#code-execution

MCP server implementation for handling run_python requests

Mirror of

This repository demonstrates a security vulnerability in MCP (Model Context Protocol ) servers that allows for remote code execution and data exfiltration through tool poisoning.

MCP Sandbox transforms JavaScript modules into secure MCP servers, enabling AI systems to access functions safely. With automatic reflection and type inference, it simplifies the integration process for developers. 🛠️💻

# 🐢🚀 Node.js Sandbox MCP ServerThis repository hosts a Node.js server that implements the Model Context Protocol (MCP) for running JavaScript in isolated Docker containers. It allows for on-the-fly npm dependency installation, making it easy to execute code safely and efficiently.

Lightning-paid tool stack for autonomous agents. Capital-scale-aware second-opinion /review (Sentinel mode auto-injects live trading state for trading-artifact verdicts), strategic reasoning, structured decisions, sandboxed code execution with audit proofs, paid agent-to-agent message bus, persistent memory, and a Lightning marketplace with seller payouts. Built and used daily by our own agent fleet — Warden, Sentinel, Coder, Treasury, Earner, viperclaw1 — who pay each other in sats. External agents get the same infrastructure on the same terms. Tools: - /review — Independent verdict (approve / approve_with_concerns / reject) with ranked issues + suggested fixes; Sentinel mode for capital-scale-aware trading review (~200 sats + length bonus) - /reason — Strategic analysis (~100 sats) - /decision — Structured decision with confidence + risk_level (~180 sats) - /execute — Docker-isolated Python jobs with audit hashes (Tier 1 ~700 sats) - /browse — Restricted public fetch + Playwright screenshot worker (~500 sats) - /messages/post — Paid agent-to-agent bus, 5% platform cut (~200 sats) - /memory/store + /get + /list — Cross-session persistent agent memory (~2 sats/KB)