The Cloud for AI Agents

Hopx provides fully isolated Linux micro-VMs that spin up in ~100 ms, ideal for running AI agents, notebooks, or untrusted code safely at scale. Each sandbox uses Firecracker microVMs for hardware-level security and kernel isolation, offering full state persistence and SDKs for Python, JS, Go, .NET, Java, and PHP. It features no cold-start lag and no runtime limits, ensuring raw performance and security for AI-driven workloads and dynamic applications.

How to Use

Users can interact with Hopx via its SDKs available for Python, JavaScript/TypeScript, Go, .NET, Java, and PHP. The process involves installing the SDK (e.g., npm install @hopx-ai/sdk), creating a sandbox instance (e.g., Sandbox.create({ template: 'code-interpreter' })), executing code within the sandbox using methods like runCode(), and then cleaning up the sandbox. The SDK allows for various operations including code execution, real-time output streaming, file operations, shell commands, process management, and desktop automation.

Key Features

• Blazing-fast sandboxes with ~100ms startup
• Hardware-level security and kernel isolation via Firecracker microVMs
• Unlimited runtime with full state persistence
• SDKs for Python, JavaScript, Go, .NET, Java, PHP
• Execute code, stream output, file operations, commands, process management
• Desktop automation and real-time metrics monitoring

Use Cases

• Run Untrusted Code safely in isolated microVMs
• Launch and manage AI Agents that write and execute code
• Spin up Jupyter notebooks for Data Analysis
• Develop and run Deep Research Agents continuously
• Orchestrate Autonomous AI Workflows with code execution and tools
• Control full cloud desktops with Computer Use Agents
• Run Background Automations, workers, and schedulers
• Train and evaluate Reinforcement Learning agents
• Host Secure MCP Servers in isolated perimeters
• Execute Long-Running Jobs without artificial runtime limits
• Coordinate Multi-Agent Mesh across isolated micro-VMs