Utah startup Phaze, founded by the team behind Parsec, has raised $8.75 million in a seed round led by Andreessen Horowitz (a16z). Its pitch: turn underused workstations into a shared pool of compute that both employees and AI agents can tap, wherever the machine sits.
🔑 Key takeaways
- Seed round of $8.75M led by Andreessen Horowitz; valuation not disclosed.
- Low-latency remote desktop platform open to both employees and AI agents.
- Agents connect through a Model Context Protocol (MCP) server, with no computer-use software required on the client machine, per SiliconReport.
- Agent connectivity is still described as experimental; no revenue or named enterprise customers have been published.
A seed round built on Parsec’s track record
Phaze Inc. announced on November 4, 2026 that it had closed $8.75 million in seed funding. Andreessen Horowitz led the round, with participation from NextView, Makers Fund and Lerer Hippeau. The company has not disclosed its valuation. Sources do not agree on the full investor list: SiliconANGLE names Index, NextView, Makers and Lerer Hippeau, while SiliconReport, Pomegra, The SaaS News and The Next Web cite Indie, NextView, Makers or Makers Fund, and Lerer Hippeau. The versions do not make clear whether “Index” and “Indie” refer to two separate investors or to one investor under two names.
The founding team brings a proven track record. Phaze was created by the engineers who built Parsec, a low-latency remote desktop application used notably for graphics-intensive workloads. Benjy Boxer is co-founder and CEO. Pomegra also names Chris Dickson as a co-founder and says that Boxer and Dickson led Parsec. Unity acquired Parsec in 2021 for $320 million. That prior experience is central to the investment thesis, according to Martin Casado, general partner at a16z.
“Benjy and the team have been working on this for more than ten years, and they are bringing that experience to a new era of remote access and machine management.”
Martin Casado, General Partner, Andreessen Horowitz
Expensive workstations, often sitting idle
Phaze starts from a simple economic observation. Companies equip employees with high-end workstations that can cost more than $10,000. These machines render complex 3D graphics or compile very large codebases, yet they are frequently unused during lunch breaks, meetings, vacations and other absences. Assigning one machine per employee therefore locks up significant hardware capital for much of the day.
Computer-use agents create a second problem. When they run on an employee’s own machine, they can seize control of the screen and disrupt the person’s work. Phaze avoids this by giving each agent its own access to a remote machine. According to The Next Web, the MCP server can run in the background without taking over the user’s personal computer.
| Metric | Figure | Context |
|---|---|---|
| Seed round | $8.75M | Led by Andreessen Horowitz |
| Unity acquisition of Parsec (2021) | $320M | Prior exit of the founding team |
| High-end workstation cost | over $10,000 | Often idle outside active work |
| Valuation | not disclosed | No revenue figures published |
MCP access without taking over the screen
To reach the machines, Phaze combines its proprietary low-latency streaming technology with a purpose-built Model Context Protocol (MCP) server. MCP is a common standard for connecting AI systems to tools. Through this layer, an agent can navigate the operating system, launch applications and carry out tasks directly on the machine, much as an employee would. The goal is to avoid confining agents to application programming interfaces (APIs), which SiliconANGLE describes as costly and brittle.
Accounts differ on where the server sits. SiliconANGLE describes an MCP server built to support the streaming layer, while Pomegra says agents connect through a local MCP server.
Authentication and permissions
Phaze adds authentication, centralised access management and enforcement of corporate security policies. User-level permissions restrict which resources agents can reach. According to The Next Web, an agent holds only the rights of the person it is linked to, and it can work in the background without occupying the user’s screen. SiliconReport adds that the system does not require installing computer-use software on the client device.
“The computer on your desk should not determine the compute you can access. People need powerful workstations to do their jobs, and now AI agents need computers they can actually operate.”
Benjy Boxer, Co-founder and CEO, Phaze
Shared sessions, competition and risks
The product supports shared sessions between a person and an agent. Both can work in the same remote session and hand over control to each other. Phaze cites an IT support bot that diagnoses a problem when a ticket is opened, attempts a fix, then transfers the session to a human engineer when the fix goes beyond its authorised actions. Pomegra describes a demo in which a bot receives a request in Slack, connects to the employee’s computer and calls in a human when it gets stuck. The Next Web notes that agent connectivity via MCP is still considered experimental.
A crowded market
The segment is already contested. According to The Next Web, Baseten acquired Blaxel, which builds isolated environments for agents, in September. Perplexity offers a platform that can turn a Mac mini into an always-on agent. Demand for that hardware is strong: Mac mini stock sold out in the United States in spring 2026, a sign, per The Next Web, of growing interest in dedicated agent hardware.
Pomegra considers the round modest next to the capital flowing into agent infrastructure, and sees a competitive market. Most isolated environments offered to agents are graphics-free containers built to run code. Phaze targets a different need: some high-value tasks require a full graphical computer, with real applications and the ability for a human to take over. The company also sets its offer apart from the virtual machines and remote desktop services already sold by major cloud providers.
The sources flag several risks:
- no named enterprise customers and no published usage data;
- the possibility that cloud or virtual desktop providers add agent features of their own;
- the need to give agents access to workstations that may hold authenticated sessions.
Buyer confidence in this security and session-sharing model will therefore be a decisive factor in adoption.
Outlook and scenarios
Phaze’s thesis rests on a simple premise: existing hardware is underused, and AI agents will need access to real machines. In the bullish scenario, the platform quickly signs a handful of enterprise customers, demonstrates recurring use cases, and buyers judge its shared-session security model to be robust.
The bearish scenario arrives if the large cloud providers rapidly ship equivalent agent features, or if companies remain reluctant to expose workstations that hold authenticated sessions. In either case, this $8.75M round is a signal of interest in remote access for agents, but on its own it does not validate a business model.
Sources
Cet article est publié à titre informatif et éducatif. Il ne constitue en aucun cas un conseil en investissement. Faites vos propres recherches (DYOR) avant toute décision.
