For decades, building a networking startup required a large team of engineers and years of development. Artificial intelligence is rewriting that equation. A new venture called Doxx is betting that AI agents can not only operate networks but also build the technology behind them. Founded by serial networking entrepreneur Barrett Lyon, Doxx emerged from stealth this week with a platform it calls Agentic Defined Networking. The company also announced a $38 million Series A round led by Andreessen Horowitz, with participation from Animo Ventures and Focal.vc.
Lyon is not a newcomer to the networking industry. Over the past two decades, he has founded and sold multiple vendors, including Prolexic, which was acquired by Akamai; defense.net, acquired by F5 Networks; and netography, acquired by Vectra AI. His latest project turns his attention to AI, promising a network that can be managed and extended by autonomous agents rather than human operators alone.
The origins of Doxx trace back to a hobby project called Darkflare. Lyon created it as a TCP tunnel over Cloudflare that could traverse public networks in a way that helped avoid detection and potential censorship. The response to Darkflare led him to consider a VPN that cannot be stopped and is autonomously managed by AI. 'The hobby turned into an obsession, and the obsession turned into a business,' Lyon said.
How the Agentic Defined Network platform works
At its core, the Doxx platform combines three main elements: custom VPN transports, a company-run mesh, and a separate control plane. Each piece is designed to work together while giving users and their AI agents granular control over private networking.
Covert transports
The Doxx client supports WireGuard and custom transports for VPN traffic over WebSocket, Session Initiation Protocol, ping, QUIC, Cloudflare, and HTTPS. These transports are engineered to get through firewalls, making it possible to establish connections in restrictive environments. The goal is resilience and reach, not just standard VPN functionality.
Mesh networking
The mesh connects every site to every site with encryption. It runs over both IPv4 and IPv6 and is orchestrated through an application programming interface. That API-driven approach means agents can provision and modify mesh connectivity programmatically, without waiting for manual configuration.
Control plane
Lyon emphasized that the client application is a device, not a control plane. Routing, port forwarding, and firewall rules are managed in a separate portal. This separation keeps policy and orchestration distinct from endpoint connectivity, which is important when agents are involved.
Addressing
Devices on the mesh are not behind network address translation or carrier-grade NAT, according to Lyon. That allows peer-to-peer applications to function as intended, without the usual NAT traversal headaches. Each device can be addressed directly within the private network.
DNS and certificates
The company runs its own Domain Name System root with 196 domains and its own certificate authority. Lyon said the team wrote its own Border Gateway Protocol daemon and runs internal BGP across the mesh. Running these core services internally gives the platform tighter control over identity, naming, and routing.
Identity and discovery
The address book holds no phone numbers, email addresses, or names. Instead, users exchange cryptographic keys and assign an alias to each contact. A directory service seeds a new device with the data needed to find its contacts. After that initial call, devices sync with each other directly. This design minimizes exposed personal information while still enabling discovery.
How AI runs the network and what agents can do on it
The need for Doxx to be run by AI is pragmatic. Lyon said he turned to AI because he could not manage the network by himself. The platform spans 31 locations around the world and includes an internal mesh network. 'The whole thing is run completely by AI, so behind the scenes, I mean, it is a network with 31 locations around the world, and internally there is a mesh network, and I cannot, as a human being, manage all of this by myself,' Lyon said.
Before building, Lyon said the team modeled every site down to the wire and the optic in a virtual model. An infrastructure management system running the company's own AI on its own hardware then ordered the installation. It orchestrated shipping and delivery through data center APIs, while human technicians performed the remote smart hands installations. That combination of AI orchestration and human physical work allowed the network to be deployed without a large operations team.
The company also built tools for agents to work with users and with the network. An agent gateway gives an AI agent an identity in the Doxx chat app. The user pastes a credential into the agent. The agent obtains its certificate and appears in the user's chat. Users can create group chats with several agents. In one example, Lyon said one agent runs BGP while others handle other tasks.
The API is self-teaching. An agent given the configuration URL can learn how to build networks without further instructions. It then asks what the user wants to do, such as link an office to a house or give a Linux box a private IP address. Agents can also create their own domains and host names, sign certificates, and run web servers and applications inside the network.
Administrators issue read-only or admin tokens for each agent. Tokens can expire or be revoked. An administrator can also cut an agent's gateway access or delete the agent in the app. 'You do not just give them your account credentials to the master account,' Lyon said. The permission model is designed to limit what an autonomous agent can do if something goes wrong.
From 1992 to 2026 and beyond
The term doxxing usually refers to publishing someone's private information, but that is the exact opposite of what Doxx is about. As it turns out, Lyon registered the domain more than 30 years ago, as it was his hacker name at the time. 'So, I am just reclaiming the word, and it is kind of funny that we are doing the exact opposite of what doxing is. And, I do not know, it just seems fitting for me,' he said.
Looking forward, the plan is to build out more enterprise offerings as well as release some code as open source. Fundamentally, the technology is not just about using AI for networking but also about using AI to build the technology itself. That distinction matters because it suggests a future in which AI systems participate in every layer of network creation, from design and procurement to configuration and ongoing operations.
For network engineers and enterprise architects, the rise of agentic networking raises important questions. If AI agents can build private networks, manage BGP, issue certificates, and host applications, what is left for human administrators? The likely answer is oversight, policy, and exception handling. Administrators still set the guardrails, issue tokens, and decide what agents are allowed to do. But the day-to-day work of connecting sites, managing routing, and responding to changes could increasingly be delegated to software agents.
The funding round suggests investors see potential in that vision. Andreessen Horowitz led the $38 million Series A, with Animo Ventures and Focal.vc participating. The company's roots in Darkflare, a project built to tunnel through public networks and avoid censorship, also hint at a broader mission around resilient, private connectivity. Doxx is not just another VPN vendor; it is an attempt to make the network itself autonomous and self-managing.
Lyon's track record gives the startup credibility. Prolexic, defense.net, and netography each addressed different parts of networking, security, and visibility. Doxx brings those threads together with AI at the center. The platform's 196-domain DNS root, internal certificate authority, and BGP daemon show a willingness to build foundational infrastructure rather than rely entirely on existing providers. That vertical integration could give the company more control over performance, privacy, and security.
At the same time, handing network controls to AI raises governance challenges. Agents need identities, permissions, and audit trails. The token system is a start, but enterprises will want detailed logs, policy enforcement, and the ability to roll back changes. The company's plan to add more enterprise offerings will likely need to address those requirements. Open-sourcing some code could help build trust and allow security researchers to inspect critical components.
Lyon framed the effort as a way to surpass human limits. 'You can have a vision of where it is going to go, and then that vision hits a wall when you hit, you know, your own personal limitations,' he said. 'The AI just takes it well beyond, running this with an agentic partner, so to speak, turns this into fire.' That metaphor captures the ambition: not just automating tasks, but creating a network that can grow and adapt faster than any human team could manage alone.
Whether Doxx becomes the template for agentic networking or remains a niche experiment, its launch signals a shift. AI agents are moving from assistants that answer questions to operators that hold credentials, configure infrastructure, and run services. The network control plane is becoming a place where humans and AI agents collaborate. The coming years will show how well that collaboration scales, and what new security models are needed to keep it safe.
Source: Network World News