performance
A full 10-locale translation run over a 5,000-message project costs ~3 s of
server time (M1 Max Mac Studio)— the pipeline is bounded by the agent's translation speed, not
the server.
## Documentation
- [DEVELOPMENT.md](DEVELOPMENT.md) — architecture, message format details,
validation rules, building, testing, benchmarking, releasing
- [COMPATIBILITY.md](COMPATIBILITY.md) — supported project setups
## License
[MIT](LICENSE)
What people ask about paraglide-messages-mcp
What is whazeted/paraglide-messages-mcp?
+
whazeted/paraglide-messages-mcp is subagents for the Claude AI ecosystem with 0 GitHub stars.
How do I install paraglide-messages-mcp?
+
You can install paraglide-messages-mcp by cloning the repository (https://github.com/whazeted/paraglide-messages-mcp) or following the README instructions on GitHub. ClaudeWave also provides quick install blocks on this page.
Is whazeted/paraglide-messages-mcp safe to use?
+
whazeted/paraglide-messages-mcp has not been audited yet by our security agent. Review the original repository on GitHub before using it in production.
Who maintains whazeted/paraglide-messages-mcp?
+
whazeted/paraglide-messages-mcp is maintained by whazeted. The last recorded GitHub activity is from yesterday, with 0 open issues.
Are there alternatives to paraglide-messages-mcp?
+
Yes. On ClaudeWave you can browse similar subagents at /categories/agents, sorted by popularity or recent activity.
Deploy paraglide-messages-mcp to your cloud
Ship this repo to production in minutes. Each platform spins up its own environment with editable env vars.
Maintain this repo? Add a badge to your README
Drop the badge into your GitHub README to show it's tracked on ClaudeWave. Each badge links back to this page and reflects the live Trust Score.
[](https://claudewave.com/repo/whazeted-paraglide-messages-mcp)<a href="https://claudewave.com/repo/whazeted-paraglide-messages-mcp"><img src="https://claudewave.com/api/badge/whazeted-paraglide-messages-mcp" alt="Featured on ClaudeWave: whazeted/paraglide-messages-mcp" width="320" height="64" /></a>More Subagents
The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.
The agent that grows with you
Java 面试 & 后端通用面试指南,覆盖计算机基础、数据库、分布式、高并发、系统设计与 AI 应用开发
Production-ready platform for agentic workflow development.
The agent engineering platform.
🤯 LobeHub is your Chief Agent Operator, organizing your agents into 7×24 operations by hiring, scheduling, and reporting on your entire AI team.