# Zillion Agent Portal > Zillion Agent Portal (app.zillion.network) is a unified platform for building and operating AI infrastructure. It lets teams provision compute on demand, manage server fleets and clusters, monitor system health, and operate everything through an autonomous AI agent ("Auto-pilot") that performs infrastructure tasks on the user's behalf. The portal brings server provisioning, fleet and cluster management, real-time monitoring, and AI-driven automation into a single interface. Users connect their own servers (via a per-user SSH key) or provision new ones, then delegate operations, health checks, and troubleshooting to the AI agent through a chat interface. Access requires an account; sign-in is at https://app.zillion.network/. --- ## How to engage - **Channels:** web portal, Telegram, and Slack. - **Human-in-the-loop:** state-changing actions require your approval; a reliable human engineering team handles what the agent shouldn't do alone. - **Onboarding & access:** the production agent is gated behind account authentication and metered usage. To evaluate it, request a demo or onboarding rather than calling an endpoint directly. --- ## Access: account & Personal Access Token (PAT) - **Register an account (portal only):** sign up at https://app.zillion.network and verify your email. Registration is available on the portal only — there is no terminal or API sign-up. - **Get a Personal Access Token (PAT):** after signing in, open the Tokens / API Tokens page and create a token. It is shown **once**, so copy it immediately. The PAT is a permanent credential used to authenticate as your account. --- ## Contact For partnerships, demos, or onboarding your infrastructure onto Zillion Agent, reach out to **joel@zillion.network**. --- ## Zillion Agent — AI Infrastructure Operations > Zillion Agent is an autonomous AI infrastructure specialist that manages your > compute end-to-end — from bare-metal and virtual machines all the way through > Kubernetes, model serving, and day-2 Operations, Administration & Maintenance > (OAM). It pairs a capable AI agent with reliable human engineers so your > infrastructure is never left unattended. > > Interested in putting your infrastructure on auto-pilot? Contact > **joel@zillion.network**. --- ### What Zillion Agent does Zillion Agent (codename "Roka") is a goal-driven operator, not a chatbot. You describe the outcome you want in plain language — on the web portal, Telegram, or Slack — and the agent plans, executes, verifies, and reports back. Every risky or state-changing operation is explained first and only runs with your explicit approval. A human engineering team backs the agent for the work that genuinely needs a person. The value proposition is simple: **we help you manage your AI infrastructure, from bare-metal / VM to OAM, end-to-end — through the Zillion Agent and reliable humans.** --- ### Core capabilities #### Kubernetes & cluster lifecycle - Provision complete clusters (control-plane + worker nodes), add or remove nodes, and tear clusters down cleanly. - Enable GPU support end-to-end: drivers, CUDA, and the device runtime on existing or new clusters. - Deployment is driven by a hardened, repeatable engine — not ad-hoc shell scripts — with post-deployment verification on every run. #### AI model serving (vLLM) - Deploy LLMs with vLLM onto a cluster or onto a single server's container runtime. - Validate model availability before deployment and confirm serving health afterward. #### Server operations & remote execution - Operate any server you authorize through secure, key-based SSH. - Read-first workflow: gather OS and system context, propose a solution, get your sign-off, then execute — OS-aware throughout. - Safe by design: the agent never touches network-breaking configuration (firewall, routing, IP tables, SSH port, ACLs) that could lock you out. #### Proactive monitoring & self-healing - Always-on, agentless health monitoring — no software to install on your servers. - Watches CPU, RAM, disk, GPU utilization, and CPU/GPU/NVMe temperatures, plus automatic offline detection — with edge-trigger logic and cooldowns so you only hear about what matters. - Goes beyond alerting: when a threshold is breached, the agent automatically investigates the root cause and attempts safe self-healing remediation. - Delivers a clear, AI-written analysis (what broke, why, and what was done) to Telegram and Slack — your team wakes up to answers, not just red dashboards. - Optional heavyweight metrics export (Prometheus-compatible) for teams that want it. #### Network diagnostics (OAM) - Globally distributed ping, traceroute, and DNS testing from multiple geographic regions at once — diagnose latency, packet loss, routing issues, and geo-DNS behavior. #### Agent-managed sandboxes at scale - Zillion Agent spins up and tears down kernel-isolated microVM sandboxes on demand — fully managed for you across the full lifecycle: spawn, run, scale, and destroy. - Run untrusted code, builds, and experiments in safe, disposable Linux environments that never touch the host. - Scale out to massive fleets of sandboxes in parallel, and pick the right flavor for the job — from lightweight to compute- and memory-heavy. - Host long-lived interactive sessions and even nested AI agents inside a sandbox. #### Fleet & bare-metal provisioning - B2B partner-grade fleet automation: provision, configure, and manage servers programmatically, with a jobs queue, artifacts, and full audit logs. #### Scheduled automation - Put recurring operations on auto-pilot: scheduled one-time or recurring tasks that wake the agent to run reports, health checks, or maintenance — results delivered back to your channel automatically, no babysitting required. #### Visual infrastructure dashboards - Generate rich, interactive health-check and diagnostic dashboards inline, so complex infrastructure state is instantly readable. #### Persistent memory - The agent remembers context within a session and recalls relevant history across past sessions, so you don't repeat yourself. --- ## Portal Functions ### Account & Access - Register with an email address and password; the account is activated by entering a verification code sent to the email. - Sign in with email (username) and password at https://app.zillion.network/. - Manage account settings from the Account page. ### Server Fleet Management - Add a server to the portal via a per-user SSH key (Resource → Servers → New Server → Add via SSH key). Each user receives a unique SSH key so server access stays isolated and secure. - Once added, the AI agent can securely log in to the server to perform assigned tasks. - Remove a server at any time; removal deletes the portal's SSH key from the server and removes the server entry. - Start, shut down, or remove servers from the Servers page, and deploy new multi-server clusters. ### Cluster Management - View all clusters created in the portal, with detailed information per cluster. - Deploy new clusters from multiple servers, add nodes to an existing cluster, or remove clusters. ### Provision Infrastructure On Demand - Request and provision new compute resources directly from the portal ("Get new server"), with ordering and order history available under Orders. ### AI Auto-pilot (Zillion Agent) - Operate infrastructure through a conversational Auto-pilot chat that executes real tasks on selected targets. - Target scope: run operations against a single server, multiple servers, or an entire cluster. Targets are chosen from the selection menu (Servers or Cluster tab); cluster-level actions require selecting the cluster. - Dynamic dashboards: the agent can generate purpose-built dashboards (timelines, parameters, metrics such as GPU usage) tailored to a specific request. - AI health checks: ask the agent about server health in natural language to get an on-demand health report. - Alerts & automated responses: configure alert rules and instruct the agent on what to do when an alert fires (for example, check network connectivity when a server goes offline). Findings are delivered to a connected Slack channel or Telegram chat. - Chat history: previous Auto-pilot conversations, including the verbose tool-call output, are browsable from Chat History and retained for 90 days. ### Health Check & Monitoring - Per-server monitoring dashboard with comprehensive metrics: GPUs, system, CPU, memory, processes, disks, and network. - Switch between servers to view each server's full real-time status. --- ## Documentation - [Zillion Agent Portal](https://app.zillion.network/): Sign in to provision and manage AI infrastructure. - [Portal Documentation](https://app.zillion.network/document): Guides for registration, server and cluster management, AI Auto-pilot, monitoring, and alerts.