Hermes Agent learns from every task you give it — 9K GitHub stars
Nous Research released Hermes Agent, a self-improving AI that remembers past conversations, creates its own skills, and runs on a $5 server. 9K stars on GitHub.
Most AI assistants forget everything the moment you close the chat window. Hermes Agent by Nous Research is different — it remembers what you taught it, creates its own skills, and gets better the more you use it. It just hit 8,800 GitHub stars with nearly 5,000 gained this week alone.
An AI That Actually Gets Smarter Over Time
The key difference is the closed learning loop. When Hermes completes a complex task, it doesn't just give you the result and forget — it analyzes what it did, packages that into a reusable skill, and stores it for next time. The more you use it, the faster and better it gets at the things you care about.
It also maintains persistent memory across sessions. Ask it about something you discussed last week? It remembers. Tell it your preferences once? It applies them every time. This isn't just chat history — it's a curated, searchable memory system with full-text search and AI-powered summarization.
Works Everywhere You Already Are
40+ built-in tools — File management, web browsing, code execution, data analysis, and more.
6 messaging platforms — Use it from Telegram, Discord, Slack, WhatsApp, Signal, or email. Start a conversation on your phone, continue on your laptop.
Scheduled tasks — Set up automated jobs using natural language: "Every Monday morning, check my inbox and summarize important emails."
Voice memos — Send a voice message and it transcribes and acts on it.
Any AI Model, No Lock-in
Unlike most AI tools that force you into one provider, Hermes lets you switch AI models with a single command. It works with:
Nous Portal, OpenRouter (200+ models), OpenAI, Anthropic, and any custom endpoint. Switch between them by typing hermes model — no code changes needed.
Runs on Almost Anything
Hermes can run on a $5/month VPS (virtual private server — a rented computer in the cloud), on your own machine, or on serverless infrastructure that only charges when you're actually using it. The serverless option automatically "hibernates" when idle, keeping costs near zero.
It also supports parallel subagents — spawning isolated copies of itself to handle multiple tasks simultaneously. Need to research three topics at once? Hermes spins up three agents that work in parallel and report back.
Who This Is For
Power users who want a personal AI assistant — If you're tired of re-explaining your preferences every time you open ChatGPT, Hermes remembers and adapts.
Teams wanting shared AI infrastructure — Deploy one Hermes instance and access it from Slack, Discord, or Telegram. Everyone gets the same trained assistant.
Developers experimenting with AI agents — The research-ready features (batch trajectory generation, RL training integration) make it a serious platform for agent development.
Install in One Command
curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash
source ~/.bashrc
hermesWorks on Linux, macOS, and Windows (via WSL2). Run hermes setup for the full configuration wizard, or hermes model to pick your AI provider.
The project is MIT licensed and backed by Nous Research, one of the most respected open-source AI labs. Skills can be shared through agentskills.io, a community hub for reusable agent abilities.
Related Content — Get Started with Easy Claude Code | Free Learning Guides | More AI News
Stay updated on AI news
Simple explanations of the latest AI developments