Hermes Agent: The Complete Developer Guide
by Marudi
otherAdd to your library first to use in Claude Code
About
A practical, code-first developer guide to Hermes Agent, an open-source autonomous AI agent framework built by Nous Research. Covers installation, tool usage, memory systems, skills creation, structured output, delegation patterns, multi-platform deployment, and production best practices.
Preview
Hermes Agent: The Complete Developer Guide
Author: Marudi Version: 1.0 — June 2026 Last updated: 2026-06-14 — current to Hermes Agent v2026.6.5 and model Hermes-4.3-36B.
Load this book when:
- Installing, configuring, or upgrading Hermes Agent (the Nous Research autonomous agent framework).
- Writing a custom plugin, skill, or MCP integration for Hermes.
- Deploying Hermes as a messaging gateway, Docker/VPS service, or serverless backend.
- Building delegation hierarchies, agent swarms, or the self-evolution loop.
- Choosing or prompting a Hermes open-weight model (Hermes-2/3/4/4.3) directly via API or Ollama.
- Debugging tool-calling, ChatML formatting, structured output, context compression, or command-approval behavior.
Not in scope: training/fine-tuning Hermes models from scratch; non-Hermes agent frameworks.
Audience: Public marketplace reference. Maintainer / custodian: Marudi.
Scope of validity: Hermes Agent release scheme is date-based; this edition is current as of Hermes Agent v2026.6.5 (June 6, 2026) and Hermes 4.3-36B (the latest model, December 2025). Refresh cadence: review every 4–6 weeks — the framework ships multiple releases per month. (Note: releases carry both a date-based version (primary, e.g. v2026.6.5) and a semver alias (e.g. v0.16.0); secondary sources sometimes cite only the semver. Both refer to the same artifact. Source: https://github.com/NousResearch/hermes-agent/releases/tag/v2026.6.5)
Preface
Most AI agents are sophisticated command-line tools. You give them a task, they complete it, and you start fresh next session. Hermes is different.
Hermes Agent — built by Nous Research — is an agent that grows over time. Every complex task it completes can become a reusable skill. Every fact it discovers gets stored in persistent memory. Every session deepens its model of who you are and how you work. The longer it runs, the more capable it becomes — not because the underlying model changes, but because the agent accumulates structured, reusable knowledge around it.
This book is a practical, code-first developer guide. It covers everything from installing Hermes and making your first tool call, to writing custom plugins, deploying multi-platform gateways, building delegation hierarchies, and running the self-evolution loop. Every code sample is production-ready.
You'll find the same format throughout: working code, concise explanations, and reference tables you'll reach for again and again.
Part I — Understanding Hermes