Agentgate

Name: Agentgate
Author: amitpaz1

@amitpaz1

Repository

About Agentgate

No overview available yet

Basic information

Config

Add this server to your MCP-compatible client using the configuration below.

{
  "mcpServers": {
    "agentgate": {
      "command": "npx",
      "args": [
        "@agentgate/mcp"
      ],
      "env": {
        "AGENTGATE_URL": "",
        "AGENTGATE_API_KEY": ""
      }
    }
  }
}

Tools

No tools detected

We auto-extract tools from the README. The maintainer can list them under a ## Tools heading to populate this section.

Overview

What is AgentGate?

AgentGate is a human-in-the-loop approval system for AI agents. It allows agents to request actions while policies decide and humans approve, deny, or ignore them. It integrates with any agent framework or Claude Desktop via a TypeScript SDK and MCP server.

How to use AgentGate?

Install dependencies with pnpm install, run database migrations, bootstrap an admin API key, and start the server and dashboard. Agents then use the SDK, MCP tools, or CLI to create approval requests, which appear in the dashboard, Slack, Discord, or email for human review.

Key features of AgentGate

Policy engine for auto-approve, auto-deny, or human routing.
Multi-channel approvals via Slack, Discord, email, and web dashboard.
TypeScript SDK + MCP integration with Claude Desktop.
Webhooks with retry and exponential backoff.
Full audit trail for every request and decision.
Docker-ready with one docker-compose up.
API key authentication with fine-grained scopes and rate limiting.

Use cases of AgentGate

Auto-approve safe actions (e.g., read operations) and auto-deny dangerous ones (e.g., delete files).
Route production deployments to a human via Slack or email for approval.
Maintain a complete audit trail of all agent actions and human decisions.
Integrate Claude Desktop with MCP to gate email sending, file operations, or API calls.
Let teams manage approval flows through a web dashboard with filtering and search.

FAQ from AgentGate

What is AgentGate?

AgentGate is a human-in-the-loop approval system for AI agents. It uses a policy engine to automatically approve safe actions, deny dangerous ones, or route requests to a human via Slack, Discord, email, or a web dashboard.

What channels can I use to approve requests?

You can approve requests through the web dashboard, Slack bot, Discord bot, or one-click links in notification emails.

How do I integrate AgentGate with Claude Desktop?

Add the @agentgate/mcp MCP server to your claude_desktop_config.json with the server URL and API key. It provides tools like agentgate_request_approval and agentgate_check_request.

What are the runtime requirements?

AgentGate requires Node.js, pnpm, and either SQLite (default) or PostgreSQL. It runs on Hono and can be deployed with Docker.

How is authentication handled?

All API requests (except /health) require a valid API key passed via the Authorization: Bearer header. API keys have scopes such as admin, request:create, request:decide, and webhook:manage.