doxx.net Builds a Private Network for People and AI Agents

doxx.net is a Miami-based cybersecurity and network infrastructure company building what it calls an Agentic Defined Network. Founded in 2025 by Barrett Lyon, the company gives people, devices, services, and software agents private channels for browsing, communication, file transfer, and network control.

The company entered open beta on October 1, 2026, alongside a $38 million Series A led by Andreessen Horowitz, with Animo Ventures and focal.vc participating. The financing matters because software agents are moving from chat windows into networks, applications, and business workflows, where every action can create a new security and privacy trail.

doxx.net's answer is architectural. Instead of treating privacy as an application feature, the company is trying to own and operate the network layers beneath it. That makes doxx.net an early test of whether privacy infrastructure can expand from protecting human traffic to governing how autonomous software connects and acts.

What doxx.net Is Building

doxx.net combines encrypted tunnels, private DNS, DNS threat blocking, firewall controls, device networking, encrypted peer-to-peer communication, file transfer, and domain management in one platform. The company's public repository documents WireGuard-based connectivity, personalized blocklists, DNS-over-HTTPS and DNS-over-TLS, per-tunnel rules, real-time security monitoring, and domain hosting across 196 supported private TLDs.

The product also gives software agents network-level controls through APIs and native Model Context Protocol support. An agent can receive an identity, connect to private channels, and manage tunnels, IP addresses, DNS, and firewall settings without forcing every action through a conventional consumer account. The company's launch materials say users can create accounts without supplying an email address or phone number.

This scope is unusually broad for a young company. doxx.net is not pitching another virtual private network with a friendlier interface. It is assembling connectivity, communications, naming, security policy, and agent control into a single operating layer. The open beta is the point where that ambition meets real usage, product clarity, and the brutal honesty of edge cases.

Why Owned Infrastructure Is the Strategic Bet

Most privacy products sit on infrastructure controlled by somebody else. Traffic may be encrypted, but it still passes through rented servers, external DNS systems, third-party identity layers, and networks with their own visibility. doxx.net says it owns its IP space, autonomous system number, DNS root, certificate authority, bare-metal servers, and global backbone.

Owning more of the stack can reduce intermediary exposure and give the company tighter control over routing, naming, security rules, and incident response. It can also make operations more capital-intensive and demanding. Infrastructure ownership is not a magic privacy spell. It is a commitment to run the difficult parts rather than outsource them and hope the contractual fine print behaves.

That commitment is central to the Agentic Defined Network thesis. Software agents can browse continuously, call APIs, move files, and act while their users sleep. The network therefore needs controls that recognize automated actors without converting every agent interaction into another surveillance dataset. doxx.net is betting that privacy and agent autonomy belong in the network architecture, not as an apology added after deployment.

Barrett Lyon Returns to the Network Layer

Barrett Lyon is the founder and CEO of doxx.net. His earlier companies include Prolexic, BitGravity, XDN, and Defense.net, spanning DDoS protection, content delivery, and network infrastructure. Those ventures were later acquired by Akamai, Tata Communications, Fortinet, and F5.

The history matters because doxx.net depends on a founder being willing to own physical and logical infrastructure that many software companies would rather rent. Lyon has repeatedly built at the layer where packets, attacks, and network economics become operational problems. The new company applies that experience to a different pressure point: private connectivity for a world where people and autonomous software share the same network.

Andreessen Horowitz led the Series A, with Animo Ventures and focal.vc participating. a16z partner Joel de la Garza joined the board. The capital gives doxx.net more room to convert a founder-heavy infrastructure thesis into a dependable product and, eventually, an enterprise offering. Axios reported that the enterprise product is planned by the end of 2026, which remains a forward-looking target.

What the Early Traction Does and Does Not Prove

doxx.net reports that its closed beta began in December 2025 and blocked more than 38 million potential threats. The company also says its infrastructure operates across six continents. Those figures show that the system is processing meaningful activity, but the threat count is company-reported and no independent audit methodology has been published.

That distinction matters. Security companies can drown readers in very large numbers because the internet produces very large volumes of noise. The more important test is whether doxx.net can turn infrastructure control into reliable privacy, comprehensible policy, and a product normal people and technical teams can operate without becoming network engineers.

The company does provide an unusually visible technical surface. Its GitHub repository documents product capabilities, API resources, and open-source tools, while the company site invites developers to report bugs and suggest changes. No official careers page or current list of open roles was found, so public code participation should not be confused with a verified hiring campaign.

Why doxx.net Matters Now

The security market has spent years separating identity, endpoint protection, networking, messaging, and privacy into different products and vendors. Agentic software stresses those boundaries. A single agent may need authenticated access to multiple services, private communication with people, controlled file movement, and policies that follow it across devices.

doxx.net is proposing a shared network layer for that problem. If the model works, agents do not have to inherit the same fragmented connectivity and data leakage patterns that humans have tolerated. If it fails, it will likely fail on the hard details: usability, interoperability, trust, reliability, and the operational cost of maintaining a global private network.

That is why the company is worth watching after the financing headline fades. doxx.net has credible founder history, a public beta, owned infrastructure, and a clear view of the problem. What it does not yet have publicly is proof of enterprise adoption, audited security outcomes, or a mature operating record. The next chapter is execution: turning a compelling network architecture into a service people and agents can trust when the stakes are real.