← Back to Hermes
Hermes front door

Hermes for operators

This is the front-door page for the Hermes operator lane: what it covers, what it already does well, and where to go next.

Reading ruleUse the subsection in order when you are new: setup first, then workflow, then model choices, then the deeper safety and recovery lanes.

Hermes is most useful to describe in operator terms, not just as a name or research topic. The practical question is: if a team wants to build, run, troubleshoot, and roll out Hermes responsibly, what guidance do they actually need first?

This Hermes subsection is built to answer that question. It turns the current internal research pack into a public, source-backed operator guide covering the most important early lanes: setup, first run, workflow, sessions, model configuration, security posture, messaging rollout, debugging, validation, and observability.

Just as important, this section does not pretend every lane is equally mature. The strongest current guidance is in the core operator pack: setup and first-run flow, real working patterns after setup succeeds, recovery paths when things break, session continuity, approval posture, messaging differences, and proof-oriented validation habits.

What you can use this section for

Honest current limits

This v1 subsection is already useful, but it is not complete in every edge lane yet. Companion-stack specifics are still thinner than the rest, especially broader IDE/editor coverage beyond the current VS Code evidence. Some model alias and auxiliary-slot examples can still get sharper. Non-default backend and container security tradeoffs also deserve a deeper later pass.

Suggested reading path

  1. Setup and first run
  2. Workflow and session operations
  3. Model configuration
  4. Security and approval modes
  5. Debugging and recovery
  6. Validation and observability