Summary
Sharbel A. presents a business automation stack built around two complementary AI tools: Claude Code as a builder (creating apps, scripts, and agents in focused keyboard sessions) and Hermes as a persistent operator (running those tools on a schedule from a server or local machine). The core thesis is a clean division of labor — build it once with Claude Code, hand it to Hermes to run indefinitely.
Three live business examples illustrate the pattern: a personal website (charbell.com) where Hermes monitors email open rates and triggers Claude Code updates when metrics slip; a video editing workflow where Hermes analyzes second-by-second viewer retention data and feeds improvement briefs back to the editing agent; and a client services operation where Hermes monitors client health and flags problems proactively — with the video opening on a claim that Hermes autonomously closed a $6,000 client overnight.
The pivotal technical piece is shared memory: pointing both tools at the same memory account so context shared with one is immediately readable by the other. Sharbel reviews four options — Memo (fast and free to start), Zep (tracks how facts change over time), Leta (fully auditable open memory nodes), and Super Memory (fully local, nothing leaves the machine). He also notes that OpenAI’s CodeX can substitute for Claude Code for users already on a ChatGPT subscription, making the entire stack accessible without additional cost.
📺 Source: Sharbel A. · Published September 15, 2026
🏷️ Format: Workflow Case Study
