循环代理工具 是 AI Skill Hub 本期精选MCP工具之一。综合评分 7.5 分,整体质量较高。我们推荐使用将其纳入你的 AI 工具库,帮助提升工作效率。
循环代理工具 是一款遵循 MCP(Model Context Protocol)标准协议的 AI 工具扩展。通过 MCP 协议,它可以让 Claude、Cursor 等主流 AI 客户端直接访问和操作外部工具、数据源和服务,实现 AI 能力的无缝扩展。无论是文件操作、数据库查询还是 API 调用,都可以通过自然语言在 AI 对话中直接触发,极大提升生产效率。
循环代理工具 是一款遵循 MCP(Model Context Protocol)标准协议的 AI 工具扩展。通过 MCP 协议,它可以让 Claude、Cursor 等主流 AI 客户端直接访问和操作外部工具、数据源和服务,实现 AI 能力的无缝扩展。无论是文件操作、数据库查询还是 API 调用,都可以通过自然语言在 AI 对话中直接触发,极大提升生产效率。
# 方式一:通过 Claude Code CLI 一键安装
claude skill install https://github.com/hsaghir/looplet
# 方式二:手动配置 claude_desktop_config.json
{
"mcpServers": {
"------": {
"command": "npx",
"args": ["-y", "looplet"]
}
}
}
# 配置文件位置
# macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
# Windows: %APPDATA%/Claude/claude_desktop_config.json
# 安装后在 Claude 对话中直接使用 # 示例: 用户: 请帮我用 循环代理工具 执行以下任务... Claude: [自动调用 循环代理工具 MCP 工具处理请求] # 查看可用工具列表 # 在 Claude 中输入:"列出所有可用的 MCP 工具"
// claude_desktop_config.json 配置示例
{
"mcpServers": {
"______": {
"command": "npx",
"args": ["-y", "looplet"],
"env": {
// "API_KEY": "your-api-key-here"
}
}
}
}
// 保存后重启 Claude Desktop 生效

Describe an agent in one paragraph. Get a working agent in five minutes.
pip install looplet
export OPENAI_BASE_URL=https://api.openai.com/v1 # any OpenAI-compatible endpoint
export OPENAI_API_KEY=sk-...
export OPENAI_MODEL=gpt-5.5
looplet new "An agent that takes a URL and returns the page title and a 2-sentence summary"
looplet run-workspace ./agent.cartridge "Summarize https://example.com"
The recording above is a deterministic --pretty trace of that same CLI flow: build an agent cartridge, then run it against a real task. The real factory path uses the same commands; the recorded trace is scripted so the docs stay stable and tiny.
Mention an existing CLI, Python module, or script in your brief, and the factory wraps it. Your team's tools already exist; looplet introspects the real surface and writes thin wrappers around them — no hallucinated signatures.
```bash
Five fully-declarative cartridges ship in examples/:
| Workspace | What it does |
|---|---|
[hello.cartridge](examples/hello.cartridge/) | Two-tool starter; load and run with any backend |
[coder.cartridge](examples/coder.cartridge/) | Coding agent — bash, read, write, edit, grep, glob |
[dep_doctor.cartridge](examples/dep_doctor.cartridge/) | Audits a repo's dependency files for security/license/maintenance risk |
[git_detective.cartridge](examples/git_detective.cartridge/) | Investigates repo health from git history |
[threat_intel.cartridge](examples/threat_intel.cartridge/) | Local-first security briefings |
Four tools is usually enough.coder.cartridgeships withbash,read,write,edit— the same four that Pi used to rank #2 on TerminalBench.grepandglobare convenience wrappers overbash; you can drop them and the agent still works. Resist the urge to add a tool until the model demonstrably can't accomplish the task with the four it has.
Load any of them:
from looplet import cartridge_to_preset, composable_loop
preset = cartridge_to_preset("examples/dep_doctor.cartridge", runtime={"workspace": "/path/to/project"})
for step in composable_loop(llm=preset.llm, config=preset.config, tools=preset.tools, state=preset.state, hooks=preset.hooks, task={"goal": "Audit dependencies"}):
print(step.pretty())
Or use them as a starting point: cp -r examples/coder.cartridge ./my_agent.cartridge, then edit. Each cartridge round-trips losslessly with an AgentPreset via preset_to_cartridge / cartridge_to_preset.
---
looplet new "Wrap the gh CLI as a triage agent that surfaces my open PRs and issues"
高质量的开源MCP工具,实现LLM代理的工具调用循环
AI Skill Hub 为第三方内容聚合平台,本页面信息基于公开数据整理,不对工具功能和质量作任何法律背书。
建议在沙箱或测试环境中充分验证后,再部署至生产环境,并做好必要的安全评估。
✅ Apache 2.0 — 宽松开源协议,可商用,需保留版权声明和 NOTICE 文件,含专利授权条款。
经综合评估,循环代理工具 在MCP工具赛道中表现稳健,质量良好。如果你已有明确的使用需求,可以直接上手体验;如果还在评估阶段,建议对比同类工具后再做决策。
| 原始名称 | looplet |
| 原始描述 | 开源MCP工具:The tool-calling loop for LLM agents; iterator-first, protocol-hooked, one depen。⭐6 · Python |
| Topics | mcpagent-frameworkai-agentsasyncio |
| GitHub | https://github.com/hsaghir/looplet |
| License | Apache-2.0 |
| 语言 | Python |
收录时间:2026-06-04 · 更新时间:2026-06-04 · License:Apache-2.0 · AI Skill Hub 不对第三方内容的准确性作法律背书。
选择 Agent 类型,复制安装指令后粘贴到对应客户端