Summary
Sharbel A. walks through the full architecture and construction of “Hermes HQ,” a custom mission control dashboard for managing the Hermes AI agent across terminals, Telegram, background jobs, and web interfaces. The video makes the case that once an agent starts doing real work across multiple systems, a standard chat thread is inadequate — and a purpose-built dashboard solves both the visibility and control problems.
The build is organized around five layers: a command layer for dispatching tasks via a Postgres message bus, an approval inbox that holds consequential actions (sending emails, deleting files, external commits) for human review before execution, a visibility layer tracking agents, cron jobs, costs, and failures, a long-term memory wiki, and an operator layer surfacing daily briefs and next-action panels. A local “Hermes Bridge” process running on a Mac mini handles actual agent execution, avoiding the need to expose the local machine to the internet.
The design philosophy throughout is ruthlessly functional: every dashboard module must answer what the workflow is, its current status, the next action, and any blockers — anything that fails that test gets cut. A free dashboard template is offered to viewers who watch to the end, making this a useful starting point for anyone building their own AI agent control plane.
📺 Source: Sharbel A. · Published August 02, 2026
🏷️ Format: Hands On Build
