mcp-orchestrator
Verified Safeby mrorigo
Overview
Orchestrates multiple Model Context Protocol (MCP) servers, enabling LLMs to generate and execute TypeScript code for complex, multi-step workflows, and integrating with Agent-to-Agent (A2A) protocols for collaborative AI.
Installation
ANTHROPIC_API_KEY=your_key npx ts-node examples/a2a_server.tsEnvironment Variables
- OPENAI_API_KEY
- ANTHROPIC_API_KEY
- ENABLE_SNIPPET_MODE
- SNIPPET_STORAGE_PATH
- OPENAI_BASE_URL
Security Notes
The project uses a secure sandboxed `vm` environment for executing LLM-generated TypeScript code. Dangerous Node.js globals (`process`, `require`, `__dirname`, `__filename`, `global`, `Buffer`, `module`, `exports`) are explicitly blocked. Execution includes a configurable timeout. All external interactions (e.g., file system, network) must occur through registered MCP tools, which can be controlled and audited. The `SamplingSecurityManager` further enhances security by implementing approval workflows, rate limiting, and policy enforcement for LLM sampling requests, ensuring human-in-the-loop oversight for sensitive operations. No direct `eval` calls in critical paths were found, and code is compiled into a script for execution. The overall design prioritizes secure execution of untrusted code.
Similar Servers
ironmanus-mcp
Orchestrates AI workflows with an 8-phase control flow and specialized tools, serving as a Model Context Protocol (MCP) server.
project-nova
Manages Paperless-NGX documents by providing an MCP server that exposes an SSE endpoint for integration with n8n, enabling AI agents to search, upload, and manage documents, tags, correspondents, and document types.
AutotaskMCP
Manage Autotask PSA (Professional Services Automation) data, including tickets, companies, contacts, time entries, and resources, through a Claude Desktop MCP server.
mcp-collection
Provides a containerized server application, likely part of a larger collection or system, designed for automated dependency management.