Harness reference

Hermes harness reference

Nous Research documents Hermes as a self-improving autonomous agent with CLI usage, toolsets, a skills system, persistent memory, and MCP integration. On HarnessRouter it runs any frontier model.

Identity

HarnessRouter status: Available
HarnessRouter ID: hermes
Provider: Nous Research
Category: Autonomous agent
Last checked: 2026-07-23

Confirmed notes

These notes are intentionally narrow. They are derived from linked official sources and should not be expanded by implication.

Hermes works on a real project workspace with shell and file access, completing tasks end to end.
It builds a persistent memory and skill library from what it learns, getting more capable the longer it runs, and supports MCP servers.
On HarnessRouter, Hermes is multi-model: it routes to any configured frontier model (the default is gpt-5.5) and ships with the same `docx`, `pdf`, `pptx`, and `xlsx` document skills.

Integration note

Use the `hermes` harness ID for open-ended, long-running work, or when the task should run on a non-OpenAI, non-Anthropic frontier model.

Run payload

await hr.run({
  harness: "hermes",
  task: "Describe the product work the user asked for"
});

Keep this backend-facing.

Users should choose product actions. Your server maps those actions to allowed harness IDs.

Sources

Read the official pages before making capability claims in product UI.

Back to harness catalog