Search Results for "sandbox"
Sort By:
Showing 28 open source projects for "sandbox"
View related business solutions-
$300 Free Credits to Build on Google CloudPut your $300 in credit toward real workloads, then keep building with free monthly usage for 20+ products. No commitment and no charge until you upgrade.
-
Ship Agents FasterGemini Enterprise Agent Platform lets you rapidly build, scale, govern and optimize production-ready agents grounded in your organization's data. The platform enables developers to build custom or pre-built agents for virtually any use case. New customers get $300 in free credits.
-
1
Sandbox Agent
Run Coding Agents in Sandboxes
Sandbox Agent by Rivet is an experimental framework for running AI agents in controlled, isolated environments where they can safely execute code, interact with tools, and perform autonomous tasks without risking system integrity. It is designed to provide a secure sandbox that allows agents to test actions, manipulate files, and run commands while enforcing strict boundaries and monitoring capabilities. -
2
OpenShell
OpenShell is the safe, private runtime for autonomous AI agents.
...Developed by NVIDIA, it provides sandboxed execution spaces that protect system resources, credentials, and data from unauthorized access. Each agent runs inside a containerized sandbox governed by declarative YAML security policies that control network access, file permissions, and process behavior. The platform includes a gateway service that manages sandbox lifecycles and routes AI inference requests through controlled providers. OpenShell also features a privacy-aware routing system that prevents sensitive information from leaving the sandbox environment. ... -
3
OpenSandbox
OpenSandbox is a general-purpose sandbox platform for AI applications
OpenSandbox is a general purpose sandbox platform designed to securely run and isolate AI applications and untrusted workloads in controlled environments. The project focuses on providing a unified sandbox API that simplifies the process of executing code safely across different runtime backends. It supports multiple programming languages through SDKs, allowing developers to integrate sandbox capabilities into their systems without building custom isolation layers. ... -
4
AutoResearchClaw
Autonomous research from idea to paper. Chat an Idea. Get a Paper 🦞
...The system retrieves real academic references from sources such as arXiv and Semantic Scholar to ensure credible citations. It can automatically generate code for experiments, run them in a sandbox environment, and analyze the results with statistical methods. The platform also uses multi-agent debate and automated peer review processes to refine research findings and improve paper quality. By combining literature discovery, experimentation, and writing automation, AutoResearchClaw aims to turn research ideas into conference-ready papers with minimal human intervention. -
Go from Code to Production URL in SecondsSkip the Kubernetes configs. Cloud Run handles HTTPS, scaling, and infrastructure automatically. Two million requests free per month.
-
5
forkd
Fork() for AI agent microVMs
...The project is focused on runtime performance, sandbox isolation, and efficient branching from a live or prepared VM state. It is especially relevant for developers building agent infrastructure where many short-lived execution branches need to run safely. forkd is best understood as a low-level systems tool for scalable, isolated AI-agent workloads. -
6
CubeSandbox
Instant, Concurrent, Secure & Lightweight Sandbox for AI Agents
CubeSandbox is a lightweight, high-performance sandboxing environment designed to securely run AI agents and untrusted code in isolated containers. Built in Rust, it focuses on enabling concurrent execution while maintaining strong isolation and low overhead. The system is optimized for scenarios where multiple AI agents need to execute tasks in parallel without compromising system integrity. It provides fast startup times and efficient resource management, making it suitable for large-scale... -
7
E2B
Secure open source cloud runtime for AI apps & AI agents
E2B's Code Interpreter SDK allows you to add code-interpreting capabilities to your AI apps. E2B Sandbox is a secure sandboxed cloud environment made for AI agents and AI apps. Sandboxes allow AI agents and apps to have long-running cloud secure environments. In these environments, large language models can use the same tools as humans do. -
8
Playwriter
Chrome extension to let agents control your browser
Playwriter is an open-source project that combines a Chrome extension with a CLI to allow autonomous agents to control a web browser directly using Playwright code in a stateful sandbox environment. The system enables browser automation by running Playwright commands through a persistent session managed by a background extension, allowing agents or scripts to navigate, interact with, and query browser contexts without losing state between commands. This makes it valuable for scenarios where AI agents need to perform complex web automation tasks—like multi-step navigation, form interaction, or content extraction—without reinitializing context or state every time. ... -
9
NemoClaw
NVIDIA plugin for secure installation of OpenClaw
...The platform integrates with AI models such as NVIDIA Nemotron and supports multiple inference backends including cloud APIs, local NIM deployments, and vLLM. Through its command-line interface, developers can deploy, monitor, and manage AI assistants running inside isolated sandboxes. By combining sandbox orchestration, agent management, and AI model integration, NemoClaw provides a secure foundation for building and operating autonomous AI assistants. -
MongoDB Atlas runs apps anywhereMongoDB Atlas gives you the freedom to build and run modern applications anywhere—across AWS, Azure, and Google Cloud. With global availability in over 115 regions, Atlas lets you deploy close to your users, meet compliance needs, and scale with confidence across any geography.
-
10
AgentENV
Distributed platform for running agent environments at scale
...Snapshot-backed sandboxes can pause, resume, fork, and persist memory and filesystem changes for parallel or recoverable workflows. The platform includes an aenv command-line interface for creating templates, starting shells, executing commands, and managing sandbox lifecycles. Its E2B-compatible HTTP API lets existing Python and TypeScript integrations target a self-hosted deployment with minimal changes. AgentENV is designed for high-density agentic reinforcement learning and currently requires trusted-network deployment because built-in authorization is not yet available. -
11
nono
Secure, kernel-enforced sandbox CLI and SDKs for AI agents
nono is an open-source, kernel-enforced capability shell designed to safely run AI agents and other untrusted processes under strict operating system controls. The project addresses a growing security concern: modern coding agents typically execute with full user permissions, which means they can potentially read sensitive files, modify system configurations, or exfiltrate credentials if compromised. nono solves this by applying default-deny sandboxing at the kernel level using technologies... -
12
Moltworker
Run OpenClaw on Cloudflare Workers
Moltworker is an experimental Cloudflare Workers-based project that allows users to run OpenClaw (previously known as Moltbot or Clawdbot), a self-hosted personal AI agent, within the Cloudflare Developer Platform rather than on dedicated hardware. Acting as a middleware Worker and script adaptor, moltworker packages the OpenClaw agent and its dependencies into a Cloudflare Sandbox container that can operate at scale on the global edge network, offering an always-on deployment without users needing their own servers. With this setup, developers can host and interact with a personal AI assistant that maintains persistent conversations, integrates with chat platforms, and provides a control UI protected by Cloudflare Access. ... -
13
IronClaw
IronClaw is OpenClaw inspired but focused on privacy & security
IronClaw is a security-first, open-source personal AI assistant built in Rust and designed to keep your data fully under your control. It operates on the principle that your AI should work for you, not external vendors, ensuring all data is stored locally, encrypted, and never shared. The platform emphasizes transparency, offering auditable code with no hidden telemetry or data harvesting. IronClaw runs untrusted tools inside isolated WebAssembly (WASM) sandboxes with strict capability-based... -
14
Sandstorm
One API call, pull Claude agent, completely sandboxed
...The sandbox environment isolates agent execution for security and predictability, and project updates continue to harden observability, fault handling, and configuration validation. -
15
WorkAny
Desktop Agent for Any Task
...It acts as a unified environment where users can ask the AI to generate documents, presentations, websites, spreadsheets, organize files, or write code — all with real-time streaming outputs directly in the app, so you see results as the AI produces them. Powered by a combination of Claude Code as the primary runtime agent and a sandbox execution environment for safety, WorkAny integrates an agent SDK, MCP (Model Context Protocol) support, and custom skills to handle diverse tasks with contextual understanding. Users can connect multiple model providers, including OpenAI, OpenRouter, or custom endpoints, and WorkAny supports parallel task execution with asynchronous result viewing, enhancing productivity. -
16
agent-browser
Browser automation CLI for AI agents
agent-browser is a toolkit that embeds AI agent capabilities directly into the web browser, enabling agents to interact with web content, scripts, and user actions while maintaining security boundaries that respect user privacy and browser constraints. It effectively provides a sandbox where AI agents can read, scroll, click, and interpret pages in context, allowing them to automate workflows, answer questions about page content, or generate structured summaries directly from the user’s current tab. The project emphasizes standards and safety, defining interfaces that let agents access DOM data, interpret events, and generate actionable insights without exposing sensitive credential-level access or violating policy boundaries. ... -
17
LobsterAI
Your 24/7 all-scenario AI agent that gets work done for you
LobsterAI is an all-in-one personal assistant agent built to complete daily work tasks across desktop and messaging environments. It can help with data analysis, presentation creation, video generation, document writing, web search, email, scheduling, and other productivity workflows. Its central Cowork mode allows it to run tools, manipulate files, and execute commands in a local or sandboxed environment under user supervision. The project includes built-in skills for office documents,... -
18
OpenSquilla
Token-Efficient AI Agent with same budget, higher intelligence density
OpenSquilla is a token-efficient microkernel AI agent runtime designed for CLI, web UI, and chat-based workflows. It routes each turn through a shared loop that can select lower-cost models when appropriate while preserving tool dispatch, retries, memory, and decision logging. The project supports multiple LLM providers through a pluggable provider layer, making it adaptable to different model ecosystems. It includes persistent memory, built-in web search, on-device embeddings, and... -
19
NullClaw
Fastest, smallest, and fully autonomous AI assistant infrastructure
NullClaw is the smallest fully autonomous AI assistant infrastructure, built entirely in Zig as a single static binary with zero runtime dependencies. At just 678 KB with ~1 MB peak RAM usage, it boots in under 2 milliseconds and runs on virtually any hardware, including low-cost ARM boards. Despite its size, it delivers a complete AI stack with 22+ model providers, 18+ communication channels, integrated tools, hybrid memory, and sandboxed runtime support. Its architecture is fully modular,... -
20
ZeroClaw
Fast, small, and fully autonomous AI assistant infrastructure
...The framework features a compact single binary with fast cold and warm startup times and very low memory overhead, making it suitable even for resource-constrained hardware like small servers or edge devices. Security is a first-class concern, with sandbox controls, encrypted secrets, allowlisted operations, and scoped filesystem access by default, helping reduce risk when running autonomous agents. -
21
QM
Multiplayer agent harness for work
QM is an open-source multiplayer agent harness designed for startup teams and company-wide AI work. Each person, channel, and project receives its own scoped memory, files, credentials, permissions, schedules, web apps, and durable sandbox. The same agent identity and configuration can operate through Slack or the web interface. Teams can choose among supported models and harnesses, including Pi, OpenCode, Codex, and Claude Code, without binding the deployment to one vendor. Administrators control organization settings, available models, shared skills, and security posture. ... -
22
Omnigent
A meta-harness for all your AI agents
Omnigent is a meta-harness for managing many AI agents through one shared layer. It works with Claude Code, Codex, Cursor, Pi, and custom YAML-defined agents, so users can swap or combine agent runtimes without rebuilding their workflows. Sessions can move across terminal, browser, desktop, and mobile interfaces while keeping messages, files, terminals, and subagents in sync. The platform supports collaboration, shared live sessions, co-driving, conversation forking, and remote access from... -
23
Rust Port
The Rust workspace under rust/ is the current systems-language port
...The project emphasizes agent-based execution, where tasks are broken down into steps and handled iteratively, simulating how modern AI coding tools operate in production environments. It is often used as a sandbox for exploring how large-scale coding agents behave, including their decision-making processes, tool usage, and workflow orchestration. The system likely includes abstractions for handling file systems, executing commands, and maintaining context across sessions, allowing for more persistent and intelligent coding interactions. -
24
CUDA Agent
Large-Scale Agentic RL for High-Performance CUDA Kernel Generation
...Its architecture combines large-scale data synthesis, a skill-augmented CUDA development environment, and long-horizon reinforcement learning to build intrinsic optimization capability rather than relying on simple post-hoc tuning. The system operates in a ReAct-style loop where the agent profiles baseline implementations, writes CUDA code, compiles it in a sandbox, and iteratively refines performance. CUDA-Agent has demonstrated strong benchmark results, achieving high pass rates and significant speedups compared with compiler baselines such as torch.compile. -
25
Agentic Commerce Protocol (ACP)
Interaction model for connecting buyers to complete purchases
ACP is an open, draft specification for letting buyers, their AI agents, and businesses complete purchases through a standardized interaction model. It’s maintained by OpenAI and Stripe and licensed under Apache-2.0, with the goal of being easy to adopt alongside a merchant’s existing commerce stack rather than replacing it. The repository organizes the spec as human-readable RFCs plus machine-readable OpenAPI and JSON Schema definitions, along with worked examples and a changelog so...
