Skip to content

Your first conversation

Let's go from a blank CLI to a useful result.

1. Start Hermes

bash
hermes

2. Give it a real task, not a vague question

Weak:

text
help me with files

Strong:

text
List all .md files in ~/Projects/FDE/hermes-course, sorted by
modification date. Show the 5 most recently changed with their sizes.

Hermes is an agent. It works best when you give it a concrete goal it can act on.

3. Watch it use tools

When configured with a tool-capable model, Hermes will:

  1. Decide which tool to call (e.g. run_command, read_file).
  2. Execute the action.
  3. Read the result and continue until the task is done.

You'll see tool calls in the output. This is normal and expected.

4. Iterate in the same session

text
Good. Now create a summary table: filename | size | last modified.
Save it as SUMMARY.md in the same folder.

Hermes remembers context within a session. Use follow-ups instead of restarting.

5. Try a slash command

Inside a session, Hermes supports slash commands:

text
/model          # switch model
/memory         # view stored memory
/skills         # list available skills
/help           # all commands

A first session to try (offline)

Try it

After offline setup is complete, paste this:

text
You are my local file assistant. Scan my ~/Documents folder,
find any PDF files modified in the last 30 days, and give me
a numbered list with filename and date. Do not open the files,
just list them.

This tests tool calling, file system access, and reasoning without needing the internet.

What you just learned

  • Brief Hermes like an agent, not a search box.
  • Tool calls are the agent working, not an error.
  • Iterate in-session for better results.
  • Slash commands give you quick control.

Next: the full Complete offline setup.

Personal learning notes on Hermes Agent. Not affiliated with Nous Research. Verify against official docs.