Agent sandbox

Bubblewrap is a Linux sandbox runner that creates lightweight, isolated execution environments for running untrusted applications. It combines Linux user, mount, network, PID, and UTS namespaces with seccomp-BPF system call filtering to restrict filesystem, network, process, and inter-process communication access. The project provides comprehensive process isolation by giving each sandbox its own private tmpfs root with selective bind-mounts, a separate network stack containing only a loopback interface, an independent process ID space, and remapped user and group identifiers. It applies secc

Microsandbox is a runtime for creating and managing lightweight, hardware-isolated virtual machines — called sandboxes — that boot directly from standard OCI container images. Each sandbox runs as its own host process with a separate kernel, filesystem, and network stack, providing process-per-sandbox isolation. The project includes a command-line tool and multi-language SDKs (Rust, TypeScript, Python, Go) for programmatic lifecycle control, and it communicates with sandbox agents over Unix sockets using a CBOR-encoded protocol. What distinguishes Microsandbox is its combination of host-manag

E2B is a cloud-based infrastructure platform designed to provide secure, isolated execution environments for code and shell commands. It functions as an ephemeral orchestrator that provisions lightweight virtual machines, allowing developers and autonomous agents to run untrusted processes within a sandbox that is completely separated from the host system. The platform distinguishes itself through its focus on programmable, serverless workspaces that support the full lifecycle of cloud-based development. By utilizing hardware-level isolation and snapshot-based resumption, it enables the near-

Dify-sandbox is a secure runtime environment designed for the execution of untrusted code snippets. It functions as a containerized sandbox that isolates processes from the host operating system, ensuring that arbitrary scripts can be run without granting them unauthorized access to sensitive data or critical system resources. The project distinguishes itself through a multi-layered security approach that combines kernel-level isolation with strict resource management. By utilizing Linux namespaces and container-based process isolation, it partitions system resources to maintain visibility bo

Monty is a sandboxed execution environment designed primarily for running Python code generated by AI models. It provides a secure, isolated runtime that blocks host access, enforces resource limits, and supports pre-execution type checking against built-in type hints to catch signature mismatches before code runs. The sandbox can persist its interpreter state at external function calls, allowing execution sessions to be serialized, stored, and later resumed from a file or database. What distinguishes Monty is its combination of stateful, resumable execution with multi-language native embeddi

Judge0 is an online code execution engine and multi-language compiler API designed to compile and run source code within isolated sandboxes. It functions as an asynchronous job processor that handles code submissions via a queue and provides a secure environment to run arbitrary programs while preventing unauthorized system access. The system distinguishes itself through a multi-stage compilation pipeline and a flexible execution model that supports both single-file submissions and multi-file program execution via archives. It employs an isolate-based sandboxing mechanism to enforce strict ha

container-use is a containerized AI execution environment and code sandbox designed to provide a secure space for AI coding agents to execute commands and build applications. It functions as a workspace orchestrator that provisions isolated containers mapped to git branches, allowing multiple agents to operate in parallel without state conflicts or affecting the host system. The project serves as a Model Context Protocol server, bridging AI agents to containerized environments for standardized tool access. It enables a workflow for reviewing and merging changes made by agents within these iso

The sandbox-sdk is a development kit designed for building secure, isolated execution environments on a global edge network. It provides a framework for creating ephemeral, containerized workspaces that allow developers to run untrusted code, manage build tasks, and host automated scripts without compromising host system security. By leveraging a serverless runtime, the platform enables the deployment of these environments directly at the network edge to ensure low-latency performance. The platform distinguishes itself by integrating language models with sandboxed execution, facilitating the

OpenHands is an autonomous agent framework designed for software engineering workflows. It provides a modular platform for orchestrating AI agents that reason, plan, and execute tasks within isolated, containerized development environments. By integrating with standard version control and development tools, the system enables agents to autonomously navigate codebases, implement features, and resolve issues through iterative reasoning and tool execution. The platform distinguishes itself through a model-agnostic orchestrator that connects diverse language models to a unified tool registry. It

Cua is an agent benchmarking and desktop automation platform designed to evaluate autonomous agents and execute repetitive tasks within isolated, virtualized environments. It provides a framework for provisioning consistent workspaces and measuring agent performance against standardized desktop operations. The platform distinguishes itself by integrating virtual machine orchestration with headless interaction capabilities. By leveraging hypervisor-based virtualization, it runs operating systems at near-native speeds, while its automation layer injects commands directly into application proces

Nanobot is an orchestration framework designed for building, deploying, and managing autonomous AI agents. It provides a secure runtime environment that supports persistent memory, multi-step workflow management, and tool integration, allowing agents to maintain context and state across long-running tasks. The platform distinguishes itself through a unified model gateway that normalizes requests across diverse local and remote language models, alongside a multi-channel integration layer that connects agents to various messaging platforms. It enforces security through containerized sandboxing

ClawWork is a suite of tools designed to monitor agent finances, provide isolated execution environments, simulate economic behaviors, and benchmark performance. It functions as an autonomous agent sandbox where AI agents can run code and generate professional business deliverables. The project focuses on the financial sustainability of AI assistants through an economic simulation environment. This includes tools for tracking token expenditures and income generation, as well as simulations that analyze the trade-offs between immediate earnings and long-term skill acquisition. The system incl

Open Interpreter is an autonomous agent runtime that translates natural language instructions into executable code to interact with local software and operating systems. It functions as an orchestration framework that connects language models to a secure execution environment, enabling the development of agents capable of managing system resources and performing complex tasks. To ensure safety, the system mandates explicit user verification before executing any generated code and provides robust isolation through containerized sandboxing. The project distinguishes itself through its deep inte

This project provides secure, containerized infrastructure designed for autonomous agents, remote code execution, and cloud development. It functions as a sandboxed environment where AI agents and external processes can execute code, run shell commands, and manage files while remaining isolated from the host system. The system distinguishes itself by implementing the Model Context Protocol, allowing it to act as a standardized tool server that exposes browser and filesystem capabilities to compatible clients. It further integrates headless browser automation, enabling programmatic web navigat

This project is a Python framework for building autonomous, event-driven agent systems. It provides a unified runtime for orchestrating multi-agent workflows, managing persistent conversation state, and executing code within secure, isolated sandbox environments. The framework is designed to handle complex task delegation, allowing agents to invoke other agents as tools while maintaining context across multi-turn interactions. The framework distinguishes itself through its deep integration with the Model Context Protocol, enabling agents to connect to external data sources and remote services

Suna is an orchestration platform designed for the deployment, management, and governance of autonomous AI agents. It provides a centralized system for defining agent behaviors and tool integrations, enabling the automation of complex business processes through a unified interface. The platform distinguishes itself by applying infrastructure-as-code principles to AI, utilizing version-controlled repositories to manage agent configurations, skills, and guardrails. It ensures secure and predictable operations by spawning ephemeral, isolated virtual machines for every individual task, preventing

Sandbox Agent is a platform designed to manage, secure, and orchestrate autonomous coding assistants. It provides a standardized infrastructure for executing untrusted code and managing agent lifecycles within isolated, containerized environments. By decoupling agent execution from client connections, the platform ensures that session states remain persistent across process restarts and network interruptions. The project distinguishes itself through a capability-based security model that enforces granular permission checks on tool usage, ensuring that autonomous processes operate within defin

This project provides a secure, containerized execution engine designed to run untrusted code within isolated environments. It functions as a library for integrating code interpretation into autonomous agents and intelligent assistant workflows, ensuring that host systems remain protected while enabling dynamic data processing and file manipulation. The platform distinguishes itself through a multi-backend architecture that abstracts diverse container runtimes, allowing for flexible deployment and automated backend failover. It supports interactive, multi-turn workflows by maintaining persist

Acontext is an LLM orchestration backend and agent memory framework designed to manage session state and knowledge for AI agents. It functions as a context manager and orchestration layer that integrates model providers with a secure code sandbox and a zero-knowledge data store. The project is distinguished by its approach to knowledge distillation, capturing agent learnings as reusable Markdown skills and structured memory files. It provides a secure execution environment where shell commands and scripts run in isolated containers with the ability to mount these persistent skill files direct