Doxx has launched a new kind of private networking platform that puts artificial intelligence in charge of the network itself. The company calls the approach Agentic Defined Networking, and it is aimed at users and AI agents that need to build, connect, and operate secure networks without waiting for a large engineering team. Doxx also announced a $38 million Series A round led by Andreessen Horowitz, with participation from Animo Ventures and Focal. The funding signals that investors see a market for networks that are not only automated at the edges but are managed by AI at the core.

The launch comes from Barrett Lyon, a serial entrepreneur who has spent more than two decades building and selling networking companies. His previous ventures include Prolexic, acquired by Akamai, Defense.net, acquired by F5 Networks, and Netography, acquired by Vectra AI. That track record gives the new project a pedigree in security, traffic management, and network defense. But Lyon says the current shift is different because AI changes the economics of building a networking startup. Tasks that once required years and a large team can now be modeled, tested, and operated with AI as a partner.

Key Facts

• Doxx launched an Agentic Defined Networking platform that lets users and AI agents build private networks.
• The company announced a $38 million Series A led by Andreessen Horowitz, with participation from Animo Ventures and Focal.
• Founder Barrett Lyon previously built Prolexic, Defense.net, and Netography, all of which were acquired.
• The platform combines custom VPN transports, an encrypted mesh, and a separate control plane.
• Doxx runs its own DNS root with 196 DNS zones, its own certificate authority, and an internal BGP daemon.
• AI manages a global network with 31 locations, including infrastructure ordering and installation orchestration.
• Agents receive identities, certificates, and scoped tokens so they can work inside the network without receiving master account credentials.

From a Hobby Tunnel to an AI-Managed Network

The roots of Doxx go back to Darkflare, a hobby project Lyon built as a TCP tunnel over Cloudflare. The idea was to move traffic through public networks in a way that could avoid detection and resist censorship. Darkflare drew attention, and Lyon began to think about a larger problem: could someone build a VPN that cannot be stopped and can be managed autonomously by AI? That question turned into an obsession, and the obsession became a business.

The origin story matters because it explains the design priorities. Doxx is not just another overlay network with a dashboard. It is built around the assumption that networks must survive hostile conditions, that connectivity must be flexible, and that the people running the network may not be able to manually manage every site, link, and route. AI is not a marketing layer added after the fact. It is the operating model for a distributed system that would be difficult for a human team to run at the same scale.

How Agentic Defined Networking Works

The platform brings together three main pieces: custom VPN transports, a company-run mesh, and a separate control plane. Each piece serves a different role, and together they create a network that can be reached, secured, and managed through software and AI agents.

Covert Transports

The Doxx client supports WireGuard as well as custom transports for VPN traffic over WebSocket, Session Initiation Protocol, ping, QUIC, Cloudflare, and HTTPS. These transports are designed to get through firewalls and restrictive networks. The goal is not merely to encrypt traffic but to give the network multiple ways to establish connectivity when conventional paths are blocked or unreliable. That flexibility is especially important for users who need private connectivity across networks they do not control.

Encrypted Mesh

The mesh connects every site to every site with encryption. It runs over IPv4 and IPv6 and is orchestrated through an application programming interface. A mesh architecture means that sites can communicate directly rather than depending on a central choke point for every path. The API-first design also makes the network programmable, which is essential for AI agents that need to request changes, create links, or inspect state without a human operator clicking through menus.

Separate Control Plane

Lyon describes the client app as a device, not a control plane. Routing, port forwarding, and firewall rules are managed in a separate portal. That separation is a security and operational choice. It keeps policy and configuration in a dedicated place while endpoints remain focused on connectivity. It also gives administrators a clearer boundary for permissions, auditing, and agent access.

Addressing Without NAT

Devices on the mesh are not behind network address translation or carrier-grade NAT, according to Lyon. That allows peer-to-peer applications to work more naturally. Many modern networks rely on NAT to stretch address space, but NAT can complicate direct connections, peer discovery, and certain real-time applications. By avoiding NAT and carrier-grade NAT inside the mesh, Doxx aims to make direct device-to-device communication a default rather than an exception.

DNS, Certificates, and BGP

Doxx runs its own DNS root with 196 DNS zones and its own certificate authority. Lyon said the company wrote its own Border Gateway Protocol daemon and runs internal BGP across the mesh. Running these components internally gives the platform control over naming, trust, and routing. It also means the network can issue certificates and resolve names without depending entirely on external systems. Internal BGP helps the mesh behave like a routed network rather than a flat collection of tunnels.

Identity and Discovery

The address book holds no phone numbers, email addresses, or names. Users exchange cryptographic keys and assign an alias to each contact. A directory service called Share seeds a new device with the data needed to find its contacts. After that initial exchange, devices sync with each other. This design reduces the amount of personal information stored in a central directory and puts identity on cryptographic proofs rather than conventional contact data.

How AI Runs the Network

The need for AI to run the network is pragmatic. Lyon said he turned to AI because he could not manage the network by himself. Doxx operates 31 locations around the world, and internally there is a mesh network. A human operator cannot personally manage every location, link, route, and certificate without automation. AI provides the leverage to run a global footprint with a small team.

Before building, 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. Human technicians performed the remote smart hands installations. That mix of AI-driven planning and human physical work shows how the company thinks about automation: AI can handle coordination, ordering, and configuration at scale, while people still perform tasks that require physical presence.

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, and 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 different tasks. This turns the chat interface into an operational console where agents can collaborate with people and with each other.

What Agents Can Do Inside the Network

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 DNS names and host names, sign certificates, and run web servers and applications inside the network. That capability moves agents beyond simple chat responses and into infrastructure operations.

The self-teaching model matters because it lowers the barrier to entry. A user does not need to memorize every API call or read a long manual before an agent can help. The agent can discover the platform's capabilities, ask clarifying questions, and then act within the permissions it has been granted. For small teams, that could compress the time needed to stand up private connectivity. For larger organizations, it could make network automation more accessible to people who are not deep networking specialists.

Permissions and Guardrails

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. Lyon said users should not just give agents their account credentials to the master account. The permission model is designed to limit blast radius while still letting agents perform useful work.

That approach reflects a broader challenge in agentic systems: how to give AI enough access to be useful without giving it unchecked control. By issuing scoped tokens, setting expiration, and providing revocation paths, Doxx tries to make agents first-class but constrained participants in the network. The gateway identity and certificate model also creates an audit trail of which agent did what, rather than blending all actions into a single human account.

The Name, the History, and the Road Ahead

The term doxxing usually refers to publishing someone's private information, but Doxx is built around the opposite idea: private networking, cryptographic identity, and controlled access. Lyon registered the name more than 30 years ago because it was his hacker name at the time. He describes the project as reclaiming the word, and he notes the irony that the company is doing the exact opposite of doxxing.

Looking forward, the plan is to build out more enterprise offerings and to release some code as open source. Fundamentally, the technology is not just about using AI for networking. It is also about using AI to build the technology itself. That distinction is important. AI can help operate a network, but it can also help design, model, and assemble the systems that make the network possible.

Lyon said a vision can hit a wall when it meets a founder's personal limitations. AI takes the work beyond that wall. Running the company with an agentic partner, he said, turns the effort into fire. The launch of Doxx shows how that idea is moving from a hobby project into a funded platform with global infrastructure, agent identities, and a permission model designed for a world where AI agents are not just users of networks but operators inside them.

Source:Network World News