convez.ai
MCP ServerAutomaçãoIntermediário
gotohuman

gotoHuman MCP Server

gotohuman/gotohuman-mcp-server

Adicione aprovações humanas a agentes IA com human-in-the-loop gerenciado

Ver no GitHub
BomPluginScore · de 0 a 100

MCP server para integrar fluxos de aprovação humana em agentes IA e workflows. Oferece gerenciamento assíncrono de aprovações com UI customizável, autenticação nativa, webhooks e recursos de equipe.

Instalação

$npx convez-ai gotohuman/gotohuman-mcp-server

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
52
Forks
10
Acompanhando
3
Saúde
28%
Último commit
há 1 mês
Versão
—
Issues abertas
1
Contribuidores
2

Casos de uso

  • Agentes IA que precisam de validação humana antes de executar ações críticas
  • Workflows multi-etapas com revisão e aprovação de conteúdo gerado
  • Automação com gates de qualidade e controle de risco

Veredito

Ferramenta prática para implementar human-in-the-loop em agentic workflows. Bem integrada ao ecosistema Claude/Cursor com API clara e suporte a workflows multi-etapas, ideal para casos que exigem validação humana antes de decisões críticas.

README

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

gotoHuman MCP Server

gotoHuman makes it easy to add human approvals to AI agents and agentic workflows.
A fully-managed async human-in-the-loop workflow with a customizable approval UI.
Enjoy built-in auth, webhooks, notifications, team features, and an evolving training dataset.

Use our MCP server to request human approvals from your AI workflows via MCP or add it to your IDE to help with integration.

Installation

npx @gotohuman/mcp-server

Use with Cursor / Claude / Windsurf

{
  "mcpServers": {
    "gotoHuman": {
      "command": "npx",
      "args": ["-y", "@gotohuman/mcp-server"],
      "env": {
        "GOTOHUMAN_API_KEY": "your-api-key"
      }
    }
  }
}

Install MCP Server

Get your API key and set up an approval step at app.gotohuman.com

Demo

This is Cursor on the left, but this could be a background agent that also reacts to the approval webhook.

https://github.com/user-attachments/assets/380a4223-ea77-4e24-90a5-52669b77f56f

Tools

list-forms

List all available review templates.

  • Returns a list of all available review templates in your account incl. high-level info about the added fields

get-form-schema

Get the schema to use when requesting a human review for a given review template.

  • Params
    • formId: The review template ID to fetch the schema for
  • Returns the schema, considering the incl. fields and their configuration

request-human-review-with-form

Request a human review. Will appear in your gotoHuman inbox.

  • Params
    • formId: The ID of the review template to use
    • fieldData: Content (AI-output to review, context,...) and configuration for the review template's fields.
      The schema for this needs to be fetched with get-form-schema
    • config: Configuration for the review template. Optional. The schema for this needs to be fetched with get-form-schema
    • title: Optional title shown in the inbox and notifications
    • webhookUrl: Optional webhook URL for this request (when the review template has no default webhook)
    • workflow: Optional object linking this review to a multi-step agentic workflow:
      • runId: Unique ID for the current workflow run. Use the same runId on every review in the same run. If workflow is sent without runId (even {}), or for manual triggers, gotoHuman creates a runId and returns it as workflowRunId for subsequent requests.
      • runName: Optional display name for the run (can be set or updated on any step)
      • prevSteps: Array of reviewIds from previous gotoHuman review steps (omit on the first step)
    • metadata: Optional additional data that will be incl. in the webhook response after review template submission
    • assignToUsers: Optional list of user emails to assign the review to
  • Returns reviewId, reviewLink, and optionally workflowRunId when gotoHuman assigned a new workflow run

Development

# Install dependencies
npm install

# Build the server
npm run build

# For testing: Run the MCP inspector
npm run inspector

Run locally in MCP Client (e.g. Cursor / Claude / Windsurf)

{
"mcpServers": {
  "gotoHuman": {
    "command": "node",
    "args": ["/<absolute-path>/build/index.js"],
    "env": {
      "GOTOHUMAN_API_KEY": "your-api-key",
      "GOTOHUMAN_AGENT_ID": "your-agent-id"
    }
  }
}
}

[!NOTE] For Windows, the args path needs to be C:\\<absolute-path>\\build\\index.js

Ver o original no GitHub

Bom para

Relacionados