Hermes Agent

Hermes Agent is Nous Research's open-source CLI coding agent (MIT license) built around a simple but unusual idea: an orchestrator that runs above other coding agents. Rather than replacing Claude Code or Codex CLI, Hermes delegates work to them while keeping ownership of session lifecycle, task tracking, and cross-session memory. It runs its own reasoning loop with its own context window, memory store, and skill format, and keeps a structured store of project layout, conventions, and in-flight work — restart the terminal and it remembers. When it can't complete a task, its self-improvement loop writes a new SKILL.md to address the gap and uses it immediately. It also runs a messaging gateway so you can talk to it from Telegram, Discord, Slack, or WhatsApp.

The honest picture: Hermes adds a layer of indirection — and latency — on top of agents that already work, so the value depends on whether persistent memory actually compounds for your workflow. The "self-improving" loop can generate skill sprawl if unsupervised. As a 2026 project in a fast-moving space, its long-term maintenance depends on Nous Research's continued investment.

pip install hermes-agent (Python 3.11+), run hermes setup to configure your LLM provider, then hermes chat to start. Optional: hermes gateway for messaging-app access.