Agent frameworks & infrastructure
Agent sandboxes & runtimes
1,721 agents · most starred first
Agent sandboxes & runtimes, page 3 of 18
Ten Engines. One local-first Harness OS.
A Runtime Framework for Agent Deployment and Tool Sandbox. AgentScope Runtime Java Implementation.
Ever® Works™ - The Workshop for AI. An open agentic runtime that autonomously researches, ships, and maintains entire businesses, 24/7 - https://ever.works
Deploy a fleet of AI agents (OpenClaw, Hermes) on your machine in 10 minutes — use your ChatGPT subscription, no cloud bills. Open-source fleet manager with browser dashboard.
Community Docker and Kubernetes packaging for DeepSeek Harness (@deepseek-ai/dsh), with a hardened image, Compose stack, Helm chart, Web UI, and headless CLI.
Containment for AI agents - user isolation, sandboxed execution, network controls, backup/rollback. TLA+ verified.
Keeping your agents out of trouble with sandboxed coding agent automation
This is a fork of the https://dockbox.dev project I made
An open-source Python SDK and CLI Starter Toolkit for deploying AI agents to Volcengine AgentKit Runtime.
Run AI Agent CLIs in your browser
A production-grade agent runtime engineered from the Harness Engineering philosophy.
Committed to being the best code interpreter in the world.
Self-hosted auto clustering AI agent OS for consumer hardware like the computer you already own, an Orange or Raspberry Pi or a Mac Mini etc. Full desktop, app store, agent deployment, distributed compute cluster. Memory by taOSmd.
Open-source managed runtime for OpenAI Codex, Claude Agent SDK, OpenCode, and Pi Agent. Run coding agents through one Agents API with isolated cloud sandboxes, durable sessions, tool streaming, and managed files.
It's like Auto-GPT met Brew. The easiest and fastest way to get started with AutoGPT with any backend of your choice & whatever plugins you may need
The open-source agent SDK that gives every agent its own VM. Each VM is a microVM sandbox where the agent installs software, runs code and uses the GPU. Python, Node.js and Rust.
Use HuggingMes to Run Harmess agent for free on HuggingFace Space
A lightweight secure Self-evolution engine built in Rust, featuring a built-in native system-level sandbox, zero dependencies, and fully local execution.
Claude Code running in sandbox. Packed as single portable executable with no dependency. Has better performance than official version.
Sandbox for AI coding agents. Runs Copilot CLI, Claude Code, OpenCode, Gemini CLI, Antigravity, Pi, goose or a plain shell inside a kernel-level sandbox, with git and gh guards and sandbox policy committed to the repository.
CortexPrism — open-source AI agent operating system
A lightweight Windows framework for launching isolated AI desktop applications and CLI tools with independent profiles, workspaces, and environments.
Safe local execution layer for AI agent tools. Build and test-run MCP tools (no pass, no run), serve them from a built-in MCP server, and use them in Agentic Chat with RAG, human-in-the-loop approval, and built-in observability. Cross-platform desktop app powered by Spring AI.
UniHarness (formerly HexAgent) – An agent harness that gives any LLM a computer to complete tasks the way humans do
MarinaBox is a toolkit for creating and managing secure, isolated environments for AI agents
Deploy secure local and cloud environments for developers and agents
Run your AI agents on machines you already own, as one private P2P mesh. OpenAI-compatible, sandboxed with Podman, signed reputation, no cloud.
Secure Nix sandbox for LLM agents
Workshops are secure, fast, and composable development environments that come agent-ready.
File Storage & Sandbox Backends for Pydantic AI: console tools for file operations, Docker-isolated sandboxes for safe execution, and permission system with presets for access control. Enables secure multi-user handling and testing in agents via in-memory, local, or containerized storage.
The AI agent sandbox runtime. Boot any OCI image as a Firecracker microVM, Linux container, or Apple Silicon VM in ~400 ms — zero-copy DAG storage, P2P image sync, native MCP for opencode/Claude Code/Cursor.
Run AI coding agents in hardened container sandboxes.
the agi compiler: records llm agent behavior, proves what repeats, and compiles it into verified, sandboxed wasm binaries that run for microdollars. nothing figured out twice, paper: https://arxiv.org/abs/2607.04542
All-in-one Docker/Podman image: hermes-agent + hermes-webui + hermes-dashboard in a single container
Open-source agent sandbox infrastructure for parallel coding agents: isolated workspaces, MCP/CLI control, observability, and atomic publication.
An operating system built for AI agents — talk to your NixOS server instead of SSH-ing in. Typed, audited tool access with atomic rollback on every change. Research-grade; run it on a disposable box, not production.
HexAgent – An Agent harness that gives any LLM a computer to complete tasks the way humans do
Self-hostable, open-source alternative to Claude Managed Agents. Multi-LLM support (Anthropic, OpenAI, Ollama), enterprise governance (org/teams/RBAC)
A web-based AI workspace with a local execution daemon, with an emerging control-plane architecture for remotely operating AI coworkers across machines.
Self-hosted AI agent runtime — personal assistants, autonomous workers, and everything in between. Built on the Claude Agent SDK.
Heterogeneous Containerization of Agents
Self-hosted app store and runtime for AI agents. Install third-party agents, run them on your infrastructure with your own model providers (Ollama, Bedrock, OpenAI, etc.). Container isolation, credential injection, default-deny egress.
AgC is the open-core platform that powers Open Agentic Compute — a new compute substrate purpose-built for deploying, running, and orchestrating AI agents at scale.
The Agentic-Native Operating System
A self-hosted AI workspace with chat, code execution, parallel multi-agent orchestration, and a skill marketplace. Runs on macOS and Windows. Everything executes inside a secure Ubuntu sandbox — no Docker required.
An open taxonomy and scoring framework for evaluating AI agent sandboxes: 7 defense layers, 7 threat categories, 3 evaluation dimensions, 27 "sandboxes" scored.
TypeScript interpreter for AI agents. Written in Rust. 2µs cold start. Sandboxed. Alternative to MCP tool calling.
self-extending sandboxed agentic harness. a loop, two tools and an llm
Run LLMs and agents on TEEs leveraging NVIDIA GPU TEE and Intel TDX technologies.
the governed runtime for agent skill workflows, off the leash but on the record
Nosis is a lightweight personal agent runtime: per-session execution boundaries with OS-level sandboxing, computer use, a full built-in toolset, Skill / Plugin / MCP extensions, one runtime shared by GUI and TUI, and file-persisted sessions, plans, and memory.
KitAgent:面向算子开发与性能优化的 Coding Agent。以 LangGraph 状态机编排开发、验证与迭代,集成工具执行、人工审批、会话恢复、Skills / MCP 扩展与 Kubernetes 沙箱。
Agent operating system
The 30 MB open-source edge AI agent runtime. Run AI agents offline on Linux (Raspberry Pi, Jetson). GPIO, UART, MQTT as first-class nodes. Industrial protocols (OPC-UA, Modbus) on the roadmap.
Portable DeepSeek Harness for Windows, macOS and Linux. Keep chats, settings, plugins and workspace in one folder. 1.x wraps the official Windows desktop. Independent community project. | DeepSeek Harness 便携版,会话、设置、插件和工作区随文件夹迁移。稳定版支持三平台,1.x 基于官方 Windows 桌面端。独立社区项目。
Wiggle is a sandbox to run code in
RVM — The Virtual Machine Built for the Agentic Age, in Rust.
Run AI agents in safe VMs scoped to a local folder
Fast, self-hostable sandboxes for agents
Sandbox for Agents
Orchestrate sandboxed agents that run in the cloud while you work
A personal VM cloud in one Go binary: persistent Linux VMs on macOS, Linux and Windows, AI coding agents working inside them, any port published to HTTPS through Cloudflare Tunnel. Driven from a Mac OS 9 desktop in the browser, over SSH, or by an agent.
Sandbox for Agents
A docker image for running AI agents in YOLO mode
Run LLM-generated one-off scripts via the "Run Code Here" option in the Windows 11 context menu.
Sandboxed code execution API for AI agents: powers LibreChat's Code Interpreter
One TypeScript API for running code in sandboxes. Write it once, run it on E2B, Daytona, Vercel Sandbox, Upstash Box, Railway, Tenki, Freestyle, Docker, and more.
Kubernetes‑native control plane for running AI agents in containers
Slightly less YOLO version of pi-coding-agent in Docker container sandbox
A secure and efficient alternative to `eval` for coding agents
A durable agent harness: sessions that survive a crash or kill -9, approvals that persist, background tasks, sessions that start and message each other (preview), any model. A 2,200-line TypeScript kernel, with a coding agent for the terminal and the browser built on it.
Example repo showcasing how to represent anything as a filesystem to an agent via FUSE
Run multiple AI agents and untrusted code in isolation. One command spins up a disposable Podman container — your SSH keys, dotfiles, and env files stay safe.
DeepSeek Harness (dsh) 自托管多用户云平台:提供隔离工作空间、持久化存储、访问控制与资源配额。Self-hosted multi-user cloud platform for DeepSeek Harness, with isolated workspaces, persistent storage, access control, and resource quotas.
The agent runtime built to operate.
ML/AI training/serving and agent sandbox operator and controller for Kubernetes
The Secure, Persistent Execution Platform for AI Agents
Infra for computer agents and software factories
智能体操作系统Agent OS
Deploy and manage AI agents in seconds.
OryxOS is an open-source Distributed AI Agent OS
Sandboxed code execution API for AI agents: powers LibreChat's Code Interpreter
DeepSeek Harness 三端兼容桌面启动器:多实例完全隔离、并行协作,协作画布编排 Agent 工作流;便携版 Exe 一键启动、仅约 20M;双隔离机制让兼容性极强
Coding Agent Sandboxes
Run AI coding agents in isolated sandboxes connected to your GitHub repositories
Agentic hypervisors for the 21st century
Android No-Root Linux Runtime · Native Agent Harness · PTY Terminal · Workspace & Tool Ecosystem
Kubernetes-native runtime for persistent, collaborative AI agents.
Use Anthropic Claude Agent SDK in a Cloudflare Sandbox
🚀 Clone your OpenClaw AI Agent to a new device in ~25 minutes — configs, memory, skills, everything.
Sandbox for running AI coding agents autonomously: isolated, network-restricted, host-safe
Privacy-first desktop AI that runs agents, automations and MCP servers locally and in parallel, with your own model keys. Being rebuilt for Mac and Windows.
Firecracker made simple. Spin up secure microVMs in milliseconds, from install to interactive shell in one command.
Fast cold-boot MicroVM sandboxes for AI agents on cocoon
CLI for managing OpenSandbox lifecycle and AIO agent tools
Sandbox for AI coding agents (Claude Code, opencode, pi): runs each project in an isolated microVM with host-side secret injection. Linux/KVM & macOS.
A sandbox for coding agents with parametrized state.
Sandboxed shell agent with access to LiteParse
A unified, 100% self-contained, open-source, and offline-first local AI agent workspace.