The self-improving AI agent
Hermes Agent is Nous Research's self-improving agent: a built-in learning loop that creates skills from experience, improves them in use, and builds a model of you across sessions.
Updated 2026-08-15. Same-day snapshot of the official repo.
How to Get Started with Hermes Agent
3 steps to launch the self-improving, memory-driven agent.
Install
The official installer provisions all dependencies.
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bashPick a model
Switch any provider with `hermes model` — no code changes.
hermes modelEnter the TUI
Run `hermes` for the TUI, assign tasks and let it build memory and skills in use.
hermesPitfalls
Heavy installer
The installer provisions uv, Python 3.11, Node.js, ripgrep, ffmpeg (plus portable Git Bash on Windows); first install is slow.
Autonomy needs guardrails
Auto-created skills, cron jobs and remote triggers mean unattended actions; run on a local backend first before connecting remote channels.
Not a coding agent
For IDE completion or terminal coding, OpenCode or Pi fit better; Hermes is autonomy + memory + assistant. Do not mix expectations.
Beside OpenCode & OpenClaw
Based on official docs and licenses. Hermes's edge is the closed learning loop — for cases where the agent should know you better over time.
| Tool | Positioning | Learning & memory | Backends | License |
|---|---|---|---|---|
| Hermes Agent | Self-improving autonomous agent | Built-in learning loop + persistent memory | Local / Docker / SSH / Modal / Daytona / Vercel Sandbox | MIT |
| OpenCode | Open coding agent | No built-in learning loop | Local CLI / TUI / desktop / IDE | MIT |
| OpenClaw | Personal assistant (multi-channel) | Persistent memory (assistant use) | Self-hosted local Gateway | MIT |
Official Directory
Official
- Official site
Official site for Hermes Agent and Desktop.
- GitHub repository
Source and multilingual README. MIT.
Docs
- Documentation
Installation, providers, skills and memory.
Community
- Discord
Official Nous Research Discord.
Ecosystem & Standards
- Nous Research
The research lab behind Hermes Agent.
- AgentSkills open standard
The open skills standard Hermes is compatible with.
Latest Updates
2026-08-14 · GitHub · NousResearch/hermes-agent
Hermes Agent ships v2026.8.3 (v0.20.0) with stronger learning loop & serverlessNous Research keeps iterating the self-improving agent: skills created from experience, self-improvement in use, cross-session memory, and near-zero idle cost via Modal/Daytona.
LaoJin's Pick
FAQ
- What is Hermes Agent?
- Hermes Agent is Nous Research's self-improving AI agent. Its signature is a built-in learning loop: it creates skills from experience, improves them in use, nudges itself to persist knowledge, and builds a model of you across sessions.
- How do I install Hermes Agent?
- Linux/macOS/WSL2/Termux: `curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash`; Windows uses a PowerShell script. The installer provisions uv, Python 3.11, Node.js and more.
- How does Hermes differ from OpenCode and Pi?
- OpenCode and Pi are coding agents; Hermes is an autonomous, self-learning agent — memory, self-built skills, scheduled automation, and cloud-VM operation controlled from Telegram.
- Which models does Hermes support?
- Any model: Nous Portal, OpenRouter, OpenAI, your own endpoint and more; switch with `hermes model`, no lock-in.
Need to deploy a team-level Agent system?
Open-source frameworks are free, but turning a runtime into an enterprise asset requires access control, RBAC sandboxes, and private RAG integrations. Laojin Global provides commercial-grade architecture and deployment support.
Cite This Hub
We encourage you to capture this hub into your knowledge base, research, or feed it straight to an AI agent under CC BY 4.0.
How this hub stays current
Version and links live in src/content/hermes-agent.ts; star snapshots sync via GitHub API.