Meet your personal ai employee
In this section
- Meet Your Personal AI EmployeeOpenClaw proved that AI Employees are real, they work, and people want them.
- The AI Employee MomentUnderstand what defines an AI Employee across five architectural dimensions, then discover why ownership is the sixth dimension that makes OpenClaw's version a Personal AI Employee.
- Install & Connect Your EmployeeInstall OpenClaw from scratch, survive the crash loop, connect WhatsApp/Telegram/Discord, and send your first message to a working AI Employee.
- Delegate Real WorkSend two tasks to your AI Employee while watching the dashboard, discover the agent loop through live observation, and hit the boundary that proves this agent has real power under real constraints.
- Customize Your Employee's BrainEdit the workspace files that compose your agent's system prompt, prove with /context list when edits actually take effect, and discover the BOOTSTRAP self-configuration pattern.
- Memory & CommandsStore preferences that survive across sessions, verify memory writes on disk and in the dashboard, use memory_search and memory_get to find old notes, learn the three real paths to MEMORY.md (manual, compaction memory flush, dreaming deep phase), enable opt-in dreaming and active memory, and discover slash commands that bypass the model entirely.
- Install & Author Agent SkillsInstall skills from skills.sh and ClawHub, invoke them as slash commands, author your own with skill-creator, and let Skill Workshop capture procedural memory automatically.
- Connect External ToolsConfigure the mcp-server-time MCP server with one command, learn gateway-level versus workspace-level config, master the silent failure pattern, and understand the three transport options (stdio for local, SSE and streamable-http for remote).
- Plugins, Bundles & The Decision TreeDiscover that channels and model providers ARE plugins, tour the bundled plugin ecosystem, install a bundle plugin for a real domain need, and earn the four-way decision tree for choosing the right extension type.
- Make It ProactiveEnable heartbeats and cron jobs to make your agent act on its own schedule. Write a HEARTBEAT.md, understand the Agent OS mental model, and learn why one heartbeat costs less than five cron jobs.
- Give It a VoiceEnable text-to-speech on WhatsApp with three bundled providers, learn the four TTS modes, and discover why letting the agent decide when to speak is the smartest option.
- Add a Second AgentCreate a second agent on the same WhatsApp number, configure channel binding and routing, set agent identity, and understand why session cache means identity changes only apply to new conversations.
- Connecting Google WorkspaceConnect your AI Employee to Gmail, Calendar, and Drive using gog, experience the moment your agent processes real email, and confront the security implications of OAuth access
- Orchestrate Other AgentsDelegate tasks with sessions_spawn, control Claude Code via ACP from WhatsApp, understand the two-layer concurrency model, and learn why model quality determines whether orchestration works at all.
- Gate Your Agent's ToolsSpecify a TypeScript plugin that intercepts tool calls and requires operator approval on WhatsApp, then have your AI Employee build it from a constraint-driven prompt referencing this lesson.
- Deploy to ProductionDeploy your AI Employee from laptop to a $5/month VPS, survive the paper cuts, and verify production health through an SSH tunnel.
- Isolate with NemoClawUnderstand NemoClaw's 4-pod architecture where API keys live in a separate process the agent cannot reach, and learn when $10 more per month is worth the security upgrade.
- Chapter 56: Meet Your Personal AI Employee QuizA 50-question assessment covering AI Employee architecture, OpenClaw platform mastery, security models, deployment patterns, and production readiness across all 17 lessons of Chapter 56.