Phaze Raises $8.75M Seed Led by a16z for Human and AI Agent Remote Access

Phaze has raised $8.75 million in seed funding led by a16z to give employees and AI agents access to powerful remote computers. The company announced the round on November 4, with participation from Indie, NextView, Makers Fund, and Lerer Hippeau.

The Utah-headquartered startup plans to expand engineering, develop the product, and support customer growth. Its founders previously built Parsec, the remote-access business acquired by Unity in 2021. Their new venture combines a high-performance remote desktop with tools for managing workstations as shared infrastructure.

The investment targets a practical question behind the rise of computer-using AI: where should an agent actually do its work? Phaze’s answer is a company-managed machine that an employee or agent can reach remotely, with access governed by the organization’s existing permissions.

Remote Desktop Meets the AI Workforce

Phaze starts with a familiar requirement: making a workstation somewhere else usable from the device in front of you. Its website emphasizes responsive input, 3D graphics, color quality, and multiple monitors for workloads such as CAD and design. Those are performance claims from the company, rather than results from an independent benchmark.

The broader proposition is that remote access should serve people and their agents through the same infrastructure. A designer might use a workstation interactively, while an agent connects to another machine to test software or monitor a long-running process. Work can continue without the agent occupying the employee’s active desktop.

That separates the device someone uses to connect from the hardware that executes the task. It also gives IT a way to assign access to existing machines rather than treat every new agent as a reason to purchase another dedicated workstation.

In his earlier interview with Unite.AI, Boxer pointed to better connectivity and hardware video encoding and decoding as foundations for modern remote desktop performance. He also highlighted software testing and end-user troubleshooting as tasks where operating a full computer can be more useful than calling an application API.

How Phaze Connects Agents to Computers

The company’s AI product page describes a local Model Context Protocol (MCP) server through which agents can operate remote computers using the desktop or command line. Phaze currently labels the MCP integration experimental.

This makes Phaze an access layer rather than a new reasoning model. The agent supplies the intelligence that decides what to do; Phaze supplies the connection to a machine and its applications. The distinction matters because giving an agent a computer does not, by itself, establish that it can complete every task correctly.

It also differs from integrating an agent with one application’s API. A computer interface can expose software workflows that do not have a purpose-built integration. As Unite.AI explains in its guide to computer-using agents, screen-based interaction expands the range of software an agent can reach while making permissions and reliable execution especially important.

A Support Bot Shows the Human Handoff

Phaze’s IT support demonstration offers a concrete example. The team built a Slackbot that responds to a request for help, connects to the user’s computer, and attempts to resolve the issue. When it needs assistance, it escalates to an IT employee who can join the remote session.

In the documented example, a person enters the password needed to log in. The support employee can then return control to the agent to finish the task. Phaze says it checks whether the employee has permission to connect and blocks unauthorized connection attempts.

The useful feature here is continuity. Escalation does not have to mean abandoning the session and reconstructing the problem somewhere else. A person can intervene at the point where judgment or authorization is needed, then let automation continue.

The company also publishes a Render Buddy example for watching long-running jobs and notifying users when a render finishes or encounters a problem. These examples illustrate possible workflows; they are not evidence that every support or rendering task can run unattended.

The API Manages Access and Machines

Phaze’s Enterprise API documentation describes a separate REST interface for administration. Available to Enterprise-plan customers, it covers members, organizations, groups, machines, locations, relays, and connection records. An API key is scoped to one enterprise.

That administrative API complements the agent connection. It lets developers automate which machine is assigned to a user or group and integrate access management with an existing workflow. Phaze’s own API announcement sketches a Slack command that allocates a workstation and returns a connection link.

The documentation also specifies an important approval boundary: connecting to a machine assigned to a user or group requires approval unless the connecting user is its assignee or belongs to its assigned group. Broad administrative access therefore should not be confused with unrestricted entry to every assigned workstation.

On its website, Phaze says agents inherit the machine access of their users. It also offers on-premises relays, directory synchronization, role-based controls, and connection and administration logs. Together, these features address the enterprise question of who may reach which computer and how that access is recorded.

Why Shared Workstations Are the Bigger Bet

In his funding announcement, CEO and co-founder Benjy Boxer argues that companies buy workstations for peak requirements while leaving part of that capacity unused. His proposed alternative is a pool of machines that can be assigned as demand changes and returned for someone else to use afterward.

The idea extends remote desktop beyond supporting distributed employees. If one employee supervises several agents, access demand can grow faster than headcount. A shared pool offers another way to meet that demand, provided available machines, software licensing, and workload requirements line up.

The commercial opportunity is in coordinating that access well. Fast streaming helps a person work comfortably; machine management helps IT allocate resources; agent connectivity lets automation use the same environment. Phaze is bringing those functions together rather than asking enterprises to manage them as unrelated systems.

The seed round gives the team funding to develop that approach. Its significance is the shift in what a workstation serves: a resource that employees and agents can access when a task needs it, with performance and permissions managed as part of the same product.