Descriptions:
Craig Hewitt, an AI agency founder, shares 10 hard-won lessons from deploying AI agents for clients, structured as a complete walkthrough for setting up the Hermes agent framework in the cloud. Hermes, built by New Research (described as a hacker collective turned organization), is presented as the leading always-on, open-source, fully configurable agent framework — succeeding earlier tools like Open Claw, which was acquired by OpenAI earlier this year.
The tutorial covers selecting a Hostinger KVM 2 VPS plan, using Docker to containerize the agent for security, and deploying via one-click install. Craig explains why cloud hosting beats a local Mac Mini for reliability — a Mac Mini that dies across town while you’re away can’t be fixed remotely — and walks through configuring Hermes via its soul.md personality file and Telegram-based management interface. A key differentiator for Hermes is its persistent cross-session memory: unlike Claude Code or similar tools, it remembers every past conversation, allowing the agent to compound knowledge about your preferences and working style over time.
The video also addresses Hermes’s self-improving skill system, where the agent autonomously writes and prunes its own capabilities, with a practical warning about skill bloat and the growing security risk of malware embedded in third-party skill hubs like skills.sh. For developers and agencies wanting maximum configurability beyond consumer-grade agent tools, this is a thorough, experience-backed guide to running a production Hermes agent securely in the cloud.
📺 Source: Craig Hewitt · Published June 12, 2026
🏷️ Format: Tutorial Demo
