What Is Hermes Agent? Memory, Skills, and What It Costs to Run

Hermes Agent is Nous Research’s open-source, MIT-licensed autonomous AI agent, publicly released in late February 2026. You self-host it on a laptop, a VPS or serverless infrastructure and interact with it from the CLI, a desktop app, or 20+ chat platforms including Telegram, Discord and Slack. Its defining feature is a closed learning loop: persistent memory across sessions plus skills the agent writes and refines itself.

The software is free and MIT licensed. You pay for model inference and for whatever infrastructure you run it on. Nous Portal offers optional paid tiers ($20, $100 and $200 per month as of September 2026) that bundle model credits and hosted tools, but you can use your own provider instead.

Two mechanisms. Memory: the agent curates MEMORY.md and USER.md inside a character budget, injected into the system prompt at each session start, with full-text search over past sessions for anything older. Skills: when it works something out, it can write a SKILL.md file with procedure and pitfalls, then load it on demand later through tiered disclosure so unused skills cost almost no context.

Yes. Hermes works with Nous Portal, OpenRouter, OpenAI or any compatible endpoint, so pointing it at an OpenAI-compatible gateway is a configuration change. That gives you per-team budgets, model routing, guardrails and unified traces without modifying the agent.

Yes. TrueFoundry runs in your VPC, on-prem, air-gapped, or hybrid, so prompts and responses never leave your domain even as you route across many providers.

Yes. It includes an MCP Gateway, an Agent Gateway, and an MCP and Agents Registry with tool-level access control, so agents from LangGraph, CrewAI, AutoGen, or a custom framework can be governed centrally.