OpenClaw - Build Your 24/7 Personal AI Agents From Zero

OpenCloud can run on a personal computer, dedicated hardware, or a VPS, each with different uptime, isolation, and cost, with VPS offering the easiest, always-on setup.

Apply a practical framework to choose among OpenCloud, Cloud Code, Cowork, and N8n for autonomous tasks, coding, desktop work, and high-volume pipelines.

explore why security matters when building an OpenCloud agent, the orchestration layer—not the AI brain—connected to Telegram, and learn layered protections: server isolation, network security, access controls, and behavioral guardrails.

Compare free and paid AI model options for OpenCloud, including OpenRouter, Gemini Flash, Kimi, and Anthropic API keys, and learn setup, budgeting, and routing basics.

Secure your OpenCloud setup by generating and safely storing the gateway token and Anthropic API key, then set a monthly spend limit and manage credits.

Launch the gateway dashboard, log in with your gateway token, and complete bootstrapped.md first run interview to define your agent, name, vibe, and work style; memory is saved.

Hardening your agent's security builds a solid foundation by applying OpenCloud's recommended settings. Enforce strict rules and safeguards to prevent runaway tasks and costs.

Connect your OpenCloud agent to Telegram via the gateway dashboard, create a Telegram bot with BotFather, and pair your account to securely chat with your agent from anywhere.

Understand how API costs work in OpenCloud, including context size, token usage, and Heartbeat implications, then learn to select cheaper models and monitor credits to stay in control.

Explore the gateway dashboard in depth, learning how to monitor chats, manage sessions and usage, and control cron jobs and agent tools from the OpenCloud control UI.

Learn three update methods for OpenCloud—gateway dashboard, agent prompts, and docker manager—and implement practical recovery steps to fix unresponsive agents, revoke API keys, and restore from backups.

Explore four model tiers, from flagship Opus to free options, and balance power and cost by reserving flagship models for planning while using cheaper models for execution, with guardrails.

Connect your ChargedGPT subscription to OpenCloud via OAuth, then set smart routing to use the cheapest capable model. Add a free fallback to keep the agent responsive during outages.

Set up a free fallback model, Kimi K2.5 from Moonshot AI via Nvidia cloud, add its API key as an environment variable, and deploy it as last resort on OpenCloud.

Learn to run OpenCloud with free and cheap models by configuring OpenRouter, minimizing costs with Minimax M2.5, Kimi K2.5, and Gemini Flash, and switch models to fit budgets.

Organize conversations by topic by creating Telegram groups, adding your agent, and enabling group chat support with clear permissions and group-specific instructions.

Equip your OpenCloud agent with skills to perform tasks, and carefully review virus total security scan and permissions via Telegram clohub before installation.

Automate Gmail with Google Workspace: an agent checks unread emails, prioritizes those needing attention, drafts replies, and delivers a daily 8am Telegram summary via a cron job.

Discover how workspace files shape your OpenCloud agent, with agents.md, sole.md, and user.md, and boost memory through daily journals, persistent memory, and session memory.

Master how to automate your AI agent with cron jobs for scheduled tasks and heartbeat for continuous background monitoring, keeping the heartbeat lean and using a cheap model.

Spin up sub agents to research tasks in parallel and compile a unified report with the main agent. Use Brave Search API for lookup and compare CRM platforms in parallel.

Build a complete landing page for the cloud ai operating system using three subagents—copy, research, and html—demonstrating parallel work, rapid assembly, and a strong starting point for a live site.

Build a structured business context document (user.md) consolidating who you are, your work style, current focus, and how AI should help—foundation level one for your business brain.

Automate your competitive intelligence with a weekly competitor watch report that uses cron jobs and sub-agents, Brave Search, and Telegram delivery, compiling insights on Zapier, Mac.com, and N8n.

Build a daily intelligence digest that automatically compiles your calendar, Gmail, and memory into a concise morning briefing with today's meetings, emails, and top priorities.

Build your own AI employees by creating a second AI brain with Obsidian in the OpenClaw AI agent masterclass.

Learn how to SSH into your server, using a browser terminal or your own computer, to troubleshoot, inspect Docker containers, and run OpenCloud commands from inside the container.

Learn to guard against prompt injection by applying sandbox rules, dedicated accounts, and a permanent data-only memory rule, then test defenses with a production deployment checklist.

Build the task board and calendar in mission control, enabling kanban columns backlog to done, task creation and drag-and-drop, a live activity feed, and a calendar of real cron jobs.

Add a project screen and a document screen to the mission control dashboard, providing a clear view of projects, progress, owners, milestones, and a searchable document archive.

Learn to extend mission control with a memory screen, a team roster, and a live, pixel-art office that visualizes real-time agent activity from the task board.

Build a financial tracker to monitor API costs, subscriptions, and revenue for your AI setup. Use server-side persistence to keep data and display a live monthly summary.

Access mission control from your phone via a three-path PDF guide (path A, B, or C) to view the dashboard, assign tasks, and review the morning briefing.

Publish mission control publicly on a domain with https and a password by setting up a reverse proxy on a VPS with Kadi and a self-signed certificate.