Back to Home
mrorigo icon

mcp-orchestrator

Verified Safe

by mrorigo

Overview

Orchestrate Model Context Protocol (MCP) servers, allowing LLMs to execute TypeScript code for complex workflows, manage multi-server coordination, and handle dual-level LLM sampling.

Installation

Run Command
cd examples/code-mode-basic && OPENAI_API_KEY=your-key npm start

Environment Variables

  • OPENAI_API_KEY
  • ANTHROPIC_API_KEY
  • OPENAI_BASE_URL
  • HOST
  • PORT
  • AUTH_HOST
  • AUTH_PORT
  • AUTH_REALM
  • OAUTH_CLIENT_ID
  • OAUTH_CLIENT_SECRET

Security Notes

The CodeExecutor provides robust sandboxing for LLM-generated TypeScript code, explicitly blocking dangerous Node.js globals (`process`, `require`, `Buffer`) and enforcing execution timeouts. The SamplingSecurityManager implements critical human-in-the-loop approval workflows, rate limiting, and policy enforcement for LLM sampling requests, significantly enhancing control and preventing abuse. No hardcoded secrets are evident, with API keys and sensitive configurations expected from environment variables. A minor concern for production deployments is the `enableDnsRebindingProtection: false` setting in `StreamableHTTPServerTransport` for development ease, which should be reviewed. The primary security risk involves potential vulnerabilities in the underlying Node.js `vm` module or exploitable *registered MCP tools* that the sandboxed code can interact with. Overall, it is well-designed for security within its specified scope.

Similar Servers

Stats

Interest Score35
Security Score8
Cost ClassMedium
Avg Tokens500
Stars2
Forks0
Last Update2025-12-06

Tags

MCPAI agentsLLM orchestrationcode executionsandboxing