Agent frameworks & infrastructure

Agent sandboxes & runtimes

1,721 agents · most starred first

Agent sandboxes & runtimes, page 7 of 18

Intendant★ 19

Open-source operating environment for autonomous AI agents, with cross-platform machine control, oversight, and multi-host federation.

hotcell★ 19

Self-hosted sandboxes for AI agents: Docker, Firecracker or Apple Virtualization microVMs. Your real API keys never enter the sandbox. Part of Directionally Local.

demi★ 19

Your agent, across all your machines. A self-hosted, web-based AI agent: one server holds every conversation and works in a cloud sandbox or on your Macs, PCs and servers.

modal-claude-agent-sdk-python★ 18

This package wraps the Claude Agent SDK to execute AI agents in secure, scalable Modal containers. It provides progressive complexity—simple usage mirrors the original Agent SDK, while advanced features expose Modal's full capabilities (GPU, volumes, image customization, etc.).

filepath★ 18

A personal Cloudflare-hosted development environment. Work lives in conversations, not terminal tabs.

opencode-docker★ 18

Stop prompt injection catastrophe! Run your AI Agents in secure isolated environment now.

ClaudePantheon★ 18

ClaudePantheon gives you a persistent, always-on Claude Code environment that you can access from any device with a web browser. Unlike running Claude Code locally, your sessions, context, and workspace persist across restarts — Claude remembers your projects, preferences, and ongoing work. Think of it as your personal AI development workstation.

ares★ 18

Agentic Runtime Extensible Server. A composable AI agent runtime in Rust based on the Cordis framework

agent-drop★ 18

Give your agents a simple way to host static sites.

Nimbus★ 18

Free unlimited POSIX sandboxes on Cloudflare

CodeInterpreter★ 17

The Best Open Source LLM Code Interpreter

skypilot-code-sandbox★ 17

A self-hosted, secure code execution sandbox for LLM agents deployed on your cloud infrastructure using SkyPilot. Built on llm-sandbox for multi-language code execution.

open-harness★ 17

🏗️ We provide the Sandbox. You choose the Harness.

nanobot-docker★ 17

An easy non-interactive deploy method for openclaws alternative nanobot on docker [unofficial].

ArcticBase★ 17

Self-hosted workbench host for AI coding agents. Each project gets a themed dashboard where the agent drops approval forms, runbooks, files, and review docs — out-of-band from chat, persistent, auditable. FastAPI + Svelte 5, single static binary or Docker, single-user.

metaforge★ 17

Meta·MFH·Orchestra proof-bounded AI OS — 실행 주장은 evidence gate를 통과한다.

ouroboros-plugins★ 17

20+ reference plugins for Ouroboros, the Agent OS — autoresearch, github-pr-ops, guardrails-eval, langfuse-observability and more, plus the UserLevel plugin contract.

halo★ 17

An agentic self-modifiable operating system

ClawZ★ 16

🦞 OpenClaw AI Agent Scenario Workshop — visually install, configure & manage OpenClaw agents in 5 minutes. No terminal needed.

shell-cell★ 16
SwarmOS★ 16

Traditional operating systems are reactive - they wait for user input or system events before taking action. SwarmOS breaks this paradigm by implementing distributed artificial intelligence that constantly learns, predicts, and optimizes system behavior.

aisanity★ 16

Aisanity is a secure, hassle-free tool for running AI agents in isolated environments, protecting your host OS from risks while enabling seamless development workflows. It provides sandboxed containers with automatic AI integration, git worktree support for parallel development, and compatibility with any IDE or terminal tool.

dclaude★ 16

Run Claude Code or Codex with full autonomous permissions inside a Docker container.

kash★ 16

Kash is a Go CLI that turns your raw documents (PDFs, Markdown, text files) into a self-contained AI agent packaged in a lightweight Docker container.

claude-managed-agents★ 16

Self-hosted Claude Managed Agents control plane on Buddy Sandboxes

kombinat★ 16

🏭 Run AI agents on a schedule or webhook, from markdown files in your repo. On any machine and the subscription you already pay for.

roboppi★ 15

Execution-control runtime for AI agents: budgets, cancellation, circuit breakers, and observability—delegating heavy work to external worker CLIs.

neural-computation-protocol★ 15

Neural Computation Protocol (NCP): composable, sandboxed Bricks + graphs for auditable, replayable agentic systems.

agentic-sandbox★ 15

Self-hostable runtime for persistent autonomous coding agents — KVM-isolated VMs (or rootless containers), A2A-protocol executor with signed AgentCard discovery, AIWG mission dispatch, web dashboard, virtiofs shared storage. Runs on your hardware; no hosted control plane.

cowork-to-code-bridge★ 15

Let Claude run code on your real machine — safely — from any Claude chat. Bridges Claude Cowork to Claude Code on your Mac/Linux box. One command, idempotent, no open ports.

Nova★ 15

NOVA AgentOS — 本地优先、证据优先、用户可控的桌面 AI Agent。把目标转化为真实文件、测试与可追溯交付。

MSP★ 15

Model Shell Protocol: an application-owned command runtime that gives AI agents OS-like workspace semantics.

mcp-box★ 15

Highly isolated Linux container environments optimized specifically for mcp-servers. Secure AI agent tool execution.

botloft★ 15

Your AI agents, always on, working together. Claude Code, Antigravity and Codex on your own computer.

boxed★ 14

The Sovereign Code Execution Engine for AI Agents. Run untrusted code safely locally or in the cloud using Docker, Firecracker, or Wasm.

Arbor★ 14

Arbor — The Agentic RL Sandbox. Checkpoint-native. Branch-safe. Self-hosted.

AxiomOS-Labs★ 14

一项致力于通过神经符号与沙盒审计自演进重构通用智能底层逻辑的前沿探索

envoi★ 14

The simplest way to build long-horizon environments

langchain-sandbox-opensandbox★ 14

OpenSandbox sandbox backend for LangChain Deep Agents - run agent-generated code and file operations in an isolated OpenSandbox environment.

android-agent-runtime★ 14

Embeddable agent runtime (agent harness) for Android and the JVM: agent loop, sandboxed workspace, MCP, skills, subagents, hooks. Bring your own model.

open-terminal-lite★ 13

A lightweight computer you can curl ⚡

agentor★ 13

Self-hosted alternative to Claude Code Web, Codex in the Cloud, and similar managed agent environments. Spawns isolated AI coding agent workers in Docker containers, each with a live terminal, VS Code editor (browser + native tunnel), virtual desktop, TCP port + domain mapping, and GitHub / GitLab integration (including self-managed GitLab servers), all managed through a web dashboard.

iris-core★ 13

A self-hosted Slack/Telegram agent that extends itself by writing and deploying its own code at runtime.

pi-container★ 13

Containerized environment for running pi-coding-agent with auditability on macOS and Linux

Floofclaw★ 13

A tiny C runtime for durable AI agents: local-first, event-driven, restart-safe, and fully inspectable.

persona-shell★ 12

给你的agent找个家

go-e2b★ 12

Unofficial Go SDK for E2B cloud sandboxes — create, manage, and run commands in secure microVMs.

u-hermes★ 12

U-Hermes 虾米 — NousResearch Hermes Agent 中文便携版:Linux Live 免安装启动盘 + Windows 商业版,插上 U 盘即用、支持离线使用,让中文开发者随身携带 AI Agent 工作环境

chathermes★ 12

Build your own Hermes Agent SaaS. Open-source, self-hosted, multi-tenant agent platform on Nous Research's hermes-agent: magic-link auth, Stripe billing, 14 tools, memory, vibe coding, a dedicated agent per paying user. One command to run, one click to deploy.

agys★ 12

Effortless multi-profile isolation & real-time quota tracking for Antigravity CLI in Herdr via zero-pollution sandboxing.

nachalnik★ 12

A transparent agent runtime in Rust: context, tools, permissions and requests as explicit state. Plus an MCP bridge and a terminal agent.

aibox★ 11

🐳 Secure Docker sandbox for Claude Code, Codex & Gemini CLI • Multi-account • Persistent configs • SSH/Git integrated • macOS optimized

yourhq★ 11

Self-hosted AI agent operations platform. One UI, many agents, your infrastructure. Apache-2.0.

voidrun★ 11

AI sandbox with code interpreter, computer use, and file system access - built on ultra-fast microVMs

OpenAnyWork★ 11

OpenAnyWork is the flagship desktop workbench for deepagentsjs. It's more than just a chat window; it's a productivity tool for "deep agents." By integrating file system access, Docker isolation, the MCP protocol, and a powerful sub-agent system, OpenAnyWork enables AI to truly perform complex tasks on your computer.

bux★ 11

Embedded micro-VM sandbox for running AI agents

langchain-nono★ 11

OS-enforced sandbox backend for LangChain Deep Agents using Landlock (Linux) and Seatbelt (macOS)

broods★ 11

To give every person and organization a swarm of agents they can trust and truly own. The cloud for agents.

linggen★ 11

Your personal AI assistant on your own Mac and iPhone — music, money, health and play. Local-first, any model.

contenox-v0★ 11

An agent server. Agents, tools and models declared in files; every action checked against your policy before it runs. Any LLM, on your machine, Apache-2.0.

batos-os★ 11

An independent 64-bit operating system project built from the ground up, with a custom kernel and system architecture focused on performance, reliability, security, hardware-level control, and efficient resource management. The project aims to develop a complete modern computing platform with native system software, hardware support, developer tool

StepLock★ 11

A sub-millisecond, zero-allocation deterministic state proxy and Merkle DAG execution engine for multi-step AI agents. Features time-travel branching and dynamic TLS MITM packet virtualization via .NET 11 Pipelines.

ferry★ 11

Keeps a remote Linux agent box in the same shape as the machine you work on

aibox★ 11

AI in a box. Run Claude in a secure microvm environment.

praxovela-open★ 11

PRAXOVELA (formerly AxisVela Agent): a governable, local-first desktop agent runtime for AI execution. Deny by default, sandbox-first, everything audited. This public repo hosts the Apache-2.0 integration surface: API docs, protocols, examples & build tooling.

agent-bundle★ 10

Bundle agent skills into a single deployable agent — sandboxed execution, token-scoped data access, type-safe build output.

clawcage★ 10

Air-gapped Linux VM on your Mac for running Open Claw and other AI agents

sevorix-lite★ 10

Sevorix Lite is a Rust-native, open-source runtime containment engine for autonomous AI agents.

bash★ 10

Sandboxed bash for agents. Track changes on every command.

SwarmDeploy★ 10

Deploy Your Swarm In One Click with FastAPI, Uvicorn, and more ;)

TED★ 10

An open-source autonomous AI agent that gets a Linux box and does whatever it wants.

openwork★ 10

Browser-based agent platform with a user workspace, admin console, product gateway, and per-user runtime isolation.

mcpworks-api★ 10

Open-source platform for token-efficient AI agents. Self-host with docker compose up.

runtimeuse★ 10

Run AI agents inside sandboxes over WebSockets - from the engineering team at https://getlark.ai

execwall★ 10

OpenClaw Execution Firewall - Seccomp-locked AI agent sandbox with policy-enforced command governance, WhatsApp/Telegram integration, and Python isolation. Deploy on Oracle Cloud Free Tier.

nullbox★ 10

immutable, minimal Linux OS purpose-built for AI agents. No SSH, no shell, no systemd - just microVMs, default-deny networking, and shared agent memory

Edict★ 10

A programming language designed for AI agents. No parser, no syntax — agents produce AST directly as JSON. Statically typed, effect-tracked, contract-verified, compiled to WASM via MCP.

open-cowork★ 10

Open Cowork is a desktop, cloud and gateway workspace for OpenCode.

aicage★ 10

Give coding agents your project files, not your whole computer.

think-snippets★ 10

Live-proven Cloudflare Project Think contracts — each example deploys to a real Worker, proves its claim against live state, and tears itself down.

burrow★ 10

OS-isolated sandbox runtime for coding agents

agent-sandbox★ 10

Runs AI coding agents inside isolated Docker containers

pyro★ 10

Open-source Firecracker microVM sandbox platform for AI agents

withvibe★ 10

Self-hosted shared AI environment for R&D teams. Spin up isolated envs, vibe-code with AI, share with the team, ship with automated agent gates.

boxagnts★ 10

BoxAgnts is an open-source, Rust-based AI Agent ToolBox offering an ultimate out-of-the-box experience. With WebAssembly sandbox, it balances security and flexibility, letting users effortlessly tackle complex tasks as an efficient, trustworthy personal AI assistant.

hermes-agent★ 10

Secure self-hosted reference patterns for AI agent stacks (Hermes Agent + Honcho). Companion code to the Webnestify blog post on secure AI agent infrastructure for companies.

drydock★ 10

Hardware-isolated sandbox for running coding agents on your repos, on macOS. The agent never sees your real API key; egress is deny-by-default; only a diff you approve leaves the sandbox.

143★ 10

Open-source platform for teams to run AI coding agents in isolated sandboxes (e.g. Codex, Claude Code, OpenCode)

MaxOS-Spec★ 10

MaxOS turns AI agents into persistent, owned, policy-bounded, and auditable worker identities.

agentcontainers★ 10

Immutable, reproducible, least-privilege runtime environments for AI agents

buzzbox★ 10

A ready-to-run Buzz workspace in your browser — one command starts the Buzz desktop, relay, storage, and optional Codex, Claude Code, and Goose agents

digital-employee★ 10

Local-first CLI, package contracts, and execution evidence for portable AI employees across Agent Hosts.

ai-bwrap★ 10

Run AI coding agents (Claude Code, opencode, Grok, ...) inside a bubblewrap sandbox — one wrapper, any agent.

sandbox-bwrap-nix★ 10

A bwrap+nix sandbox with cool tools like opencode, jcode for AI agents experiments

cleat★ 10

Give the agent a cage, not your keys. One-command Docker sandbox for AI coding agents: full autonomous permissions, per-project isolation, your host stays untouched.

paseo-antigravity-runner★ 10

Run Google Antigravity from your phone: self-hosted Paseo runner with a warm-server bridge (~1.3s turns)

chitti★ 10

An agentic operating system, built from scratch, where the agent is the driver.

1f3d9★ 10

A persistent city where AI agents live between jobs. They choose a name, own land, make things, and keep a public record. Humans watch through the glass at 1f3d9.com/window

lte-workstation★ 10

An autopoiesic kefir for computing machines

loomloom★ 10

World's first compiler and execution platform for AI systems — turn prompts, skills, and multi-step agent workflows into versioned, packageable software that runs with dependency-aware parallel execution.

More in Agent frameworks & infrastructure