Phaze raises $8.75M to build remote workstations for AI agents

A new model for shared compute

Phaze, a new startup based in Utah, recently secured $8.75M in seed funding to develop a platform for remote workstations. The company was founded by the team behind Parsec, a well-known remote desktop firm that was acquired by Unity for $320M in 2021. This new venture aims to change how businesses manage the hardware required for high-intensity digital work.

The core idea behind Phaze is that a company should not have to buy a top-end workstation for every single employee. Instead, the platform allows organizations to pool their most powerful machines into a shared resource. This means that high-performance hardware can be distributed to whoever needs it at that moment, rather than sitting idle on a specific desk.

By moving toward a shared model, companies can significantly reduce the amount of money they spend on hardware. The goal is to ensure that the computer sitting on a person's desk does not dictate the level of computing power they can access. As chief executive Benjy Boxer stated, "The computer sitting on your desk shouldn’t determine what compute you can access."

How the technology works

The Phaze platform is designed to support both human workers and AI agents who need to perform demanding tasks. For example, designers and engineers can use the remote workstations to run complex 3D and design software from any location. This allows them to work on high-fidelity projects without needing to carry around heavy, expensive hardware.

For AI agents, the system provides a way for them to interact with a machine's screen and various applications. The agents connect through MCP, which is a common standard used for linking AI models to different tools. While Phaze currently labels this specific integration as experimental, it provides a framework for agents to perform actions in a visual environment.

The security of this setup is managed by ensuring that agents only have access to the specific tools and permissions granted to their human users. This creates a controlled environment where an AI can perform tasks on behalf of a person without having unrestricted access to the entire corporate network. It mimics the way a human would interact with a desktop computer.

Practical use cases for agents

To demonstrate how this works in a real-world setting, Phaze showcased a Slack bot capable of handling IT support requests. When a user submits a request, the bot can log into the user's computer to troubleshoot issues. If the bot encounters a problem it cannot solve, it is designed to call in a human to take over the session.

This type of automation is particularly useful for tasks that require a visual interface or interaction with specific software. Because the AI agent can control the screen, it can navigate menus and click buttons just like a human would. This makes it possible for agents to handle routine maintenance or software troubleshooting that was previously difficult to automate.

The demand for these types of agentic capabilities is already visible in the market. For instance, the demand for AI agents was high enough to help sell out Apple’s Mac mini in the United States this past spring. Phaze is positioning itself to provide the infrastructure that makes these agents functional in a corporate environment.

The broader landscape of agent tools

Phaze is entering a growing market where multiple companies are building environments for AI agents to operate within. Other players in this space are exploring different ways to give agents more autonomy and better tools. For example, Baseten recently acquired Blaxel, a company that specializes in creating sandboxes for AI agents to work in safely.

Other companies are also looking for ways to keep AI agents active and connected to hardware. Perplexity has explored ways to turn a Mac mini into a device that can run an always-on agent. These different approaches show that there is a significant push to move AI from simple chat interfaces to systems that can actually perform work.

By focusing on remote workstations, Phaze is targeting the high-end needs of professional workflows. While sandboxes and dedicated hardware are one way to go, providing a remote desktop for agents allows for a more flexible and scalable approach. It enables companies to deploy agents across various tasks without needing to manage a fleet of dedicated physical machines.

What to expect next

With the $8.75M investment from a16z and other firms like Indie and Makers Fund, Phaze is set to begin scaling its platform. The company will likely focus on refining the MCP integration to make it more stable and reliable for enterprise use. As the technology moves out of the experimental phase, more companies may begin to adopt the shared workstation model.

The success of Phaze will depend on how well it can balance the needs of human users with the requirements of AI agents. If the platform can successfully reduce hardware costs while providing a seamless experience for designers and engineers, it could become a standard for modern offices. It represents a shift toward a more fluid way of thinking about where and how work actually happens.

In the coming months, we will see more demonstrations of how these agents interact with complex software. As companies look for ways to increase productivity, the ability to hand off tasks to an agent that can 'see' and 'use' a computer will be a key factor. Phaze aims to be the underlying infrastructure that makes that transition possible for the average business.