AI Coding Hub

Explainer

What Is Hermes Agent?

Hermes Agent is an open-source, self-improving AI agent built by Nous Research. Rather than a single-shot coding tool, it is designed to run continuously — it learns skills from experience, keeps a curated memory, and can work from your terminal, a cron schedule, or chat apps like Telegram, Discord, and Slack.

By DK, Editor  ·  Last verified: 2026-06-20  ·  How we test  ·  Facts only — no hands-on score

01 / Definition

Hermes Agent is described by Nous Research as "the self-improving AI agent" — it creates skills from experience and improves them during use through a built-in learning loop. It is open source under the MIT license, and the current release is v0.17.0 (build v2026.6.19), published on 2026-06-19.

For developers, the relevant part is that Hermes is a general autonomous agent you run on your own machine or server, not an IDE plugin. It exposes a full terminal UI, can spawn isolated subagents for parallel work, run scheduled automations, and operate across multiple execution backends — so coding tasks are one of many things it can carry out, alongside research, automation, and messaging workflows.

02 / What it does

  • Self-improving skills + curated memory

    Hermes creates skills from real experience and refines them as it works, backed by an agent-curated memory with periodic nudges. The longer it runs, the more it accumulates — the core idea behind the "self-improving" framing.

  • Subagents and scheduled automations

    It can spawn isolated subagents to run parallel workstreams, and it ships a built-in cron scheduler for automations that fire on a schedule — useful for recurring jobs without an external orchestrator.

  • Full TUI + chat gateways

    Work from a terminal UI with multiline editing and slash-command autocomplete, or reach the same agent through Telegram, Discord, Slack, WhatsApp, Signal, and email gateways.

  • Multiple execution backends

    Hermes runs across several terminal backends — local, Docker, SSH, Singularity, Modal, and Daytona — so you can place the agent close to where the work needs to happen.

03 / How it works

Hermes runs as a long-lived agent rather than a per-session tool: a learning loop turns experience into reusable skills, an agent-curated memory persists what it learns, and a built-in scheduler lets it act on a cron cadence. The same agent core is reachable from a TUI, from chat gateways, and programmatically, and it can fan work out to isolated subagents across local, Docker, SSH, Modal, Daytona, and other backends. The one-line installer provisions its dependencies (Python 3.11, Node.js, ripgrep, ffmpeg).

[screenshot: Hermes Agent TUI running in a terminal — showing the slash-command autocomplete and an active subagent / scheduled task. Dark theme, 2x.]

Hermes Agent vs Claude Code in one line

Claude Code is a coding-specific terminal agent for editing a codebase; Hermes Agent is a general, self-improving agent with persistent memory that lives on your server and works across chat platforms — coding is one of many jobs it handles.

04 / Who it's for

  • Developers who want a persistent, self-hosted agent that accumulates skills over time rather than starting fresh each session.
  • People who want to reach an agent from chat apps (Telegram, Discord, Slack) or trigger it on a schedule, not only from an IDE.
  • Tinkerers comfortable installing and running an open-source agent on their own machine or server and giving it real execution backends.
  • Anyone evaluating open-source autonomous agents alongside coding-specific tools like Claude Code or OpenCode.

05 / Getting started

Hermes installs with a single shell command on Linux, macOS, WSL2, or Termux (a PowerShell one-liner covers Windows), then you launch it by running hermes. Because the project is moving fast — v0.17.0 shipped on 2026-06-19 — confirm the current install command and capabilities on the official Nous Research repository before you rely on them. A dedicated setup guide is on the way.

Our AI agents

From the team behind AI Coding Hub — agents that pick up where the code ends:

FAQ

Who makes Hermes Agent?
Hermes Agent is built by Nous Research and published as open source under the MIT license. The official source repository is github.com/NousResearch/hermes-agent. Note that several third-party sites and "awesome-hermes" lists exist — always verify commands and claims against the official Nous Research repo.
Is Hermes Agent a coding agent?
Not exclusively. Nous Research positions it as a general, self-improving AI agent rather than a coding-specific tool. It can perform coding work — it has a terminal UI, can run commands, and spawns subagents across execution backends — but it is also built for research, automation, and messaging workflows. If you only want to edit a codebase from your terminal, a coding-specific agent like Claude Code or OpenCode may be a closer fit.
Is Hermes Agent free and open source?
Yes. Hermes Agent is open source under the MIT license, so you can read, run, and modify it. As with any agent, running it incurs whatever model/provider and infrastructure costs you configure — confirm those on the official repo and your provider.
How do I install Hermes Agent?
The official repo lists a one-line installer: on Linux, macOS, WSL2, or Termux you run a curl | bash command from hermes-agent.nousresearch.com, and on Windows a PowerShell one-liner; then you start it with hermes. Because the project updates frequently, confirm the current command on github.com/NousResearch/hermes-agent before running it.
What can Hermes Agent connect to?
Beyond the terminal, Hermes can be reached through chat gateways including Telegram, Discord, Slack, WhatsApp, Signal, and email, and it can execute across multiple backends (local, Docker, SSH, Singularity, Modal, Daytona). It also has a built-in cron scheduler for recurring automations.

Sources: github.com/NousResearch/hermes-agent