TechnicalDiscussion

Agent OS: Obsidian + Multi-Agent Orchestration + GLM 5.2!

Julian Goldie SEO

This video answers community questions about building an Agent Operating System (Agent OS) that combines multiple AI agents like Claude and Hermes with tools like Obsidian for memory management. The speaker covers practical implementations including multi-agent orchestration, video/animation agents, lead generation automation, and GLM 5.2 model optimization.

Summary

The speaker addresses questions from the AI Profit Bottom community about constructing and operating an integrated Agent OS. Key discussion points include how to organize multiple projects within Obsidian by creating separate note systems and memory spaces for each project to prevent overlap, while maintaining interconnected knowledge graphs through color-coding and linking. The speaker shares real examples from community members: Andrew's project organization strategy, Robbie's lead generation system called 'Deal Desk' that finds prospects and reaches out after approval, and implementations using multi-agent orchestration with tools like Paperclip to create hierarchical agent teams. The transcript covers technical solutions like using Open Montage for video generation and video agents with specific style instructions, explains GLM 5.2's advantages as an open-source alternative to Claude Opus that performs comparably or better on many tasks while being cheaper and faster through Ollama integration. The speaker addresses whether beginners should learn N8N versus jumping directly to Agent OS, recommending Agent OS for experienced users comfortable with CLIs and Claude code, while N8N serves better for learning fundamentals. The community aspect is emphasized, including the /learn command in Hermes Agent for training agents on custom guides, VPS setup options for running Hermes remotely, and the support structure of daily tutorials, 24/7 community help, weekly coaching calls, and local meetup maps.

Key Insights

  • Andrew's approach separates each project with its own notes and memory space to prevent overlap, while maintaining automatic agent synchronization between projects through proper organization, allowing agents to distinguish context across multiple projects without manual updates.
  • GLM 5.2 outperforms Claude Opus 4.8 on many tasks, is open-source, significantly cheaper with higher token allocations, and achieves up to 2x faster responses when run through Ollama for local deployment.
  • Robbie built a custom lead generation system within the Agent OS that automatically finds prospects for contracting jobs, routes them for user approval, and then automatically sends outreach messages through an 'hourly system' without requiring code understanding.
  • The speaker recommends Agent OS over N8N for users with AI experience and CLI/Claude code comfort because Agent OS automates more without requiring HTTPS configuration or individual module setup, making it more efficient than mastering N8N fundamentals.
  • Hermes Agent's new /learn command allows users to input tutorials or guides directly into the chat, which the agent saves as reusable skills for future sessions, enabling dynamic agent training without modifying system architecture.

Topics

Agent Operating System architecture and setupObsidian integration for memory management and knowledge graphsMulti-agent orchestration and hierarchical agent teamsGLM 5.2 model as open-source Claude alternativeLead generation automation workflowsVideo and animation agent coordinationN8N versus Agent OS learning pathsHermes Agent training with /learn commandVPS deployment for remote agent systemsCommunity-driven AI automation development

Transcript

[0:00] So, today we're going to be answering the latest questions on our agent operating system. This basically how we combine all of our agents together. So, for example, we've got like Hermes and the Hermes Oracle, we've got our lead generation machine over here, we have Claude ready to go, we've got our mission control system, and everything we need to win with AI inside one single place. Even got, for example, a Kanban notebook lamp all synced inside one place. So, today what we're going to be doing is answering some of the latest questions that I've got inside the AI [0:30] Profit Bottom community because I know if people have questions inside here, you probably have similar…

Full transcript available for MurmurCast members

Sign Up to Access

More from Julian Goldie SEO

Get AI summaries like this delivered to your inbox daily

Get AI summaries delivered to your inbox

MurmurCast summarizes your YouTube channels, podcasts, and newsletters into one daily email digest.