convez.ai
MCP ServerAutomaçãoIntermediário
Usman1Abbas

Workflow Automation Scout

Usman1Abbas/Custom-built-MCP

Identifica workflows automáveis, calcula economia e esboça código

Ver no GitHub
InicialPluginScore · de 0 a 100

Servidor MCP em TypeScript que identifica oportunidades de automação em fluxos de trabalho, analisa o ROI e gera código-template para implementação. Integra-se com clientes MCP como Claude Desktop e Cursor via stdio, com suporte a containerização Docker.

Instalação

$npx convez-ai Usman1Abbas/Custom-built-MCP

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
0
Forks
0
Acompanhando
1
Saúde
28%
Último commit
há 5 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Analisar rotinas repetitivas para encontrar oportunidades de automação
  • Gerar templates de código para tarefas manuais recorrentes
  • Estimar retorno sobre investimento (ROI) de automatizações

Veredito

MCP server bem estruturado e prático para identificar e automatizar workflows repetitivos. Implementação sólida em TypeScript com suporte Docker, mas depende de integração com cliente MCP compatível para funcionar.

README

A tradução para o português deste README ainda está na fila.

Workflow Automation Scout is a custom Model Context Protocol (MCP) server, written in TypeScript on top of the official @modelcontextprotocol/sdk. It plugs into any MCP-capable LLM client (Claude Desktop, Cursor, etc.) over stdio and gives the agent two purpose-built tools: one that scores a described workflow for automation potential, and one that drafts implementation code for the tasks worth automating. It ships with a Dockerfile for containerized deployment.

Point an agent at the boring parts of your week — it tells you what's worth automating, hands you a starting template, and shows you the time you'd get back.


✨ Tools & Resources exposed

This server registers two MCP tools and does not currently expose any MCP resources.

🔧 analyze-workflow

Analyzes a natural-language description of a workflow to surface automation opportunities. It accepts an array of workflow steps, each with:

  • description — what the task is
  • frequency — daily · weekly · monthly · quarterly · yearly
  • duration — minutes per occurrence
  • tools — current tools used (optional)

It returns, per step, an automation score, the potential time savings, recommended automation tools, and an overall ROI analysis.

🔧 generate-automation-code

Generates an implementation code template for a task you've decided to automate. It accepts:

  • task — the task description
  • tool — the selected automation tool
  • language — target language (optional, defaults to Python)

It returns a code template with the basic implementation structure for the chosen tool.


🏗️ Architecture

A thin, single-process MCP server that talks to the host LLM client over stdio — no network service is required for normal use.

┌──────────────────────────┐        stdio (MCP)        ┌──────────────────────────────┐
│   MCP client / LLM agent  │ <───────────────────────> │  Workflow Automation Scout    │
│  (Claude Desktop, Cursor) │       JSON-RPC             │  (TypeScript, MCP SDK)        │
└──────────────────────────┘                            │                              │
                                                        │   ├─ analyze-workflow         │
                                                        │   └─ generate-automation-code │
                                                        └──────────────────────────────┘
  • Language / build: TypeScript (strict mode), compiled with tsc from src/ → build/.
  • Module system: ES modules (Node16 resolution), entry point build/index.js.
  • Transport: stdio, per the MCP SDK.
  • Tests: Jest + ts-jest (ESM preset).
  • Container: node:18-slim, builds on image start; EXPOSE 3000 is reserved for an optional HTTP transport.

🚀 Run it

Requirements: Node.js 18+.

# 1. install dependencies
npm install

# 2. build the TypeScript sources (src/ -> build/)
npm run build

# 3. start the server (stdio transport)
npm start

Useful during development:

npm run dev     # tsc in watch mode
npm test        # run the Jest test suite (ESM)

🐳 Docker

# build the image
docker build -t workflow-automation-scout .

# run it
docker run --rm -i workflow-automation-scout

The image is based on node:18-slim, installs dependencies, compiles the TypeScript, and starts the server via npm start. Port 3000 is exposed for an optional HTTP transport.


🔧 Config / client integration

The server speaks MCP over stdio, so any MCP-capable client can launch it directly. After npm run build, register the compiled entry point with your client.

Claude Desktop / Cursor — add an entry to your MCP servers config:

{
  "mcpServers": {
    "workflow-automation-scout": {
      "command": "node",
      "args": ["build/index.js"]
    }
  }
}

Environment configuration is loaded via dotenv from a .env file (e.g. ILLEGAL_AGENTS_API_KEY). For VS Code development and debugging, the server can also be wired up through a .vscode/mcp.json file.

Deployment metadata (image name, capabilities, tool manifest) lives in illegalagents.yml.


📋 Example

A request to analyze-workflow:

{
  "steps": [
    {
      "description": "Extract data from weekly sales reports (PDF) and compile into Excel",
      "frequency": "weekly",
      "duration": 45
    },
    {
      "description": "Send automated email reports to stakeholders",
      "frequency": "daily",
      "duration": 15,
      "tools": ["Outlook"]
    }
  ]
}
Ver o original no GitHub

Bom para

Relacionados