The OpenClaw Setup Nobody's Talking About

Artem Zhutov ยท Beginner ยท๐Ÿง  Large Language Models ยท1mo ago
๐Ÿ“š Claude Code Lab Mar 17โ€“ 6-week program for beginners: https://lab.artemzhutov.com I connected OpenClaw to my Obsidian vault with 4,000 notes. Not as a 24/7 employee - as a personal assistant that knows my context. ๐Ÿ’ฌ Discord: https://discord.gg/g5Z4Wk2fDk You've seen the demos - booking restaurants, sending emails overnight. But that's not how I use it. I use skills I trust, for tasks with predictable outcomes. Morning routines. Research that lands in my vault. Tasks extracted from meetings. In this video: what actually works, what doesn't, security considerations, and how to set it up โ€ฆ
Watch on YouTube โ†— (saves to browser)

Chapters (25)

Not a 24/7 employee - a personal assistant
0:54 What is OpenClaw vs Claude Code?
1:37 The model problem (Opus 4.5, ToS, costs)
3:02 Solution: Claude Code as backend
4:02 Black box problem โ†’ Skills you trust
5:17 Demo: Bookmark skill in action
6:11 Query your second brain
6:51 Extract tasks from meetings
8:01 Deep research โ†’ PDF to Telegram
10:47 Morning routine from phone
14:15 Intelligent inbox (images, voice, text)
16:46 What DOESN'T work
18:13 Rough edges (voice transcription issues)
19:55 Costs: API vs Claude Code backend
20:22 Security: Exposed servers
21:06 Prompt injection attacks
22:00 Security best practices
22:21 Installation walkthrough (Mac Mini)
25:51 Telegram bot setup
27:00 Syncing Obsidian vault (GitHub + Obsidian Sync)
28:06 Voice messages with Groq
28:33 Security audit command
29:14 Sandbox mode and permissions
29:57 Disable default skills
30:24 Final thoughts
5 Levels of AI Agents - From Simple LLM Calls to Multi-Agent Systems
Next Up
5 Levels of AI Agents - From Simple LLM Calls to Multi-Agent Systems
Dave Ebbelaar (LLM Eng)