agentgate

Overview

AgentGate provides a comprehensive approval workflow system for AI agents, allowing them to request human approval for potentially risky actions. It includes a policy engine that can auto-approve safe actions, auto-deny dangerous ones, and route everything else to humans via multiple channels including Slack, Discord, email, or a web dashboard. The system features a complete audit trail, API key management, webhook support, and is designed with production-hardened security features including SSRF protection and ReDoS defense.

Try asking AI

After installing, here are 5 things you can ask your AI assistant:

• agentgate_request_approval

  Create a new approval request

  Try:I need to send an email to customer@example.com with subject 'Order shipped'Try:Request approval to delete the temp fileTry:Can I deploy to production?
• agentgate_check_request

  Get status of an approval request

  Try:Check the status of request req_abc123Try:Has the file deletion request been approved?Try:What's the current status of my deployment request?
• agentgate_list_requests

  List pending approval requests

  Try:Show me all pending requestsTry:List all approval requests with high urgencyTry:What requests are waiting for approval?
• agentgate_list_policies

  List configured policies

  Try:Show all configured policiesTry:List all approval rulesTry:What policies are currently active?
• agentgate_create_policy

  Create a new policy rule

  Try:Create a policy to auto-approve email sendingTry:Add a rule to deny file deletion requestsTry:Set up a policy for production deployments
• agentgate_update_policy

  Update an existing policy

  Try:Update the email approval policy to include attachmentsTry:Modify the file deletion policy to require approvalTry:Change the urgency level for deployment requests
• agentgate_delete_policy

  Delete a policy

  Try:Remove the old email approval policyTry:Delete the temporary deployment policyTry:Unset the auto-approve rule for file operations

Comparable tools

Installation

pnpm install

pnpm --filter @agentgate/server db:migrate

pnpm --filter @agentgate/server bootstrap

pnpm dev

{
  "mcpServers": {
    "agentgate": {
      "command": "npx",
      "args": ["@agentgate/mcp"],
      "env": {
        "AGENTGATE_URL": "http://localhost:3000",
        "AGENTGATE_API_KEY": "agk_..."
      }
    }
  }
}

FAQ

How does the policy engine work?

The policy engine evaluates requests against configurable rules that determine if actions should be auto-approved, auto-denied, or routed to humans for review. Policies can be based on action type, parameters, agent identity, and other contextual factors.

Can I integrate AgentGate with existing AI agent frameworks?

Yes, AgentGate provides a TypeScript SDK and MCP server, making it compatible with most AI agent frameworks, including direct integration with Claude Desktop via MCP.

Compare agentgate with