Workflows MCP

MCP开发工具

提供方 AgentDeskAI

开发工具 stdio · 本地进程 无需鉴权 社区 / 第三方
查看接入方式

orchestrate \u0026 combine prompts + MCP servers into compound MCP tools

一句话选型结论

按资料字段整理

Workflows MCP 属于开发工具类 MCP 服务器,stdio · 本地进程,无需鉴权,有可直接执行的安装命令。接入前先确认权限范围;涉及写入、删除或付款的操作,建议先在测试环境验证。

分类开发工具 传输方式stdio · 本地进程 鉴权无需鉴权 部署方式本地部署
GitHub 星标 未标注 仓库热度,无公开仓库时显示未标注
工具数 未标注 暴露给模型的工具数量
资源 / 提示词 0 / 0 资源与提示词模板数量
维护方 社区 / 第三方 核验等级 仅资料核对

工具与能力

开发工具
数量来自来源目录的标注,具体工具清单以官方文档为准
工具数量未标注暴露给模型可调用的工具个数
资源数量0可读取的资源条目
提示词数量0内置提示词模板数量

能力标签

适用场景

developer-tools

安装与配置

stdio · 本地进程 无需鉴权
客户端在本机拉起子进程,数据不出本机
npx -y mcpn@latest init。

这条命令只是入口,具体启动参数请按官方文档补齐,不要直接照搬到别的服务器。

  1. 1
    准备运行环境

    stdio 服务在本地起进程,需要先装好 Node.js,再按安装命令拉取依赖。

  2. 2
    安装服务

    复制上面的安装命令执行;命令里带占位符的位置,换成自己的密钥或路径。

  3. 3
    写入客户端配置

    在客户端的 MCP 配置里填好启动命令和参数,保存后完全退出客户端再打开,让新配置生效。

  4. 4
    验证是否可用

    连接成功后随便调一个只读工具,确认能返回结果,再接入正式流程。

典型使用场景

3 类
按分类与标签整理的常见用法,具体以官方文档为准
代码与仓库操作

在对话里直接查代码、提 issue、走 PR,省掉来回切工具。

本地命令与脚本

把构建、测试、格式化这类重复命令交给模型执行。

接口联调与测试

让模型调你的接口、跑一遍用例,快速定位问题。

选择建议

选之前先看这几条
按条目字段给出的判断,不替你决定用不用

值得选的地方

  • 有现成的安装命令,不用自己找入口,照着跑就能起服务。
  • 资料标注无需鉴权,装好就能连。

先想清楚的地方

  • stdio 方式要求本地装好运行时(如 Node.js),换机器要重新装一遍。
  • 资料没标注工具数量,实际能力要装好后确认。
  • 版本更新与行为细节以官方仓库为准。

更适合

  • 需要在开发工具场景里让模型直接调用外部能力的人。
  • 想先快速试一下、能接受在本地或自有环境跑服务的人。

不太适合

  • 完全不想在本地装环境的人。
  • 需要厂商级 SLA 和技术支持的生产环境。

同类对比

6 个同类
同分类里挑最接近的几条,字段以各自详情页为准
MCP 服务器工具数传输方式部署方式安装命令最后核验
Workflows MCP当前 未标注 stdio · 本地进程 本地部署 有 2026-05-18
Git 838e2ec1 12 stdio · 本地进程 本地部署 有 2026-05-17
Markitdown MCP 2c653a5f 1 stdio · 本地进程 本地部署 有 2026-05-17
JavaGuide 未标注 stdio · 本地进程 本地部署 有 2026-04-11
awesome-claude-skills 未标注 stdio · 本地进程 本地部署 有 2026-04-26
Context7 831acf40 未标注 stdio · 本地进程 本地部署 有 2026-05-17

服务器档案

开发工具
字段来自项目仓库与公开索引,中文说明由本站标注
提供方AgentDeskAI上游标注的维护方 AgentDeskAI
分类开发工具本站按上游分类展示
连接方式stdio · 本地进程客户端在本机拉起子进程,数据不出本机
鉴权方式无需鉴权远程服务建议用最小权限凭据
维护方社区 / 第三方核验等级 仅资料核对
包名 / 运行时未标注 · Node.js本地安装时用到的包标识与运行时
部署类型本地部署上游给出的部署形态
最近核验2026-05-18上游最后核验时间

详细介绍

14.2 KB
来自项目仓库与上游目录的原文,未做删改

mcpn.ai helps you create, organize, and share AI workflows and prompts.

Like a dynamic prompt library you can share & version across MCP tools and servers for specific tasks.

Created by

Usage

Basic Example```yaml
plan_prd:
description: "Create a product requirements..."
prompt: |
You are an expert product manager/business
analyst and are now entering "PRD Mode"
Begin by deeply reflecting upon the notes,
requirements and images provided...
tools: createPRD, createStoriesFromPRD, createRoadmap

coding_max:
description: "Highly effective Coding Agent"
prompt: |
You are now entering "Developer Mode" -
deeply reflect upon the changes being asked and
analyze existing code to map the full scope...
tools: getConsoleLogs, getErrorLogs, getLinter

...


Workflows MCP (mcpn) are the building blocks of software development, an orchestration of multiple MCP servers and prompts.

Think of it like a dynamic prompting library that you can easily share and version control through yaml files that also lets you define how to best use many MCP tools across many MCP servers for specific tasks.

## Installation

Install the MCP server into an MCP client using the following command or JSON:```bash
npx -y mcpn@latest init

If using JSON in Cursor to setup your MCP server, you can use the following config:```json
{
"mcpServers": {
"workflows-mcp": {
"command": "npx",
"args": ["-y", "mcpn@latest server"]
}
}
}


## Documentation

Documentation: https://mcpn.ai

## Examples

... TBD

## License

... TBD

## Links

... TBD

========================================================================

Move below to documentation

## Overview

The key to effective MCP use is to know when/how to use the right tools. Workflows MCP helps make this process MUCH easier and faster. With Workflows MCP, you can:

✅ Combine prompts with MCP servers into reusable & easy to edit/share/organize workflows

✅ Trigger workflows with custom commands like: "enter debugger mode" or "use thinking mode"

✅ Define custom strategies for how to use many tools in a workflow (sequential or situational/dynamic)

✅ Easily onboard your team to the best prompts + ways to use many MCP tools w/ version control.

_We also provide a number of useful presets to get you started 🫡_

### So why use this?

The goal of this project is to turn prompts into programmable tools and rules for AI in a code editor more deterministic while reducing the overall amount of token use across all requests. Since stuffing a global set of rules for AI can eat up your context window, using MCP to route to prompts and workflows can help reduce the overall amount of tokens used across all requests.

Workflows for MCP are also **incredibly** powerful! For instance, you can create an **"production incident workflow"** tool that has a special system prompt to reason about an incident in a particular way and then use a series of tools in a specific order to analyze + potentially resolve the incident. This may involve things like:

- Gathering logs across many MCP tools for Kubernetes, Cloudwatch, Splunk, etc
- Checking Github issues or a vector database for similar issues in the past
- Perform web searches for useful information on the web
- Analyze your code with all the gathered context from the prior steps
- Write a detailed incident report with all the context from the prior steps
- Gather input from you to better understand the problem space
- Implement a fix for issue if possible

And since these workflows are all defined in YAML and can be referenced from anywhere on your machine, you can **easily** rally your team around when/how to use the best prompts and sequences of MCP tools for specific tasks 🦾

## Features

### 📝 Custom Configuration

- Load custom configs from multiple YAML files in a `.workflows` or `.mcp-workflows` directory
- Easily save these yaml files to a git repo and share with your team
- Since MCP servers can have custom configs for each project, you can easily pick and choose which configs to use for each project - setting up custom workflow folders for each project
- Support for typed parameter inputs in tool configurations makes building custom MCP prompting tools a breeze

### 🚀 Ready-to-Use Presets

Workflows MCP includes multiple preset workflow modes out of the box for thinking, coding, and github use:

**Thinking**: General purpose tools to improve reasoning on any task

- **Thinking Mode**: Reflect on thoughts and produce structured analysis (inspired by Anthropic's latest research into thinking tools)
- **Deep Thinking Mode**: Comprehensive multi-perspective analysis with detailed reflections

**Coding**: Universally applicable tools for common coding tasks
Universally applicable tools for common coding tasks

- **Debugger Mode**: Systematic multi-step reasoning and debugging with hypothesis creation, telemetry gathering and testing
- **Architecture Mode**: System design prompt with tradeoff analysis and implementation planning
- **Planner Mode**: Systematic code change planning with codebase analysis
- **PRD Mode**: Structured product requirements documentation for features, user stories and spike analysis
- **Save Note**: Document ongoing work with comprehensive progress tracking. Useful for when you need to step away for a bit and want to save your thoughts for later.

**GitHub**: Tools to simplify common GitHub tasks. Great for anyone who is a beginner with source control using Git and the Github CLI

- **PR Review Mode**: Comprehensive pull request analysis with security considerations
- **PR Creation Mode**: Structured PR creation process using GitHub CLI
- **Create Branch**: Smart branch creation with contextual naming
- **Save Changes**: Systematic git commit and push workflow

## Installation

Install the MCP server into an MCP client using the following command or JSON:```bash
npx -y mcpn@latest init

If using JSON in Cursor to setup your MCP server, you can use the following config:```json
{
"mcpServers": {
"workflows-mcp": {
"command": "npx",
"args": ["-y", "mcpn@latest server"]
}
}
}


To provide custom configurations, you can use the `--config` flag to point to a directory containing YAML configuration files. The directory must be named `.workflows` or `.mcp-workflows` as so:```bash
npx mcpn@latest server --config /path/to/.workflows

If you want to enable presets, you can use the --preset flag to specify which presets to load:```bash
npx mcpn@latest server --preset thinking,coding,github


Here's what this would look like in a Cursor config all combined:```json
{
  "mcpServers": {
    "workflows-mcp": {
      "command": "npx",
      "args": [
        "mcpn@latest server",
        "--config",
        "/path/to/.workflows",
        "--preset",
        "thinking,coding"
      ]
    }
  }
}

Note:

  • If you update your config, you must refresh the MCP tool
  • If refreshing doesn't work, make sure your config is valid YAML
  • If you're still having issues, then try removing & renaming the MCP tool in your client
  • If no config or preset is provided, it will default to using the thinking preset.
  • If you're still not able to get this working, open an issue ticket

Custom Workflow Configs

Create a .workflows or .mcp-workflows directory in your project and add YAML configuration files with any name (must end with .yaml or .yml). These configurations will also override the preset defaults if named the same as a preset tool.

Example Configuration Files

Basic Workflow Configuration```yaml

workflow_name:
description: "Description of what this workflow does"
prompt: |
Enter your multi-line
prompt here like this
toolMode: "situational" # can be omitted as it's the default
tools: "analyzeLogs, generateReport, validateFindings"


#### Tools Configuration Styles

You can define tools in several ways.

Here's an example of a tool that debugs a problem in a codebase with tools used situationally / as needed:```yaml
web_debugger_mode:
  description: Debug my codebase my web application with browser logs and BrowserTools MCP
  prompt: |
    Deeply reflect upon all of this and think about why this isn't working. Theorize 4-6 different possible sources of the problem.

    Then, deeply reason about the root cause and distill your theories down to the 1-2 most probable sources of the problem before suggesting next steps.
  tools: getConsoleLogs, getConsoleErrors, getNetworkLogs, getNetworkErrors, takeScreenshot

This will return the following MCP response:```
Deeply reflect upon all of this and think about why this isn't working. Theorize 4-6 different possible sources of the problem.

Available Tools

Use these tools as needed to complete the user's request:

  • getConsoleLogs
  • getConsoleErrors
  • getNetworkLogs
  • getNetworkErrors
  • takeScreenshot

After using each tool, return a 'Next Steps' section with a list of the next steps to take / remaining tools to invoke along with each tool's prompt/description and 'optional' flag if present.


#### Parameter Injection

A powerful feature is the ability to inject parameters into your prompts using the `{{ parameter_name }}` syntax:```yaml
custom_mode:
  description: "Workflow with parameter injection"
  parameters:
    thought:
      type: "string"
      description: "A thought to deeply reflect upon"
      required: true
    idea:
      type: "string"
      description: "An additional idea to consider"
  prompt: |
    Deeply reflect upon the provided thought.
    Here's the thought: {{ thought }}

    Additional idea to consider: {{ idea }}

    Reflect upon the implications/tradeoffs it may have as it relates to my current goals.

Parameters are automatically validated based on their type definitions and injected into your prompts at runtime.

Sequential Tool Configuration

In sequential mode, tools are executed in a specific order:```yaml
web_debugger_mode:
description: Debug my codebase my web application with browser logs and BrowserTools MCP
prompt: |
Deeply reflect upon all of this and think about why this isn't working. Theorize 4-6 different possible sources of the problem.
toolMode: sequential
tools: getConsoleLogs, getConsoleErrors, getNetworkLogs, getNetworkErrors, takeScreenshot


This will return the following MCP response:```
Deeply reflect upon all of this and think about why this isn't working. Theorize 4-6 different possible sources of the problem.

## Available Tools
If all required user input/feedback is acquired or if no input/feedback is needed, execute this exact sequence of tools to complete this task:

1. getConsoleLogs
2. getConsoleErrors
3. getNetworkLogs
4. getNetworkErrors
5. takeScreenshot

After using each tool, return a 'Next Steps' section with a list of the next steps to take / remaining tools to invoke along with each tool's prompt/description and 'optional' flag if present.

Advanced Tool Configuration

Includes ability to define prompts for each tool and an optional flag to indicate if the tool is optional or not in the sequence:```yaml
deep_thinking_mode:
description: Reflect on a thought and produce a reflection/new set of thoughts
parameters:
thought:
type: string
description: A thought to deeply reflect upon
required: true
prompt: |
Deeply reflect upon the provided thought.
Reflect upon the implications/tradeoffs it may have as it relates to my current goals, challenges and our conversation.
Do not change anything in our system, just return some thoughts/considerations/analysis based on your reflection of the provided thought.
toolMode: "sequential"
tools:
analyze_thought: analyze a previously generated thought
explore_perspectives: think about additional perspectives given the analysis
apply_findings:
propmt: implement the findings of the analysis
optional: true


### Configuration Structure

Each YAML file should contain a mapping of tool names to their configuration. Configurations can be loaded from two sources:

1. Internal presets (located in the `presets` directory)
2. User-defined configurations (in `.workflows` or `.mcp-workflows` directory)

### Basic Tool Configuration

For each tool, you can specify:

- `name`: Optional name override for the registered tool (default is the config key)
- `description`: Description of what the workflow/tool does
- `prompt`: Custom prompt (completely replaces default prompt if workflow name is also an active preset)
- `context`: Additional context to append to the prompt (does not replace default prompt for presets)
- `tools`: Array or object of tools available in this mode, with flexible definition styles
- `toolMode`: Mode of tool execution, either "sequential" or "situational" (defaults to "situational")
- `parameters`: Object mapping of parameters as input to the tool - supports template injection using {{ parameter_name }}
- `disabled`: Boolean to disable the tool

### Input Parameter Configuration

Tools can accept typed parameters that an MCP Client / agent can provide to improve reasoning capabilities for your prompt/workflow. These inputs are automatically converted to Zod schemas for validation and type safety. Each parameter is defined with the following properties:

| Property      | Description                                                            | Required                 |
| ------------- | ---------------------------------------------------------------------- | ------------------------ |
| `type`        | Data type: "string", "number", "boolean", "array", "object", or "enum" | Yes                      |
| `description` | Human-readable description of the parameter                            | No                       |
| `required`    | Whether the parameter is required                                      | No (defaults to `false`) |
| `default`     | Default value if not provided                                          | No                       |
| `enum`        | Array of valid values (required for `enum` type only)                  | Yes (for `enum` type)    |
| `items`       | For array types, defines the type of items in the array                | No                       |
| `properties`  | For object types, defines the properties of the object                 | No                       |

#### Example Input Parameter Types```

常见问题

7 问
答案由本站依据条目字段整理,不引用上游原文
Workflows MCP 怎么安装?

在客户端按下面的安装命令接入:npx -y mcpn@latest init。。装完在客户端配置里引用同一个包或端点。

Workflows MCP 怎么在客户端里配置?

它的传输方式未标注,在客户端 MCP 配置里按官方文档填写启动命令和参数,保存后重启客户端生效。

Workflows MCP 收费吗?

本站不代收费用,也没有标注具体价格。是否收费取决于服务本身和背后的接口额度,正式使用前按官方说明确认。

Workflows MCP 需要 API Key 吗?

资料标注为无需认证,装好就能连;如果实际调用时报鉴权错误,按官方文档补上凭据。

Workflows MCP 能提供哪些能力?

资料显示能力标签包括 版本控制、JavaScript、AI工作流。具体工具清单以官方文档为准。

Workflows MCP 需要什么运行环境?

运行时为 Node.js;属于本地进程,需要在本机装好运行环境。

Workflows MCP 是本地还是远程?

它是本地进程(stdio),由客户端在本机拉起。传输方式为stdio · 本地进程。

热门专题

按标签挑服务器
全部专题 →

服务器信息以作者仓库与官方文档为准;分类、传输方式与鉴权方式的中文说明为中文整理,最近核验 2026-05-18。