MCP server is used to create a secure code sandbox environment, execute code in Docker containers, and provide code execution capabilities for AI applications.

Secure Node.js execution sandbox for AI. Allows coding agents & LLMs to dynamically run JavaScript, install NPM packages, and retrieve results, facilitating code generation, testing, and interactive assistance. MCP-compatible.

Python sandbox escape detection, SQL injection taint analysis, Seccomp BPF filter generation, and ReDoS regex scanning.

# 🐢🚀 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.

Easy, fast, isolated microVM sandboxes for AI agents and untrusted code on AWS Lambda MicroVMs — Python SDK, asb CLI, and MCP server.

Local AI agent sandbox — run isolated Linux VMs on your Mac in ~60ms. No cloud, no containers sharing your kernel. Works with Claude Code, Cursor, and any MCP client.

LangGraph agent with Docker-based sandboxed code execution (Python, Node.js, R). Hardened containers with persistent kernels, vision support, provider-agnostic LLMs, cross-session file transfer, and session GC. Available as CLI, MCP server, REST API (Aegra), and Streamlit UI.

MCP server for code operations with sandboxed execution

MCP server for secure, sandboxed code execution using Nix Flakes and Bubblewrap

MCP server that orchestrates code execution in a sandboxed Deno environment with access to all MCP servers

Sandboxed TypeScript runtime for AI coding agents to query AWS — full SDK access with in-sandbox aggregation, credentials flow via IMDS and never enter the VM. Ships as a Claude Code plugin (MCP) and a standalone CLI, backed by Shuru microVMs or Docker.

Rust pre-apply verification gate for AI coding agents — sandbox-checks edits before they touch disk. MCP server + daemon + CLI.

Enterprise-grade Zero-Leak Secret Defense & Sandbox for Autonomous AI Coding Agents

Turn ChatGPT Web Chat, Grok, Claude Code, Codex, Cursor or VS Code into a coding agent for one folder on your machine, sandboxed and logged on a live dashboard. MCP server.

Code Execution Sandbox Platform Solution for Individuals and Small Teams with MCP

Reference implementation of Anthropic's 'Code execution with MCP.' Reuses Claude Code's own sandbox and config to keep intermediate tool output out of the context window on every workflow — up to 99% fewer tokens.

Stop flooding your agent's prompt with MCP tools: MCPtoCodeMode puts every MCP server behind 3 tools and a sandbox, for Codex, Claude Code, Cursor, Grok, OpenCode, Pi, or any MCP client.

The DAG-Native Symbolic AST Slicing Hypervisor & Safety Sandbox for AI Coding Agents

Load-tested Isolyth's MCP tool-calling endpoint with 100 concurrent virtual users via Locust, sustaining ~535 req/sec at p95 latency of 160ms with 0% error rate; benchmarked WASM-sandboxed code execution against plain tool calls, showing only ~15-30ms of added latency overhead from sandboxing —proving security isolation came at min performance cost

Universal code & logic calculator for AI models: 51 MCP tools across 31 languages. Rust-sandboxed execution with verdicts, sessions, artifacts, exact arithmetic, and verified translation/optimization. No LLM, no gateway, no telemetry: the caller is the model. Apache-2.0.

PRIMS is a lightweight, open-source Model Context Protocol (MCP) server that lets LLM agents safely execute arbitrary Python code in a secure, throw-away sandbox.

Self-tuning, disposable Docker sandboxes for AI agents. Runs code in locked-down containers sized to your hardware, with GPU sharing, crash cleanup, a CLI, a Python API and an MCP server for Claude.

Safety-first Database Guardrail & Sandboxing MCP Server for AI Coding Agents (Cursor, Claude Desktop, Antigravity). Protects databases from unconstrained mutations, drops, locks, and PII leaks.

Official Mailtrap MCP server — over 125 tools to send email from Claude, Cursor, VS Code & other AI assistants. Send transactional and bulk email, test in Email Sandbox, manage templates, contacts, domains and webhooks, and inspect delivery logs and stats.

Realistic data for your Salesforce sandboxes: clone a record and its graph or a whole org, generate, anonymize and replay datasets, empty objects or an org, checkpoint an org and rewind it. A VS Code extension with production guardrails.

MCP server for sandboxed code execution with context window reduction

ToolsEnabled Fleet: agent fleet management and mediation for Codex and Claude Code inside NVIDIA OpenShell sandboxes. Public source for each published build.

MCP server for Errata: sandbox-verified corrections where AI models write outdated code for PyPI and npm libraries

Run several Claude Code agents in sandboxed containers on a Linux VM, and drive them from your own Claude Code via MCP.

Codex (OpenAI) as an independent second opinion inside Claude, Claude Code and Cowork: parallel plans, critiques, code and diff reviews, red-teaming. Read-only sandbox.

Run code in isolated cloud containers. ConTree gives AI agents secure sandboxed execution environments with full root access, network, and persistent images.

Serving access to a JupyterHub as code interpreter sandbox for LLMs using MCP.

One MCP server for all the APIs your model needs (OpenAPI, GraphQL, SOAP/WSDL, atproto). Sandboxed code execution keeps your context clean.

🔨 Throwaway Docker sandbox for agents: run code/commands in isolated, resource-limited containers with structured results. CLI + web + MCP. Zero deps.

Your coding agent's change, verified on a real desktop. A fresh agent uses the app in a sandboxed Hyprland and leaves a video for the pull request.

Kern , The local, deterministic code-intelligence engine for AI agents. Fast AST indexing, precise dependency graphs, sandboxed execution, and MCP tools with zero network calls and zero telemetry.

Policy hook (fail-closed for checked tools) and tamper-evident evidence log for Claude Code. The code runs locally. Evidence supporting audits. Not a sandbox. PolyForm Noncommercial.

Enterprise-Grade Zero-Leak Secret Defense & Sandbox for Autonomous AI Coding Agents (Claude Code, Cursor, Windsurf, Codex, Antigravity)

iMessage MCP server for macOS without Full Disk Access. Sandboxed, no network, you approve every connection and every message. Works with Claude Code, Codex and any MCP client.

Meta Global AI Developer Hackathon. Show Tracera a bug. It investigates your code, proposes a fix, verifies it in a sandbox, and shows you what changed.

MCP intelligence layer — intent routing, perception, judgment, and sandboxed code execution across child MCP servers.

MCP server that turns Codex into a conductor for a multi-model coding pipeline: split plan → code with GLM → review with another model → auto-fix, all sandboxed with git commit/push blocked.

Stateless spreadsheet calculations via HTTP + remote MCP. Formulas instead of code. MCP: https://hyperformula-mcp.handsontable-sandbox.workers.dev/mcp

Autonomous Cognitive Amplifier for AI Coding Agents — CPG AST Radar, Sandboxed RLM, OKF Semantic Memory & Swarm Concurrency

Connecting coding agents across people and machines. Codex, Claude Code and Pi work together in one sandboxed workspace and, peer to peer, with your colleagues' agents. Each person directs their own agents and chooses what to share.

Let your AI coding agent (Claude Code or Codex CLI) drive your real Chrome — clicks, fills forms, grabs API keys, writes them to .env. Real session, real cookies, no fresh browser sandbox.

Aegis Jail MCP: least-privilege sandbox for AI coding agents and MCP servers. Hides ~/.ssh, secrets and the network, proves the walls hold, finds the minimal permissions. Linux, macOS, Windows. Claude Code plugin + MCP server.

A compiler for the code AI agents write: a flat SSA language with JSON diagnostics, compiled in under a millisecond to sandboxed native code, as MCP tools. Its Llama 3.2 1B prefills 1.7-2.4x faster than llama.cpp.

Free, local plugin that keeps tool output out of your coding agent's context window. Sandbox and SQLite FTS5 search, 99% fewer bytes in context. Works with Claude Code, Codex, Pi, OMP, OpenCode, Gemini CLI, Copilot CLI and 4 more agents. Zero telemetry.

Small MCP server for running PowerShell and CMD on Windows over WinRM: accurate exit codes, real timeouts, multi-host runs, sandboxed file transfer, security policy.

BitPull client: desktop app, CLI and MCP server. Join your computers and your friends' to run open AI models, reuse shared code and run sandboxed jobs.

Free security & privacy scanner for AI-coded apps. 766 rules, 84 MCP tools, 13-layer defense, zero-trust sandbox, AI firewall. Runs locally, never sends data anywhere. Scan your AI-generated app for leaked keys, open databases, and injection holes in 5 seconds.

Self-hosted AI API key manager and LLM gateway. Store OpenAI, Gemini, Claude, OpenRouter and Groq keys once, call any model by name. Key rotation, OpenAI-compatible proxy, web dashboard, and an MCP server with sandboxed Code Mode for Claude Code, Cursor and Codex.

The anti-hallucination gate for AI coding agents: 9 MCP tools (undefined-symbol verification in 17 languages with real toolchain adapters, hallucination-word scan, sandboxed run, evidence receipts), a cross-client hook and a CI gate — so 'done' means obse

MCP server that lets Claude Code and Codex see and operate a GNOME Wayland desktop: screenshot, click, type, drag by pixel coordinates. Headless sandbox by default so your mouse and keyboard stay free; real-desktop mode via uinput.

Secure MCP gateway that lets ChatGPT, Claude, Grok & other AI agents remotely code, run commands, manage Git, control browsers and desktop apps, operate multiple workspaces, and supervise your device/server from anywhere—with OAuth, approvals, sandboxing, DLP, audit logs & rollback.

Plugsky MCP — one remote endpoint for 20 AI tools (chat behind 7-tier failover, web search, memory, browsing, code sandbox) + npm stdio bridge. OAuth 2.1 + DCR.

Hardened MCP server giving AI agents real-browser automation: CDP-level control, dynamic Python hooks, full element cloning, AST-sandboxed for safe AI-supplied code.

Open-source alternative to Claude Code's hidden multi-agent Workflow tool. Fan out dozens of subagents from one Python script and keep only the final answer in context. Runs standalone or as an MCP Workflow tool inside Claude Code. Zero core deps, OS-sandboxed, MCP-native.

Scanners flag CVEs. CheckExploit proves which ones actually reach your code — by running the real exploit against your actual code inside an isolated sandbox. Supports Python,Node repos; generates and verifies remediation patches. Built on OpenCode + MCP.