Pré-requisitos

• Python 3.10+
• Chave API Pro do CleanTextLab
• Strands Agents instalado

python -m venv .venv
source .venv/bin/activate
pip install strands-agents

---

Opção A: MCP local via npx (recomendado para dev local)

from strands import Agent
from strands.tools.mcp import MCPClient
from mcp import stdio_client, StdioServerParameters

cleantextlab = MCPClient(
    lambda: stdio_client(
        StdioServerParameters(
            command="npx",
            args=["-y", "cleantextlab-mcp"],
            env={"CLEANTEXTLAB_API_KEY": "ctl_live_YOUR_KEY"},
        )
    )
)

with cleantextlab:
    agent = Agent(tools=cleantextlab.list_tools_sync())
    agent("Remove line breaks from: Hello\nworld")

---

Opção B: MCP hospedado (SSE)

from strands import Agent
from strands.tools.mcp import MCPClient
from mcp import sse_client, SSEServerParameters

cleantextlab = MCPClient(
    lambda: sse_client(
        SSEServerParameters(
            url="https://cleantextlab.com/api/mcp/sse",
            headers={"x-api-key": "ctl_live_YOUR_KEY"},
        )
    )
)

with cleantextlab:
    agent = Agent(tools=cleantextlab.list_tools_sync())
    agent("Generate an ASCII tree from: src/app/page.tsx")

---

Por que MCP para agentes?

• Saída determinística para tarefas de formatação
• Menos tokens ao delegar operações pesadas
• Ferramentas reutilizáveis em workflows de agentes

---

Próximos passos

• Guia MCP: /docs/mcp-setup
• Referência da API: /docs/api
• Automation Hub: /automation