# ZeroOps Cloud — Complete Platform Reference > ZeroOps Cloud is a SaaS control plane that provisions isolated AI agent runtimes on bare-metal k3s servers. Users get their own pod per run inside a Kata microVM. LLM access is dual-mode: BYOK (user-provided API keys) or Managed (pick a model, buy credits, no API key needed). Deploy any supported AI framework in 60 seconds with zero DevOps. Website: https://0ps.cloud Blog: https://blog.0ps.cloud --- ## What is ZeroOps Cloud? ZeroOps Cloud eliminates the infrastructure burden of running AI agents. Instead of managing servers, Docker, networking, TLS certificates, and storage yourself, you pick an AI framework, choose a plan, and your agent is live in 60 seconds. Every agent runs inside its own hardware-isolated Kata microVM on bare-metal servers. This means: - Full VM-level isolation between users (not just containers) - Persistent storage that survives restarts, updates, and version upgrades - Automatic HTTPS with individual TLS certificates per agent - Daily off-site backups to a separate data center - No noisy-neighbor performance issues The platform runs on dedicated bare-metal servers (not shared cloud instances), giving consistent performance for compute-intensive AI workloads. --- ## Supported AI Agent Frameworks ### Agent Zero - **Category:** Autonomous Agent - **What it does:** General-purpose AI framework with full access to its own Linux environment. It can write and execute code, install packages, browse the web, manage files, and interact with external APIs. Built with a self-developing approach where the agent learns and adapts. - **Key features:** Full Linux system access, autonomous web browsing, persistent memory across conversations, multi-LLM support (OpenAI, Anthropic, Google, local models), file management, self-developing agent - **Best for:** Automated code generation, system administration, data processing pipelines, web scraping, research automation, file conversion - **License:** MIT - **GitHub:** https://github.com/agent0ai/agent-zero - **Website:** https://agent-zero.ai - **Deploy:** https://0ps.cloud/agents/agent-zero ### OpenHands (formerly OpenDevin) - **Category:** AI Software Developer - **What it does:** AI-powered software development agent. Writes code, runs terminal commands, browses the web, and interacts with APIs in a sandboxed environment. The leading open-source Devin alternative with 50k+ GitHub stars. - **Key features:** Autonomous code writing/testing/debugging, terminal command execution, web browsing, GitHub issue resolution and PR creation, code refactoring and migration - **Best for:** Fixing bugs, implementing features from natural language, code review, framework migrations, writing tests and documentation - **License:** MIT - **GitHub:** https://github.com/OpenHands/OpenHands - **Website:** https://openhands.dev - **Deploy:** https://0ps.cloud/agents/openhands ### Open WebUI - **Category:** Chat Platform - **What it does:** Self-hosted AI chat interface supporting multiple LLM providers. A polished, production-ready alternative to ChatGPT that keeps your data private. - **Key features:** RAG with document uploads (PDF, DOCX, TXT), web search integration, tool/function calling, multi-model switching (GPT-4o, Claude, Gemini), conversation history, multi-user support with roles - **Best for:** Private ChatGPT for teams, document Q&A, research with web-augmented AI, multi-model comparison, secure AI chat for regulated industries - **License:** BSD-3-Clause - **GitHub:** https://github.com/open-webui/open-webui - **Website:** https://openwebui.com - **Deploy:** https://0ps.cloud/agents/open-webui ### OpenClaw - **Category:** Personal AI Assistant - **What it does:** Open-source personal AI assistant that integrates with messaging platforms. Connect it to WhatsApp, Telegram, Slack, or Discord and interact with AI through your everyday chat apps. - **Key features:** WhatsApp/Telegram/Slack/Discord integration, email and calendar management, task automation, always-on 24/7 operation, plugin ecosystem, multi-LLM support - **Best for:** Personal AI secretary, automated email triage, calendar scheduling, cross-platform task automation, customer support bots - **License:** MIT - **GitHub:** https://github.com/OpenClaw/OpenClaw - **Website:** https://openclaw.ai - **Deploy:** https://0ps.cloud/agents/openclaw --- ## Pricing All plans include VM isolation, persistent storage, automatic TLS, daily off-site backups, and managed LLM support. No free tier — billing starts when you create an instance. Annual billing saves 20% (pay for 10 months, get 12). ### Instance Types | Plan | vCPU | RAM | Storage | Monthly | Annual (per month) | |------|------|-----|---------|---------|-------------------| | Starter | 2 vCPU | 4 GB | 15 GB SSD | $34/mo | $27/mo ($326/yr) | | Standard | 4 vCPU | 8 GB | 15 GB SSD | $49/mo | $39/mo ($470/yr) | | Performance | 8 vCPU | 16 GB | 15 GB SSD | $89/mo | $71/mo ($854/yr) | ### Storage Add-On - $2/month per 10 GB block (or $20/year) - Maximum 100 GB total per instance (base + extra) - Expand anytime without downtime — storage never shrinks ### Managed LLM Credits - Pay-as-you-go: buy credits from $1 to $500 - Real provider prices (no per-token markup) - 4.5% platform fee on credit purchases (minimum $0.80) - Auto-recharge available when balance drops below threshold - 28 models across 10 providers: Anthropic (Claude Sonnet 4, Haiku 4.5, Opus 4), OpenAI (GPT-4o, o3, o4-mini), Google (Gemini 2.5 Pro/Flash), and more ### Billing Model - Charged at instance creation (not at first start) - Adding instances: prorated charge for remaining billing period - Stopping instances: billing continues (data and resources preserved) - Deleting instances: credit for unused days applied to next invoice - Resizing: atomic swap — credit for old plan + charge for new plan --- ## Platform Architecture ### Infrastructure - **Servers:** Dedicated bare-metal (Hetzner) — not shared cloud - **Orchestration:** k3s (lightweight Kubernetes) - **Isolation:** Kata Containers 3.27.0 with Cloud Hypervisor — each agent gets its own hardware-isolated microVM - **Storage:** Longhorn distributed block storage with daily off-site backups via restic - **TLS:** cert-manager with Let's Encrypt — individual certificate per agent subdomain - **Routing:** Traefik handles all HTTP/HTTPS traffic with automatic TLS termination - **Monitoring:** Grafana + Prometheus + Alertmanager with custom ZeroOps dashboard ### Security - VM-level isolation via Kata Containers (not just container namespaces) - Network policies restrict agent-to-agent and agent-to-host communication - Secrets encrypted at rest in Kubernetes - Agent images validated against a framework-specific registry whitelist - Daily off-site backups to geographically separate data center - SSH key-only access, fail2ban, restricted CORS, OpenAPI docs disabled in production ### LLM Access Modes **BYOK (Bring Your Own Key):** - User provides their own API keys in the agent's settings - Keys stored inside the agent's microVM only — never touch the platform - Works with any provider the agent framework supports **Managed:** - User selects a model from the platform's model catalog - Buys prepaid credits via Stripe - Platform injects LLM credentials into the agent pod automatically - Agent talks to a LiteLLM proxy (OpenAI-compatible) which routes to the actual provider - User never sees real provider API keys --- ## How It Works 1. **Sign up** at https://0ps.cloud/signup (email + password, no credit card required) 2. **Choose a framework** from the agents marketplace (Agent Zero, OpenHands, Open WebUI, OpenClaw) 3. **Select an instance type** (2 vCPU/4 GB, 4 vCPU/8 GB, or 8 vCPU/16 GB) 4. **Pay via Stripe** — first instance goes through Stripe Checkout, additional instances are prorated automatically 5. **Your agent is live** — accessible at `agent-{id}.0ps.cloud` with automatic HTTPS 6. **Configure LLM** — either paste your own API key (BYOK) or buy managed credits and pick a model 7. **Use your agent** — web terminal, real-time logs, resource monitoring, version updates, and backups all available from the dashboard --- ## Frequently Asked Questions **What happens to my data when I stop an instance?** Your data is preserved on persistent storage. Stop/start is purely operational — your files, conversations, and settings remain intact. Billing continues while stopped because resources are reserved. **Can I switch between BYOK and Managed LLM?** Yes, you can switch anytime from the instance settings. BYOK mode uses your own API keys stored inside the agent. Managed mode uses platform credits and the built-in model catalog. **How is this different from running Docker locally?** ZeroOps handles the infrastructure: Kubernetes orchestration, VM isolation, TLS certificates, DNS, persistent storage, backups, monitoring, and multi-provider LLM routing. You just pick a framework and use it. **What if my agent needs more storage?** You can expand storage anytime in 10 GB blocks ($2/month each) up to 100 GB total. Storage expansion happens online with no downtime. **Can I use my own domain?** Each agent gets a subdomain at `agent-{id}.0ps.cloud` with automatic TLS. Custom domains are on the roadmap. **Is my data backed up?** Yes. Daily off-site backups run automatically to a geographically separate Hetzner Storage Box. You can also trigger manual snapshots from the dashboard. **What AI models are available in managed mode?** 28 models across 10 providers including: Claude Sonnet 4, Claude Opus 4, GPT-4o, o3, o4-mini, Gemini 2.5 Pro, Gemini 2.5 Flash, and more. You pay real provider prices with no per-token markup. --- ## Contact & Links - **Website:** https://0ps.cloud - **Blog:** https://blog.0ps.cloud - **Sign Up:** https://0ps.cloud/signup - **Twitter/X:** @0pscloud - **Domain:** 0ps.cloud (pronounced "zero-ops")