Hermes / Public Guide / v2
Hermes Zero-to-Hero — from chat user to agent operator
Hermes Zero-to-Hero

Don’t just prompt better.
Operate better.

The real upgrade is not better prompts. It’s becoming an agent operator.
Use Hermes to build a personal operating system for execution: durable context, inspect-before-act discipline, real tool use, skills, memory, cron jobs, webhooks, gateway workflows, and verification that compounds instead of resetting to zero every time.
Audience
AI-native founders & operators
People already using AI, but still trapped in isolated chats.
Primary hook
Inspect → Plan → Act
Simple enough to remember. Strong enough to change outcomes.
Transformation
Chat user → agent operator
Less prompt roulette. More repeatable leverage.
Section 1

Break the wrong model first

If the mental model is wrong, every feature gets used badly. Hermes is not most powerful as a chat box. It becomes valuable when it works inside a real operating frame.

Wrong model
  • Ask isolated questions
  • Restart context every time
  • Hope one prompt solves everything
  • Request action before inspection
  • Repeat successful work manually forever
This creates plausible answers, inconsistent outcomes, and chaos disguised as productivity.
Right model
  • Give durable context
  • Inspect the current state first
  • Plan before changing anything
  • Use tools to verify reality
  • Capture what works and automate repetition
This turns Hermes from a clever responder into part of your operating system.
Section 2

ASCII cinema — a living Hermes agent

This embedded ASCII video narrates the full process: from chaotic chat usage to durable context, inspection, planning, execution, system capture, and automation. The background changes with every scene so the visual field always matches the meaning.

Playback
Hermes ASCII cinema / living agent walkthrough
Scene 1 / 7

              
Wrong model
Theme: chat chaos
Most people start in chat chaos.
Hermes is being used like a chatbot, so context keeps resetting and the system never compounds. The agent is alive, but it is trapped in shallow requests.
Action Recognize the wrong model
Background message storms and broken threads
agent pulse unstable
visual mood fractured
Scenes / 7 beats
Section 3

First 15 minutes with Hermes

Start practical. Install or open Hermes, pick a model, verify the system, then run one grounded task with tools instead of staying in pure chat.

01 / Setup

Run the wizard once.

Use the interactive setup to configure provider, model, terminal, gateway, tools, and agent defaults.

hermes setup
02 / Health

Check before you operate.

Use health and status commands to catch missing dependencies, auth issues, and gateway/tool configuration problems.

hermes doctor
hermes status
03 / Model

Choose the right engine.

Hermes is provider-agnostic. Switch models/providers without changing the workflow around them.

hermes model
04 / One-shot

Use non-interactive mode for scripts.

For automation and repeatable checks, run a single query from the shell and capture the final answer.

hermes chat -q "Inspect this repo and summarize the risks"
Section 4

The current operating model

Hermes is strongest when durable context, tools, skills, memory, scheduling, and messaging all serve the same operator loop.

01 / CLI

Terminal-first, automation-ready.

Use hermes for interactive work, hermes chat -q for one-off execution, and --worktree when parallel agents need isolated git workspaces.

The CLI is the control room; chat is only one interface.
02 / Toolsets

Ground answers in reality.

Enable focused toolsets such as terminal, file, web, browser, vision, image generation, TTS, todo, cronjob, delegation, memory, session search, and skills.

If reality can be checked, don’t settle for guessing.
03 / Skills + memory

Compound what works.

Memory keeps stable facts and preferences. Skills keep reusable procedures, commands, pitfalls, and verification steps.

hermes skills search github
hermes skills install <identifier>
hermes memory status
04 / Gateway

Operate from where you already are.

The gateway brings Hermes to Telegram, Discord, Slack, WhatsApp, Signal, Email, SMS, Matrix, Mattermost, Home Assistant, Feishu, WeCom, BlueBubbles, Weixin, API Server, and webhooks.

hermes gateway setup
hermes gateway status
05 / Cron + webhooks

Turn checks into recurring operators.

Use cron for scheduled runs and webhooks for event-driven work. This is where useful prompts become operational routines.

hermes cron create "0 9 * * *"
hermes webhook subscribe deploy-watch
06 / MCP + plugins

Extend without rewriting the agent.

Connect MCP servers, install plugins, and expose Hermes itself as an MCP server when your workflow needs more specialized tools.

hermes mcp list
hermes plugins list
Section 5

The safe compounding workflow

Good Hermes usage is not a sequence of lucky chats. It is a controlled loop: context → inspect → plan → act → verify → compound.

1
Context

Load stable information: project, constraints, source-of-truth docs, success criteria, and do-not-touch zones.

2
Inspect

Read files, docs, logs, webpages, issues, and system state before touching anything.

3
Plan

Break work into reversible steps, surface risks, and choose the minimum safe move.

4
Act

Use tools deliberately. Ask for approval before destructive, broad, or external-impact operations.

5
Verify

Run tests, builds, browser checks, API probes, or live smoke tests. Report exact commands and results.

Section 6
Build your personal operating system.

Hermes becomes powerful when it stops being a chatbot and starts becoming part of how you execute: terminal, files, browser, memory, skills, schedules, webhooks, and messaging all tied to one disciplined loop.

Progression
Chat user
Then
Tool-aware operator
Then
Context + skills builder
Then
Automation owner