You hooked up Telegram. You picked a model. You type prompts, get answers, close the tab.
You're using 8% of Hermes.
The other 92%, persistent memory, session branching, file rollbacks, voice mode, 17-platform reach, custom slash commands, sits there untouched while you treat a fully-loaded agent like a slightly smarter ChatGPT.
This article is the 92%. 15 features, ranked by impact. Most Hermes users haven't touched a single one.
Quick context for newcomers: Hermes is a provider-agnostic AI agent. 100+ pre-built skills. Runs on 17 messengers from one process. Swaps models mid-session without restarting. Built portable, extensible, yours, not locked to one vendor like every other agent on the market.
Now the list.
---
## PART 1: The Setup You Skipped
1. /personality + SOUL.md
Hermes reads SOUL.md at boot. Whatever's in it becomes your agent's voice forever, every session, every platform. /personality switches between named personas on the fly.
Stop typing "you are a senior X expert" at the start of every chat. Write it once.
2. MEMORY.md + USER.md
Two persistent files, read every session. MEMORY.md = project notebook. USER.md = what it knows about you. Indexed with FTS5 + LLM summarizer, so a memory from 8 weeks ago surfaces in today's session.
You stop re-explaining yourself.
3. /insights [days]
Cross-session analytics. Tokens burned, providers used, where the agent stalled, what you keep returning to. /insights 30 = last month at a glance.
4. /snapshot
Save full Hermes state before doing something risky. Break things on purpose. /snapshot restore to come back. Rollbacks for the agent itself.
---
Generated by Thread Navigator
Press ⌘ + S to quick-export