Published: 2026-06-30
Summary

Hermes Agent Setup: Skills, Gateways, Tasks & Memory Walkthrough

Chapters / key moments (click to jump — plays here on the page)

Matthew Berman walks through a full Hermes agent setup — install, adding skills from GitHub, wiring up a Telegram gateway, scheduling tasks, and configuring memory, profiles, and model routing. The video highlights what makes Hermes different from OpenClaw: a stack of skills preloaded out of the box, self-healing when a skill hits an unexpected error, and built-in provider/model routing.

Source video

""The best thing since OpenClaw" (Hermes Tutorial)" by Matthew BermanWatch on YouTube →

Key Takeaways

  • Preloaded skills. Hermes ships with many skills enabled out of the box (Claude Code, Codex, Excalidraw diagrams, Manim animation video, and more). Each appears in the Skills tab, toggles on/off with a click, and is plain, readable markdown you can open and edit.
  • Installing a skill is copy-paste. Grab a skill.md raw file from a GitHub repo, click the + in the Skills tab, name it, choose a category, paste, and save. When a skill needs a companion file it doesn't have, Hermes self-heals — fetching a fresh copy of the repo into a temp directory and running from there.
  • Chat gateways from the CLI. Open the command line and run hermes gateway setup to connect Telegram, Slack, Matrix, Mattermost, WhatsApp, Signal, email, and more. Telegram needs a bot token from @BotFather (/newbot) and your numeric user ID from @userinfobot to restrict who can message the agent.
  • Tasks, memory, and profiles. Tasks are scheduled automations (e.g. a daily calendar brief every 24h). Memory stores facts about you and holds the agent's SOUL document (identity.md / soul.md style) editable in-app. Profiles let you silo separate agents — a marketing agent and a development agent — instead of one bloated assistant.
  • Broad provider support + model routing. Settings expose OpenAI, Anthropic, DeepSeek, Gemini, Kimi, LM Studio (fully local), Mistral, and Nvidia NIMs, plus per-task model routing — assign specific models for vision, compression, web extraction, or session search, or leave the default "auto" to let Hermes decide.

Commands & Code Mentioned

hermes gateway setup      # open the CLI and connect a chat gateway (Telegram, Slack, WhatsApp, Signal, ...)
/newbot                   # in Telegram @BotFather — creates a bot and returns a bot token
/last30days               # example of invoking an installed skill from Hermes chat (type / to list skills)

Weekly Digest — In Your Inbox

Get the week's top AI agent news, updates, and guides — every Friday.