#Hermes Agent Introduction

#What is Hermes Agent?

Hermes Agent is a self-hosted AI agent framework open-sourced by Nous Research in February 2026 (MIT license). It runs on your own machine, keeps persistent memory across sessions, proactively calls tools, connects to many messaging platforms, and learns and self-improves as you use it.

Unlike a chatbot where "the conversation ends and it forgets," Hermes behaves more like a long-term digital employee:

• Self-hosted / privacy-first: all data stays in local ~/.hermes/; officially no telemetry, no tracking
• Persistent memory: remembers your preferences, projects, and environment across sessions
• Self-improving: hard problems it solves are distilled into reusable "skills"
• Multi-channel reach: one gateway connects Telegram, Discord, Slack, and a dozen more
• Model-agnostic: Nous Portal, OpenRouter (200+ models), local vLLM, and more

#Design Philosophy: A Growing Agent

Traditional AI assistants are "amnesiac" — every conversation starts from zero. Hermes's core idea is that it grows with you:

Traditional assistant            Hermes Agent
─────────────────────            ─────────────
Starts from scratch       →      Remembers history & prefs (Memory)
Only preset abilities     →      Writes new skills after solving (Skills)
Waits passively           →      Runs on a schedule (Crons)
Fixed persona             →      Customizable "soul" + calibration (SOUL)
Single entry point        →      Unified multi-platform reach (Gateway)

#The Five Pillars

Hermes's capabilities are organized around five core pillars:

                ┌───────────────────────────┐
                │       Hermes Agent        │
                └───────────────────────────┘
        ┌──────────┬──────────┬──────────┬──────────┐
     Memory      Skills      Soul       Crons    Self-Improve
   (remembers) (does work) (who it is)(acts)   (gets better)

#Who It's For

Great fit:

• You want a long-term, private AI assistant, not a fresh chat window each time
• You need the AI to act proactively (daily briefings, monitoring alerts, scheduled backups)
• You value data privacy and want to own both the model and the data
• You like commanding the AI through familiar chat apps like Telegram / Discord
• You do MLOps / research and need trajectory generation and RL data export

Be careful when:

• You must supply your own server and LLM access (you own cost and ops)
• Powerful tool access means you must take permissions and security seriously
• It iterates fast — some features and docs may change

#Comparison

Choosing:

• Want a growing, proactive, cross-platform private AI → Hermes Agent
• Focused on in-terminal coding with deep vendor integration → Claude Code / Codex
• Just need occasional Q&A → use that vendor's chatbot directly

#Key Facts (from the official repo)

• Creator: Nous Research　License: MIT　First release: February 2026
• Built-in tools: 40+　Models: 200+ (via OpenRouter, etc.)
• Messaging platforms: Telegram, Discord, Slack, WhatsApp, Signal, and more
• Execution backends: Local / Docker / SSH / Singularity / Modal / Daytona

#Next Steps

• Architecture & Features — how Hermes works inside
• Installation & Usage — install and run your first chat
• Model Configuration — connect your LLM