As artificial intelligence takes on a growing share of software development, the infrastructure supporting coding agents is becoming just as important as the models powering them. Dutch startup Boxd has raised $2 million in pre seed funding to develop cloud computing infrastructure that allows developers and autonomous AI agents to run code, execute tests and manage software workflows in persistent virtual machines.
The funding round was led by BlueYard Capital, with participation from OVNI Capital, Antler, S20 and Script Capital. Business angels from General Intuition and other technology companies also joined the round.
Boxd was founded by Michiel Voortman, Laurentiu Ciobanu and Hidde Kehrer. The company plans to use the new capital to expand its team and further develop its custom virtualisation engine.
Infrastructure for Autonomous Development
AI coding agents can perform tasks such as writing software, debugging code, running tests and exploring alternative solutions. However, these systems often require dedicated computing environments that can remain active for extended periods and support several agents working simultaneously.
Traditional development infrastructure generally relies on individual virtual machines or containers. While these approaches are widely used, Boxd argues that they can become difficult to manage when developers operate multiple agents in parallel.
Running autonomous agents also creates security concerns. An agent that executes code without continuous human supervision may interact with files, systems and services in ways that introduce new risks. Boxd is building its infrastructure to address both the operational and security challenges associated with this new development model.
Persistent Virtual Machines
Boxd’s platform enables teams to create as many cloud machines as they need for development and testing. These machines are designed to provide persistent environments in which code, processes and workflows can continue running.
A key feature of the platform is its support for snapshotting and live forking. Developers and AI agents can create copies of running machines, including their memory and active connections, in under 100 milliseconds, according to the company.
This capability allows multiple agents to work from the same starting environment while testing different approaches in parallel. Instead of setting up a separate machine for every experiment, teams can create copies of an existing environment and continue development from that point.
The approach is intended to reduce configuration work, improve testing speed and make it easier to manage complex agent based workflows.
A Simplified Virtualisation Model
Boxd’s architecture focuses on the core resources needed for software development, including CPU, memory, storage and networking. It does not attempt to emulate additional hardware such as audio devices, webcams or USB ports.
The company says this simplified design reduces the software attack surface and can help mitigate security risks when autonomous agents are given access to computing environments.
Boxd co founder and chief technology officer Hidde Kehrer said AI coding agents need access to real computers rather than disposable sandboxes. He argued that persistent, hardware isolated virtual machines can allow developers and agents to build, test and deploy software more safely while preserving the state of ongoing processes.
Funding Product Development
The $2 million pre seed investment will support Boxd’s next stage of development. The company plans to hire additional employees and continue building its custom virtualisation engine.
As AI coding agents become more capable and are increasingly used to perform multiple software tasks at once, Boxd is positioning its infrastructure as a foundation for persistent, secure and scalable development environments. Its focus is on giving both human developers and autonomous agents the computing resources needed to experiment, test and execute code without repeatedly rebuilding their environments.
