接入你的智能体
發布者自己的指令。不是 PlugMCP 端點。
用我們的金鑰與日誌託管一份副本。
npx -y [email protected]Claude Code 沒有本機指令的安裝連結。複製下方指令並在終端機中執行。
claude mcp add glooit -- npx -y glooit@0.6.7在終端機中執行。
{ "mcpServers": { "glooit": { "command": "npx", "args": [ "-y", "[email protected]" ] } }}claude_desktop_config.json
遠端 URL 還是本機指令？ 兩種安裝方式的差異
來自發布者
Sync AI coding assistant configurations across Claude Code, Cursor, Codex, OpenCode, and Roo Code/Cline.
🧴 Sync your AI agent configurations and rules across platforms with ease
Teams today use different AI coding assistants - some prefer Claude Code, others use Cursor, OpenCode, and CI/CD might run on Codex. Each tool has its own config format and location:
CLAUDE.md, commands in .claude/commands/, skills in .claude/skills/.cursor/rules/*.mdc with frontmatter, commands in .cursor/commands/AGENTS.md, commands in .opencode/command/, agents in .opencode/agent/AGENTS.md.roo/rules/glooit lets you write your rules once and sync them everywhere. Perfect for:
# Homebrew (macOS/Linux)
brew tap nikuscs/glooit https://github.com/nikuscs/glooit
brew install glooit
# npm
npm install -g glooit
# bun
bun install -g glooit
# pnpm
pnpm install -g glooit
glooit init # Create config
glooit sync # Sync rules to all agents
Create .agents/main.md with your rules, then configure glooit.config.ts:
Legacy support:
.glooitis still supported. IfconfigDiris not set, glooit will use.agentswhen present, otherwise it falls back to.glooit.
import { defineRules } from 'glooit';
export default defineRules({
rules: [
{
file: '.agents/main.md',
to: './',
targets: ['claude', 'cursor', 'codex']
}
]
});
Run glooit sync (or bunx glooit sync / npx glooit sync) and it creates:
CLAUDE.md for Claude Code.cursor/rules/main.mdc for CursorAGENTS.md for Codex*Codex uses .codex/prompts for commands
**OpenCode uses Claude-compatible skills path (.claude/skills/)
***Factory uses "droids" (.factory/droids/) for agents
Override output locations per-agent:
{
file: '.agents/backend.md',
to: './',
targets: [
'claude',
{ name: 'cursor', to: './backend/.cursor/rules/api.mdc' }
]
}
Sync commands, skills, and agents directories at the top level:
export default defineRules({
rules: [...],
// Simple string path (defaults to claude, cursor, opencode)
commands: '.agents/commands',
// Or with explicit targets
skills: {
path: '.agents/skills',
targets: ['claude']
},
agents: {
path: '.agents/agents',
targets: ['claude', 'cursor']
}
});
Output mappings:
commands → .claude/commands, .cursor/commands, .opencode/command, .codex/promptsskills → .claude/skills, .cursor/skills, .codex/skills, .factory/skills (OpenCode uses .claude/skills)agents → .claude/agents, .cursor/agents, .opencode/agent, .factory/droidsUse symlinks instead of copying. This keeps your source of truth in .agents and updates targets instantly.
export default defineRules({
mode: 'symlink', // or 'copy'
rules: [
{
file: '.agents/main.md',
to: './',
targets: ['claude', 'cursor']
},
{
file: '.agents/codex.md',
to: './',
mode: 'copy', // per-rule override
targets: ['codex']
}
]
});
Why use symlink mode?
glooit sync)Limitations: Symlinks point directly to source files, so transformations cannot be applied:
file: string[]) automatically fall back to copy mode (cannot symlink to multiple files)When limitations apply, glooit shows a warning and provides guidance. Use glooit unlink to convert symlinks back to regular files.
Combine multiple files into one output:
{
file: [
'.agents/coding-standards.md',
'.agents/testing-guidelines.md'
],
to: './',
targets: [
{ name: 'claude', to: './GUIDELINES.md' }
]
}
Limit rule scope to specific files:
{
file: '.agents/frontend.md',
to: './apps/frontend',
globs: 'src/**/*.{ts,tsx}',
targets: ['cursor']
}
Transform content during sync using placeholders in your source files.
Source file (.agents/main.md):
# Project Guidelines
Last updated: __TIMESTAMP__
## Project Structure
__STRUCTURE__
## Environment
- Node version: __NODE_VERSION__
- API URL: __API_URL__
Config (glooit.config.ts):
import { defineRules, transforms } from 'glooit';
export default defineRules({
rules: [
{
file: '.agents/main.md',
to: './',
targets: ['claude'],
hooks: ['replaceStructure', 'addTimestamp', 'replaceEnv']
}
],
transforms: {
after: [transforms.compact({ removeFillerWords: true })]
}
});
Built-in transforms:
Usage:
hooks: ['transformName'] to individual rulestransforms.after array to run on all rulesConfigure lifecycle hooks that run inside Claude Code and Cursor when the AI uses tools.
export default defineRules({
rules: [...],
hooks: [
// Run prettier after file edits
{
event: 'afterFileEdit',
command: 'npx prettier --write',
targets: ['claude', 'cursor']
},
// Run a TypeScript script before shell commands
{
event: 'beforeShellExecution',
script: '.agents/hooks/check-command.ts',
targets: ['claude']
},
// Block writes to sensitive files
{
event: 'PreToolUse',
script: '.agents/hooks/block-env-writes.ts',
matcher: 'Edit|Write', // Claude Code matcher
targets: ['claude']
}
]
});
Supported events:
Script types:
.ts files: Run with bun run.js files: Run with node.sh files: Run directlyMerge shared settings into provider-native settings files:
export default defineRules({
rules: [...],
settings: {
targets: ['claude', 'cursor', 'codex', 'opencode'],
env: ['GEMINI_API_KEY', 'OPENAI_API_KEY'],
envFiles: ['.env.agents', '.env.local', '.env'], // optional (default order shown)
permissions: {
allow: ['Read', 'Grep']
},
merge: true // optional; defaults to true
}
});
Resolution and precedence rules:
merge defaults to true (set merge: false to overwrite instead of merge).envFiles is optional; default lookup is ['.env.agents', '.env.local', '.env'].settings.env, the first matching file in envFiles order wins..env.agents is checked first, giving you a dedicated file for agent-specific secrets.process.env is used as a fallback (useful for CI and shell-exported vars).claude/codex), sync fails with a warning to prevent committing secrets.Default settings output files:
claude → .claude/settings.local.jsoncursor → .cursor/cli.jsoncodex → .codex/config.tomlopencode → opencode.jsonConfigure Model Context Protocol servers:
export default defineRules({
rules: [...],
mcps: [
{
name: 'postgres',
config: {
command: 'uvx',
args: ['mcp-server-postgres'],
env: { DATABASE_URL: process.env.DATABASE_URL }
},
targets: ['claude', 'cursor']
}
]
});
Automatic backups before sync:
export default defineRules({
rules: [...],
backup: {
enabled: true,
retention: 10
}
});
glooit init # Initialize configuration
glooit sync # Sync rules and MCPs
glooit link # Zero-config symlink sync (no config file needed)
glooit unlink # Replace symlinked outputs with real files
glooit validate # Validate configuration
glooit clean # Clean .gitignore entries
glooit reset --force # Remove all generated files
glooit upgrade # Upgrade glooit to latest version
glooit backup list # List available backups
glooit backup restore <timestamp> # Restore from backup
The link command provides zero-config symlink-based syncing without requiring a config file:
glooit link # Auto-detect .agents/ or .glooit/ and sync to all agents
glooit link .my-rules # Use custom source directory
glooit link -t claude,cursor # Sync to specific agents only
It automatically discovers and symlinks:
CLAUDE.md → Claude CodeAGENTS.md → Codex, OpenCode, Factorycommands/ → All agents that support commandsskills/ → All agents that support skillsagents/ → All agents that support agents/droidsThe upgrade command auto-detects your package manager and whether glooit is installed locally or globally:
glooit upgrade # Auto-detect and upgrade
glooit upgrade -g # Force global upgrade
glooit upgrade -l # Force local upgrade
All commands work with npx, bunx, or pnpx:
npx glooit sync
bunx glooit sync
pnpx glooit sync
See examples/full.config.ts for a complete configuration with all features.
# Install dependencies
bun install
# Run tests
bun run test
# Type check and lint
bun run check
# Build
bun run build
# Build local binary
bun run install:local
bun run release
This runs checks, prompts for version bump, creates a git tag, and pushes. CI handles npm publish and GitHub release automatically.
Without these amazing projects, this project would not be possible.
MIT
{ "mcpServers": { "glooit": { "command": "npx", "args": [ "-y", "[email protected]" ] } }}工具
尚無工具清單。遠端伺服器會在首次健康檢查後取得工具清單。
常見問題
在本頁複製適用於 Cursor、Claude Desktop、Claude Code 或 VS Code 的 stdio 設定。它使用發布者的安裝方式，而不是 PlugMCP 代理伺服器。
這筆記錄未列出必要的環境變數。
Glooit 透過 stdio 在本機執行。
上次檢查時間：2026-10-28。 🧴 Sync your AI agent configurations and rules across platforms with ease