Agentic cloud infrastructure startup E2B has raised $21 million in early-stage funding to build out an entirely new, open-source infrastructure for running artificial intelligence agents securely in the cloud. E2B’s vision is to provide businesses with a dedicated, open-source cloud infrastructure stack for AI agents. It believes that the best place for AI agents to be hosted is in secure sandboxed environments in the cloud, where it will provide them with safe computer and browser-use features to automate various complex business tasks. Co-founder and Chief Executive Vasek Mlejnsky explained that enterprises have massive expectations for AI agents due to all the hype around them, but he believes that they won’t easily scale if they’re built on legacy infrastructure that’s not designed to be used by them. “E2B solves this by equipping AI agents with safe, scalable and highly-performant cloud infrastructure, as well as tools that help agents to scale in production,” he added. With E2B’s sandboxed cloud environments, AI agents get access to the same computational capabilities as human workers do, meaning their own computer plus a browser and other tools that enable them to retrieve information, as well as a file system to store that data, and compute platforms for executing AI-generated code. Companies can quickly spin up and shut down these sandboxed cloud environments as needed, and they can scale rapidly to millions of virtual personal computers, allowing enterprises to unleash armies of agents that can work even more efficiently than humans do. It’s all hosted in a simple and secure runtime that enables reinforcement loops for AI training and agentic workflows to run rapidly. E2B aims to position its open-source sandbox protocol as the universal standard for AI agents, and has plans to add additional functionality such as “secrets vaults” and orchestration tools for managing fleets of AI agents from a single console.