Built-in tools

Hermes ships with 60+ built-in tools across multiple toolsets. With a tool-capable local model, the agent can act on your machine, not just talk.

Core tool categories

Category	Examples	Offline?
Filesystem	read, write, edit, list files	Yes
Shell	run commands, scripts	Yes
Code	execute_code, programmatic tool calling	Yes
Delegation	spawn subagents for parallel work	Yes
Web	search, browse, extract	Needs internet
Vision	analyze images	Yes (local model)
Media	image gen, TTS	Needs API or local model

For offline setups, filesystem + shell + code tools are the workhorses.

What agentic tools look like

When you ask Hermes to edit a file, you'll see something like:

[tool: read_file] path=docs/index.md
[tool: write_file] path=docs/index.md (edited)

This is the agent working. A model without tool support will describe edits instead of making them.

Programmatic tool calling

Hermes can use execute_code to collapse multi-step pipelines into a single inference call. Useful for:

• Batch file processing.
• Data transformations.
• Generating multiple outputs at once.

Example prompt:

Write a Python script that finds all .md files in docs/ without
corresponding sidebar entries, run it, and show me the results.

Subagent delegation

For parallel work:

Spawn two subagents: one scans docs/getting-started/ for broken
internal links, the other scans docs/use-cases/. Combine results.

Each subagent runs in isolation. Good for large offline audits.

Checking available tools

/tools
/tools filesystem

Disable toolsets you don't need in config.yaml to save context tokens.

Next: Toolsets & approval.