Hermes Agent

✓ Editorially verified

Open-source multi-channel AI agent with persistent memory, scheduling, and subagent delegation from Nous Research.

Freemium· MIT-licensed self-host free; managed Portal tiers Free/Plus/Super/Ultra with creditsAgentsMulti-model (300+ via Nous Portal)

Visit website →

Best for

Pick Hermes Agent if you want one self-hostable, memory-persistent agent that follows you across every messaging app and can run scheduled, sandboxed jobs.

Skip if

Skip it if you just want a hosted ChatGPT-style web UI and have no interest in self-hosting, scripting, or wiring up messaging integrations.

Hermes Agent is Nous Research's open-source personal AI agent that lives across the messaging channels you already use, including Telegram, Discord, Slack, WhatsApp, Signal, email and a local CLI. It pairs persistent memory, natural-language task scheduling and subagent delegation with web browsing, search and vision so the same agent can pick up a thread on your phone and continue it on your laptop.

What differentiates Hermes from generic chat clients is its execution model: it runs across five backends (local, Docker, SSH, Singularity and Modal) with container hardening and namespace isolation, and it auto-generates reusable skills as it learns your projects. The codebase is MIT-licensed and you can self-host it free, but Nous also sells managed Portal tiers (Free, Plus, Super, Ultra) that bundle monthly credits and access to 300+ models behind one set of credentials. Desktop apps ship for macOS 12+, Windows 10/11 and Linux, with a terminal install for headless boxes.

It's aimed at developers and power users who want one agent that remembers context, dispatches isolated subagents, and runs scheduled jobs without being chained to a single chat UI. Python RPC scripting and the multi-backend executor make it more of an automation framework than a chatbot wrapper.

Editor's take

This is one of the more serious open-source attempts at a cross-channel personal agent, and the Nous Research pedigree plus MIT license make it worth a look over closed alternatives. The five-backend executor and subagent isolation hint at real automation ambition, but it's still a pre-1.0 project, so expect rough edges and shifting skills.

— The AI Tool Bible editorial team