来自的
- 作者仓库星标 0
- 作者仓库 skills-registry
- 领域
- 人工智能
- 作者声明的智能体
- 未找到明确声明;不据此推断已兼容或已测试
- 静态检查
- 88 / 100 · 启发式扫描,不代表运行安全
- 运行环境证据
- 未执行;当前没有隔离运行测试证据
- 作者 / 版本 / 许可证
- @tomevault-io · 未声明许可证
- 上下文用量评级
- 低消耗
- 接入复杂程度
- 需简单配置
- 是否需要外部接口密钥
- 不需要
- 兼容的系统
- macOS · Linux · Windows
- 运行环境要求
- Node.js · Python
- 检测到的文件与系统行为
-
- 只读
- 允许写入 / 修改
- 命令执行
- 网络行为
- 允许外网请求
- 安装命令数
- 无(仅作为资料)
需要注意: 未声明工具权限范围;实际权限由运行环境决定。
# /n8n-cli-from-mcp — MCP / Skill → n8n Workflow
Take an MCP server config or a Claude Code skill and generate an n8n workflow that does the same thing without needing an agent at runtime.
## Procedure
1. **Identify the source**:
- **MCP server**: ask for the server name, command, and tools it exposes (or read from `~/.config/claude/mcp.json` style config)
- **Claude Code skill**: ask for the skill name or path to its SKILL.md
2. **Read the source's actual capabilities**:
- For an MCP server: list the tools, their input schemas, what they return
- For a skill: read the procedure section, identify what commands/APIs it calls
3. **Map MCP tools / skill steps to n8n nodes**:
- HTTP API call → HTTP Request node
- Slack post → Slack node
- Notion read/write → Notion node
- File read/write → Read/Write Files node
- Code execution → Code node (JS or Python)
- LLM call → OpenAI / Anthropic node
- Branching logic → IF / Switch node
4. **Identify what the workflow's trigger should be**:
- If the original MCP/skill was invoked by an agent on demand → Manual Trigger or Webhook
- If it was meant to run on a schedule → Schedule Trigger
- If it was reactive to events → Webhook Trigger or specific event trigger node
5. **Build the workflow JSON**:
- Start with the trigger
- Chain the mapped nodes
- Add credentials where needed (use existing instance credentials, not new ones)
- Add error handling for any external API calls
6. **Show the proposed workflow** before importing:
- Print the node-by-node plan
- Highlight any places where the mapping is approximate ("MCP tool X has no direct n8n equivalent — using Code node with the same logic")
7. **Import** (only after user approval):
- `n8n-cli wf import /tmp/from-mcp.json` (don't activate)
- Walk the user through testing it manually first
- Then activate
## Output format
- Show the source's tools/steps as a numbered list
- Show the mapped n8n nodes as a parallel numbered list
- Highlight any mappings that need user judgment
- Confirm before importing
## Tips
- The most natural conversions are MCP servers that just call HTTP APIs or run scripts — those map 1:1 to HTTP Request or Code nodes.
- MCP servers with deep state (like memory MCPs) don't map well — n8n is stateless between runs unless you wire up a database. Surface this limitation up front.
- Some Claude Code skills are pure prompts (e.g. "summarize this") — those need an LLM node in the n8n workflow, which means a credential and per-run cost.
- For complex MCPs (browser automation, etc.), suggest keeping them as MCPs and having n8n call out to them via HTTP if exposed, rather than rebuilding them.
- Document the resulting n8n workflow with /n8n-cli-document so the team can maintain it without referring back to the original MCP.
---
> Source: [8Dvibes/n8n-cli](https://github.com/8Dvibes/n8n-cli) — distributed by [TomeVault](https://tomevault.io).
<!-- tomevault:4.0:skill_md:2026-05-23 -->
章节与文件索引
按说明文件中的顺序排列。
程序
输出格式
温馨提示
文件/命令 -> ~/.config/claude/mcp.json · n8n-cli wf import /tmp/from-mcp.json · .config/claude/mcp.json · commands/APIs · read/write · Read/Write · MCP/skill · tmp/from-mcp.json
内容校验值 -> c36e5ba797ce
当前原文的自动摘译,尚未人工审校。
程序
识别源: MCP 服务器:询问其公开的服务器名称、命令和工具(或从
~/.config/claude/mcp.json样式配置中读取) Claude 代码技能:询问技能名称或 SKILL.md 的路径 读取源的实际功能: 对于 MCP 服务器:列出工具、其输入模式、返回内容输出格式
将源的工具/步骤显示为编号列表 将映射的 n8n 节点显示为并行编号列表 突出显示任何需要用户判断的映射 导入前确认
温馨提示
最自然的转换是仅调用 HTTP API 或运行脚本的 MCP 服务器——它们 1:1 映射到 HTTP 请求或代码节点。具有深度状态的 MCP 服务器(如内存 MCP)不能很好地映射 — n8n 在运行之间是无状态的,除非您连接数据库。预先说明这一限制。一些 Claude Code 技能是纯粹的提示(例如“总结一下”)——这些技能需要 n8n 工作流程中的 LLM 节点,这意味着凭证和每次运行的成本。