

Hermes Agent
The self-improving AI agent built by Nous Research. It's the only agent with a built-in learning loop — it creates skills from experience, improves them during use, nudges itself to persist knowledge, searches its own past conversations, and builds a deepening model of who...
Cost / License
- Free
- Open Source (MIT)
Application typeApplication types
Platforms
- Mac
- Windows
- Linux
- Android
- Self-Hosted
- Cloudron
- Docker
- Termux
Features
Hermes Agent News & Activities
Recent News
- Fla published news article about Hermes Agent
- Fla published news article about Hermes Agent
Recent activities
- tylerlikesps4 liked Hermes Agent
aivrar added Hermes Agent as alternative to Portable Hermes Agent- TBayAreaPat replied to a comment / review on Hermes Agent
- TBayAreaPat replied to a comment / review on Hermes Agent
- TBayAreaPat replied to a comment / review on Hermes Agent
I've run Hermes Agent in production since May 2026, and it currently runs my business. Not a weekend project or a demo. It manages my vault, files my records, runs my scheduled jobs, and does work I used to do by hand. I'm at a few thousand sessions on it and I stopped counting the hours it saves.
Why it stuck when other agent tools didn't: the memory persists. Most tools forget everything when the session ends, so you spend the first ten minutes of every conversation re-explaining your own...
- Christian Sass added Dark Mode as a feature to Hermes Agent
- Christian Sass added Support for MarkDown as a feature to Hermes Agent
- Christian Sass liked Hermes Agent
maidmor added Hermes Agent as alternative to POCKETCORP









Comments and Reviews
I've run Hermes Agent in production since May 2026, and it currently runs my business. Not a weekend project or a demo. It manages my vault, files my records, runs my scheduled jobs, and does work I used to do by hand. I'm at a few thousand sessions on it and I stopped counting the hours it saves.
Why it stuck when other agent tools didn't: the memory persists. Most tools forget everything when the session ends, so you spend the first ten minutes of every conversation re-explaining your own context. Hermes keeps it, and it gets noticeably sharper the longer you run it. That single difference is most of why I stayed.
What I actually use it for: one gateway across every surface, terminal, Discord, Telegram, email, same agent and same memory wherever I am. Scheduled work in plain language, running unattended while I sleep. Delegation into subagents so a long conversation doesn't bloat. Terminal backends, local and containerized, and pointing at remote machines. And model flexibility. I've changed models more than once without touching my setup, and no provider lock-in mattered to me more than I expected.
Where it's genuinely hard, and you should know before you start: it's terminal-first, so the power is in the CLI and it will fight you at first if you hate a terminal. Configuration is real work, extensible to the point where you can build almost anything, which means you can build a mess. Expect a learning weekend, not an afternoon. Model costs are yours unless you use their hosted tiers, and agents burn tokens faster than chat does. And it's a fast-moving project, good for features and occasionally annoying for stability.
Who it's right for: anyone who wants an agent that actually persists and improves, and who will learn one tool properly instead of trying ten shallowly. Self-hosters, developers, anyone running real workloads rather than experimenting.
Who it's wrong for: anyone who wants a chat window. This is infrastructure, not a chatbot.
I run it on Linux daily against my own hardware. Disclosure: I'm a heavy user and I build my own tooling on top of it, so I have a stake in it being good. I'm not affiliated with Nous Research. I tried several agent tools before this one and this is the one that actually replaced my workflow.
Nous Research is a privately held company. Founders / leadership Jeffrey Quesnelle – CEO and co-founder Headquarters listed as New York, NY.
Hermes Agent — Agent with memory and reusable skills Names of other self-improving agents and related frameworks: Agent Zero --General-purpose assistant retains memories Letta Code— Terminal agent for coding retains memories OpenClaw--Self-hosted assistant improves skills AutoSkill — Automatic reusable-skill development. MemSkill — Memory and skill improvement. EvoAgentX — Agent workflow optimization. HyperAgents — Research into agents modifying themselves and their improvement