From a 5-minute quickstart, to real-world scenarios, to architecture deep-dives and custom development — transform a "chat model" into an agent system that actually gets things done for you.
Why It Calls Itself Cuttlefish
Hermes Agent does not have an official mascot. "Cuttlefish" is the name the AI co-author of this tutorial, Hermes Agent powered by GLM-5.1, gave itself while writing this guide.
A cuttlefish adapts, changes color, uses many tentacles, and is remarkably smart. Hermes Agent works the same way: multi-channel access, tool use, cross-session memory, scheduled execution, and steady evolution. You are not merely prompting a model. You are training a system that can work.
Core Difference
You: Help me organize the emails in my inbox
ChatGPT:
I can suggest a few ways to organize them,
such as sorting by sender, topic, or priority...Great at explaining, summarizing, and suggesting approaches, but you still have to do the execution yourself.
You: Help me organize the emails in my inbox
Hermes:
[Connecting to mail service]
[Read 127 unread emails]
[Grouped by topic and summarized]
Done: 3 important emails marked.It connects tools, edits files, talks to messaging platforms, and runs on schedules, turning “answers” into “execution”.
Three Learning Tracks
From installation and model setup to gateway access and automation, get a working Agent running for real.
Scenario guides built around real problems, so the Agent enters your workflow instead of staying in demo mode.
Understand the Hermes architecture, learn Skills, MCP, and sub-agent orchestration, and build your own version step by step.
Where It Lands
Pull news, weather, schedules, and todos on a timer and produce a briefing you would actually read.
Review PRs, run checks, summarize risks, and reduce repetitive manual steps.
Turn research, topic selection, drafting, polishing, and publishing prep into one connected chain.
Connect Lark, Telegram, Discord, and email so your Agent does not live only in a terminal.
Who It Is For
You want an AI assistant that can actually handle tasks for you, not just give abstract suggestions.
You are already experimenting with automation and want to connect Lark, Telegram, email, and schedules.
You are genuinely curious about agent skills, workflows, and memory systems.
You want more than out-of-the-box use and plan to build Skills, add MCP integrations, and extend your own version.
Ready to Begin