- LLM Sandbox @vndee/llm-sandboxSecurely run LLM-generated code in isolated containers across 7 languages and 3 container backends.5% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from PyPI.1,130 GitHub stars.
- Agent Swarm @desplega-ai/agent-swarmFull Agent Swarm API and MCP server with local SQLite storage for multi-agent orchestration.6% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from npm.873 GitHub stars.
- Runno @taybenlor/runnoMCP Server for the Runno Sandbox4% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from npm.773 GitHub stars.
- SandBase Harness @sandbaseai/sandbase-harnessStdio MCP bridge for SandBase Harness agents, sessions, turns, artifacts, and cancellation.5% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from Docker.707 GitHub stars.
- haiku.rag @ggozad/haiku-ragLocal-first agentic RAG with citations - hybrid search, reranking, multimodal document retrieval4% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from PyPI.622 GitHub stars.
- Flyto Core @flytohub/flyto-coreExecution engine for AI agents. 412 modules: browser, file, Docker, data, crypto.6% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from PyPI.483 GitHub stars.
- Superserve @superserve-ai/mcpMCP server for Superserve sandboxes: create, exec, and manage Firecracker microVMs9% chance this is malicious. Assessed by AI, which can make mistakes.A remote server; it needs your API key.465 GitHub stars.
- Open Computer Use @yambr/open-computer-useGive any LLM its own computer — Docker sandboxes with bash, browser, docs, and sub-agents9% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from a package registry.126 GitHub stars.
- Codex Control Plane MCP @aresyn/codex-control-plane-mcpDurable MCP control plane for long-running Codex Desktop tasks.4% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from PyPI.124 GitHub stars.
- Flow Nexus @ruvnet/flow-nexusCloud-powered AI platform with multi-agent swarms, sandboxes, and workflow automation7% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from npm.106 GitHub stars.
- Ipybox @gradion-ai/ipyboxUnified execution environment for Python code, shell commands, and programmatic MCP tool calls5% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from PyPI.76 GitHub stars.
- Cybersec Toolkit @26zl/cybersec-toolkitAuthorization-gated MCP server to discover and run 670+ security tools for CTF, pentest, and DFIR.9% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from Docker.69 GitHub stars.
- Js @r33drichards/mcp-jsRun JavaScript and TypeScript in sandboxed V8 isolates with policy-controlled host capabilities.7% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from Docker.61 GitHub stars.
- Codex MCP Go @w31r4/codex-mcp-goMCP server wrapping Codex CLI for stdio-based tool calls.7% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from npm.52 GitHub stars.
- OpenSTAAD MCP Server @bentleysystems/openstaad-mcpMCP server for Bentley STAAD.Pro via the OpenSTAAD API3% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from MCPB.49 GitHub stars.
- Mobile Device MCP @srmorete/mobile-device-mcpControl iOS and Android devices with multi-device and seamless Native/WebView support.7% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from npm.48 GitHub stars.
- Ephemora Cell MCP — WASM Sandbox @michaels1011/ephemora-cell-mcpWASM sandbox for MCP tools. WASI-isolated, fuel/memory/I/O capped, attested.4% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from PyPI.46 GitHub stars.
- Vinv @vinvai/vinvVinv runs, tests, and finds issues in your services — with zero code changes.4% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from npm.46 GitHub stars.
- Proletariat Cli @chrismcdermut/proletariat-cliAgent orchestration platform - multi-agent spawning, tickets, boards, and workflows6% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from npm.36 GitHub stars.
- E2b @mcparmory/e2bCreate and manage secure sandboxes for running AI agent code8% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from PyPI.32 GitHub stars.
- Solidity Compiler IDE @nirholas/solidity-compilerBrowser Solidity compiler IDE - smart contract development testing deploy5% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from a package registry.27 GitHub stars.
Code execution sandboxes MCP servers
A server here runs code or commands in an isolated sandbox or container for an AI agent.
Not here: deploying apps to production.
107 remote200 localTagged by Odel's AI, which can make mistakes
Most popular in Code execution sandboxes
By GitHub stars, at most two per publisher- Number 1: AIHawk@feder-cr/aihawkAI browser agent: browses, clicks, types, and reads real web pages from plain-English instructions.9% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from PyPI.31,941 GitHub stars.
- Number 2: Bernstein@sipyourdrink-ltd/bernsteinThe open-source governance layer for AI agents. Byte-identical run receipts, 40+ adapters, air-gap.7% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from PyPI.1,215 GitHub stars.
- Number 3: Magic Cloud@polterguy/magicGenerate secured CRUD APIs over your database, run SQL, manage files, tasks and a headless browser8% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from Docker.1,200 GitHub stars.
